avg_L3/checkpoint-40000

avg_L3/checkpoint-40000/config.json

@@ -0,0 +1,233 @@
+{
+  "_name_or_path": "",
+  "action_dit_size": "DiT-B",
+  "action_future_slots": null,
+  "action_head_model_type": "dit",
+  "action_horizon": 16,
+  "action_loss_weight": 1.0,
+  "action_text_kv": false,
+  "add_cross_attention": false,
+  "add_pos_embed": true,
+  "apply_sincos_state_encoding": false,
+  "architectures": [
+    "WAMDiT4DiT"
+  ],
+  "attn_dropout": 0.2,
+  "backbone_embedding_dim": 3072,
+  "backbone_trainable_params_fp32": true,
+  "backbone_type": "wan22_diffusers",
+  "bad_words_ids": null,
+  "begin_suppress_tokens": null,
+  "bos_token_id": null,
+  "chunk_size_feed_forward": 0,
+  "color_jitter_params": null,
+  "conditional_frame_timestep": 0.0001,
+  "cosmos_model_path": "nvidia/Cosmos-Predict2.5-2B",
+  "cosmos_model_revision": "diffusers/base/pre-trained",
+  "crop_fraction": 0.95,
+  "cross_attention_hidden_size": null,
+  "decoder_start_token_id": null,
+  "diffusion_model_cfg": {
+    "attention_head_dim": 64,
+    "dropout": 0.2,
+    "final_dropout": true,
+    "interleave_self_attention": true,
+    "norm_type": "ada_norm",
+    "num_attention_heads": 12,
+    "num_layers": 16,
+    "output_dim": 2560,
+    "positional_embeddings": null
+  },
+  "diversity_penalty": 0.0,
+  "do_sample": false,
+  "dtype": "bfloat16",
+  "early_stopping": false,
+  "encoder_no_repeat_ngram_size": 0,
+  "eos_token_id": null,
+  "exponential_decay_length_penalty": null,
+  "fadistill_target_tau": 1.0,
+  "fadistill_teacher_nstep": 4,
+  "fadistill_teacher_traj": true,
+  "feature_extraction_aggregate_all": false,
+  "feature_extraction_aggregate_layer_window": 0,
+  "feature_extraction_layer": 18,
+  "finetuning_task": null,
+  "fm_high_sigma_min": 0.98,
+  "fm_high_sigma_ratio": null,
+  "fm_loss_weight": 1.0,
+  "fm_time_distribution": "uniform",
+  "forced_bos_token_id": null,
+  "forced_eos_token_id": null,
+  "formalize_language": true,
+  "frame_downsample_factor": 2,
+  "freeze_modules_pattern": "text_encoder,vae",
+  "freeze_step_weight_predictor": false,
+  "freeze_whole_backbone": false,
+  "future_loss_type": "flow_matching",
+  "future_num_inference_steps": 1,
+  "hidden_size": 2560,
+  "history_num_clean_frames": 1,
+  "history_obs_noise_timestep": 0.0,
+  "id2label": {
+    "0": "LABEL_0",
+    "1": "LABEL_1"
+  },
+  "image_crop_size": null,
+  "image_resize_backend": "cv2_linear",
+  "image_target_size": null,
+  "input_embedding_dim": 768,
+  "is_decoder": false,
+  "is_encoder_decoder": false,
+  "label2id": {
+    "LABEL_0": 0,
+    "LABEL_1": 1
+  },
+  "length_penalty": 1.0,
+  "load_bf16": true,
+  "max_action_dim": 32,
+  "max_length": 20,
+  "max_num_embodiments": 36,
+  "max_seq_len": 1024,
+  "max_state_dim": 64,
+  "meta_queries_mode": "meta_only",
+  "min_length": 0,
+  "model_dtype": "bfloat16",
+  "model_type": "WAM_DIT4DIT",
+  "multiview_layout": "tmosaic",
+  "n_meta_queries": 8,
+  "n_tune_layers_below_feature_extraction": 0,
+  "new_embodiment_train_ratio": 0,
+  "no_repeat_ngram_size": 0,
+  "noise_beta_alpha": 1.5,
+  "noise_beta_beta": 1.0,
+  "noise_s": 0.999,
+  "nstep_agg": false,
+  "nstep_agg_layers": 4,
+  "nstep_base_residual": false,
+  "nstep_base_residual_gate": "layerscale",
+  "nstep_certainty_arch": "mlp",
+  "nstep_certainty_ckpt": "",
+  "nstep_certainty_decalib": false,
+  "nstep_certainty_drop_cond": false,
+  "nstep_certainty_fusion": false,
+  "nstep_certainty_hidden": 1024,
+  "nstep_certainty_layers": 2,
+  "nstep_certainty_step_bias": false,
+  "nstep_certainty_tau": 1.0,
+  "nstep_compress_k": 64,
+  "nstep_compress_n_frame": 3,
+  "nstep_compress_query": false,
+  "nstep_latent_cond": false,
+  "nstep_n": 2,
+  "nstep_noise_ensemble": false,
+  "nstep_per_step_gated": false,
+  "nstep_query_len": 588,
+  "nstep_readout_exclude_step0": false,
+  "nstep_residual_warmup": false,
+  "nstep_residual_warmup_full": 20000,
+  "nstep_residual_warmup_start": 10000,
+  "nstep_step_dropout": false,
+  "nstep_step_embed": true,
+  "nstep_train_integration": false,
+  "num_beam_groups": 1,
+  "num_beams": 1,
+  "num_frames_in": 25,
+  "num_frames_out": 41,
+  "num_inference_timesteps": 4,
+  "num_return_sequences": 1,
+  "num_timestep_buckets": 1000,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "output_scores": false,
+  "pad_token_id": null,
+  "prefix": null,
+  "problem_type": null,
+  "pruned_heads": {},
+  "random_rotation_angle": null,
+  "remove_invalid_values": false,
+  "repeated_diffusion_steps": 4,
+  "repetition_penalty": 1.0,
+  "reproject_vision": false,
+  "return_dict": true,
+  "return_dict_in_generate": false,
+  "selfflow_ema_decay": 0.9999,
+  "selfflow_mask_ratio": 0.1,
+  "selfflow_rep": true,
+  "selfflow_rep_layer": -1,
+  "selfflow_rep_layer_student": -1,
+  "selfflow_rep_layer_teacher": -1,
+  "selfflow_rep_weight": 0.8,
+  "selfflow_teacher_noise": 0.0,
+  "sep_token_id": null,
+  "shortest_image_edge": null,
+  "state_additive_noise_scale": 0.0,
+  "state_dropout_prob": 0.0,
+  "step_weight_exclude_current_tokens": false,
+  "step_weight_predictor_ckpt": "",
+  "step_weight_predictor_heads": 16,
+  "step_weight_predictor_layers": 4,
+  "step_weight_sigmas": "0.995,0.9631,0.594,0.0099",
+  "step_weight_target_tau": 0.7,
+  "suppress_tokens": null,
+  "task_specific_params": null,
+  "temperature": 1.0,
+  "tf_legacy_loss": false,
+  "tie_encoder_decoder": false,
+  "tie_word_embeddings": true,
+  "tokenizer_class": null,
+  "top_k": 50,
+  "top_p": 1.0,
+  "torchscript": false,
+  "training_mode": "joint",
+  "transformers_version": "4.57.0",
+  "truncate_above_feature_extraction": false,
+  "tune_crossattn_proj": false,
+  "tune_diffusion_model": true,
+  "tune_layers_above_feature_extraction": true,
+  "tune_llm": false,
+  "tune_patch_embed": false,
+  "tune_projector": true,
+  "tune_time_embed": false,
+  "tune_top_llm_layers": 4,
+  "tune_vgln": true,
+  "tune_visual": false,
+  "typical_p": 1.0,
+  "use_albumentations_transforms": true,
+  "use_bfloat16": false,
+  "use_meta_queries": false,
+  "use_meta_queries_in_fm": false,
+  "use_noise_frames_in_extraction": true,
+  "use_percentiles": false,
+  "use_relative_action": true,
+  "use_step_weight_fusion": false,
+  "use_vgln": false,
+  "use_vlm_meta": false,
+  "video_context_exceeds_action_horizon": false,
+  "video_resolution": 256,
+  "video_train_crop": false,
+  "vl_adapter_num_layers": 0,
+  "wan_block_causal": false,
+  "wan_context_pool": "avg",
+  "wan_context_pool_layer": 3,
+  "wan_context_pool_n_cond_lat": 4,
+  "wan_context_pool_xattn_ffn_dim": 0,
+  "wan_context_pool_xattn_heads": 0,
+  "wan_diffusers_model_path": "Wan-AI/Wan2.2-TI2V-5B-Diffusers",
+  "wan_fadistill_action_from_teacher": false,
+  "wan_fadistill_align_weight": 0.01,
+  "wan_fadistill_single_step_index": -1,
+  "wan_fadistill_teacher_shift": 5.0,
+  "wan_fadistill_teacher_stepper": "euler",
+  "wan_feature_extraction_layer": 18,
+  "wan_fm_train_shift": 5.0,
+  "wan_learnable_token": false,
+  "wan_learnable_token_count": 64,
+  "wan_learnable_token_layers": 2,
+  "wan_learnable_token_queries": 192,
+  "wan_model_path": "Wan-AI/Wan2.2-TI2V-5B",
+  "wan_text_pad_zero": true,
+  "wan_tokenizer_model_id": "Wan-AI/Wan2.1-T2V-1.3B",
+  "wan_v2v_fm_mix": false,
+  "wan_v2v_student": true,
+  "wan_v2v_teacher": true
+}

avg_L3/checkpoint-40000/embodiment_id.json

@@ -0,0 +1,39 @@
+{
+  "new_embodiment": 0,
+  "fractal20220817_data": 1,
+  "kuka": 2,
+  "bridge_orig": 3,
+  "taco_play": 4,
+  "jaco_play": 5,
+  "berkeley_cable_routing": 6,
+  "roboturk": 7,
+  "viola": 8,
+  "berkeley_autolab_ur5": 9,
+  "toto": 10,
+  "language_table": 11,
+  "stanford_hydra_dataset_converted_externally_to_rlds": 12,
+  "austin_buds_dataset_converted_externally_to_rlds": 13,
+  "nyu_franka_play_dataset_converted_externally_to_rlds": 14,
+  "furniture_bench_dataset_converted_externally_to_rlds": 15,
+  "ucsd_kitchen_dataset_converted_externally_to_rlds": 16,
+  "austin_sailor_dataset_converted_externally_to_rlds": 17,
+  "austin_sirius_dataset_converted_externally_to_rlds": 18,
+  "dlr_edan_shared_control_converted_externally_to_rlds": 19,
+  "iamlab_cmu_pickup_insert_converted_externally_to_rlds": 20,
+  "utaustin_mutex": 21,
+  "berkeley_fanuc_manipulation": 22,
+  "cmu_stretch": 23,
+  "bc_z": 24,
+  "fmb_dataset": 25,
+  "dobbe": 26,
+  "droid": 27,
+  "agibot_dexhand": 28,
+  "agibot_gripper": 29,
+  "galaxea": 30,
+  "humanoid_everyday_g1": 31,
+  "humanoid_everyday_h1": 32,
+  "action_net": 33,
+  "neural_gr1": 34,
+  "simulation_gr1": 35,
+  "libero_panda": 2
+}

avg_L3/checkpoint-40000/experiment_cfg/conf.yaml

@@ -0,0 +1,359 @@
+load_config_path: null
+model:
+  model_type: WAM_DIT4DIT
+  model_dtype: bfloat16
+  backbone_type: wan22_diffusers
+  wan_model_path: Wan-AI/Wan2.2-TI2V-5B
+  wan_diffusers_model_path: Wan-AI/Wan2.2-TI2V-5B-Diffusers
+  wan_tokenizer_model_id: Wan-AI/Wan2.1-T2V-1.3B
+  wan_feature_extraction_layer: 18
+  cosmos_model_path: nvidia/Cosmos-Predict2.5-2B
+  cosmos_model_revision: diffusers/base/pre-trained
+  backbone_embedding_dim: 2048
+  num_frames_in: 25
+  num_frames_out: 41
+  frame_downsample_factor: 2
+  video_resolution: 256
+  feature_extraction_layer: 18
+  feature_extraction_aggregate_all: false
+  feature_extraction_aggregate_layer_window: 0
+  truncate_above_feature_extraction: false
+  vl_adapter_num_layers: 0
+  tune_layers_above_feature_extraction: true
+  n_tune_layers_below_feature_extraction: 0
+  tune_patch_embed: false
+  tune_time_embed: false
+  tune_crossattn_proj: false
+  freeze_modules_pattern: text_encoder,vae
+  freeze_whole_backbone: false
+  fm_loss_weight: 1.0
+  action_loss_weight: 1.0
+  conditional_frame_timestep: 0.0001
+  history_obs_noise_timestep: 0.0
+  history_num_clean_frames: 1
+  selfflow_rep: true
+  selfflow_rep_layer: -1
+  selfflow_rep_layer_student: -1
+  selfflow_rep_layer_teacher: -1
+  selfflow_mask_ratio: 0.1
+  selfflow_ema_decay: 0.9999
+  selfflow_rep_weight: 0.8
+  selfflow_teacher_noise: 0.0
+  fadistill_teacher_traj: true
+  fadistill_target_tau: 1.0
+  fadistill_teacher_nstep: 4
+  wan_fadistill_teacher_shift: 5.0
+  wan_fadistill_single_step_index: -1
+  wan_fadistill_align_weight: 0.01
+  wan_fadistill_teacher_stepper: euler
+  wan_fadistill_action_from_teacher: false
+  wan_v2v_teacher: true
+  wan_v2v_student: true
+  wan_v2v_fm_mix: false
+  wan_block_causal: false
+  wan_learnable_token: false
+  wan_learnable_token_count: 64
+  wan_learnable_token_queries: 192
+  wan_learnable_token_layers: 2
+  wan_context_pool: avg
+  wan_context_pool_layer: 3
+  wan_context_pool_n_cond_lat: 4
+  wan_context_pool_xattn_heads: 0
+  wan_context_pool_xattn_ffn_dim: 0
+  wan_fm_train_shift: 5.0
+  wan_text_pad_zero: true
+  nstep_agg: false
+  nstep_n: 2
+  nstep_query_len: 588
+  nstep_step_embed: true
+  nstep_agg_layers: 4
+  nstep_train_integration: false
+  nstep_noise_ensemble: false
+  nstep_base_residual: false
+  nstep_base_residual_gate: layerscale
+  nstep_compress_query: false
+  nstep_compress_k: 64
+  nstep_compress_n_frame: 3
+  nstep_readout_exclude_step0: false
+  nstep_per_step_gated: false
+  nstep_step_dropout: false
+  nstep_residual_warmup: false
+  nstep_residual_warmup_start: 10000
+  nstep_residual_warmup_full: 20000
+  nstep_certainty_fusion: false
+  nstep_certainty_ckpt: ''
+  nstep_certainty_tau: 1.0
+  nstep_certainty_drop_cond: false
+  nstep_certainty_arch: mlp
+  nstep_certainty_hidden: 1024
+  nstep_certainty_layers: 2
+  nstep_certainty_decalib: false
+  nstep_certainty_step_bias: false
+  nstep_latent_cond: false
+  use_step_weight_fusion: false
+  step_weight_target_tau: 0.7
+  step_weight_sigmas: 0.995,0.9631,0.594,0.0099
+  step_weight_predictor_layers: 4
+  step_weight_predictor_heads: 16
+  step_weight_predictor_ckpt: ''
+  freeze_step_weight_predictor: false
+  step_weight_exclude_current_tokens: false
+  multiview_layout: tmosaic
+  fm_time_distribution: uniform
+  fm_high_sigma_ratio: null
+  fm_high_sigma_min: 0.98
+  future_loss_type: flow_matching
+  future_num_inference_steps: 1
+  use_noise_frames_in_extraction: true
+  use_meta_queries: false
+  n_meta_queries: 8
+  meta_queries_mode: meta_only
+  use_meta_queries_in_fm: false
+  use_vlm_meta: false
+  use_vgln: false
+  tune_vgln: true
+  action_text_kv: false
+  state_dropout_prob: 0.0
+  state_additive_noise_scale: 0.0
+  training_mode: joint
+  action_head_model_type: dit
+  max_state_dim: 64
+  apply_sincos_state_encoding: false
+  use_percentiles: false
+  video_train_crop: false
+  image_resize_backend: cv2_linear
+  max_action_dim: 32
+  action_horizon: 16
+  video_context_exceeds_action_horizon: false
+  action_future_slots: null
+  hidden_size: 2560
+  input_embedding_dim: 768
+  repeated_diffusion_steps: 4
+  add_pos_embed: true
+  max_seq_len: 1024
+  attn_dropout: 0.2
+  action_dit_size: DiT-B
+  diffusion_model_cfg:
+    num_attention_heads: 12
+    attention_head_dim: 64
+    num_layers: 16
+    output_dim: 2560
+    dropout: 0.2
+    final_dropout: true
+    interleave_self_attention: true
+    norm_type: ada_norm
+    positional_embeddings: null
+  num_inference_timesteps: 4
+  noise_beta_alpha: 1.5
+  noise_beta_beta: 1.0
+  noise_s: 0.999
+  num_timestep_buckets: 1000
+  tune_projector: true
+  tune_diffusion_model: true
+  formalize_language: true
+  use_albumentations_transforms: true
+  crop_fraction: 0.95
+  random_rotation_angle: null
+  color_jitter_params: null
+  image_crop_size: null
+  image_target_size: null
+  shortest_image_edge: null
+  max_num_embodiments: 36
+  new_embodiment_train_ratio: 0.0
+data:
+  datasets:
+  - dataset_paths:
+    - /data/shared_dataset/robocasa_mg_gr00t_300
+    embodiment_tag: new_embodiment
+    mix_ratio: 1.0
+    dataset_type: physical_embodiment
+    val_dataset_path: null
+  modality_configs:
+    new_embodiment:
+      video:
+        delta_indices:
+        - -24
+        - -22
+        - -20
+        - -18
+        - -16
+        - -14
+        - -12
+        - -10
+        - -8
+        - -6
+        - -4
+        - -2
+        - 0
+        - 2
+        - 4
+        - 6
+        - 8
+        - 10
+        - 12
+        - 14
+        - 16
+        modality_keys:
+        - left_view
+        - right_view
+        - wrist_view
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+      state:
+        delta_indices:
+        - 0
+        modality_keys:
+        - end_effector_position_relative
+        - end_effector_rotation_relative
+        - gripper_qpos
+        - base_position
+        - base_rotation
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+      action:
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        modality_keys:
+        - end_effector_position
+        - end_effector_rotation
+        - gripper_close
+        - base_motion
+        - control_mode
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs:
+        - rep: DELTA
+          type: EEF
+          format: DEFAULT
+          state_key: null
+        - rep: DELTA
+          type: EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: DELTA
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+      language:
+        delta_indices:
+        - 0
+        modality_keys:
+        - annotation.human.action.task_description
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+  dataset_mode: sharded
+  download_cache: false
+  shard_size: 512
+  episode_sampling_rate: 0.1
+  num_shards_per_epoch: 100000
+  override_pretraining_statistics: false
+  mode: single_turn
+  random_chop: 0.0
+  mock_dataset_mode: false
+  shuffle: true
+  seed: 42
+  multiprocessing_context: fork
+  allow_padding: true
+  subsample_ratio: 1.0
+  image_crop_size:
+  - 244
+  - 244
+  image_target_size:
+  - 224
+  - 224
+  video_backend: ffmpeg
+training:
+  output_dir: /data/rlwrld-unified-checkpoints/hyeonmo/checkpoints/ctxpool_bmethod/wam_dit4dit_robocasa_kitchen_ditjoint_wan22_fadistillV2V_shift5_g001_vstu_sL3_future_tf025_sl3_latin4_ctxpoolavgL3_L18_res256_tmosaic_pd4_ga2_gb16_100k_local_jnojid
+  experiment_name: rc_bmethod_ctxpool_avg_L3_4latin_pd4ga2_4gpu
+  max_steps: 100000
+  global_batch_size: 16
+  batch_size: null
+  gradient_accumulation_steps: 2
+  learning_rate: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  min_lr: 5.0e-07
+  weight_decay: 0.0
+  warmup_ratio: 0.0
+  warmup_steps: 5000
+  max_grad_norm: 1.0
+  new_param_warmup_steps: 0
+  reset_stage2_lr_scheduler: false
+  adam_beta1: 0.9
+  adam_beta2: 0.95
+  adam_epsilon: 1.0e-08
+  lr_per_module_enabled: false
+  lr_base: 3.0e-05
+  lr_backbone: 1.0e-05
+  lr_action_head: 0.0001
+  optim: adamw_torch
+  start_from_checkpoint: null
+  tf32: true
+  fp16: false
+  bf16: true
+  eval_bf16: true
+  logging_steps: 10
+  save_steps: 10000
+  save_total_limit: 5
+  save_vl_model: false
+  upload_checkpoints: false
+  upload_every: 1000
+  upload_last_n_checkpoints: 5
+  max_concurrent_uploads: 2
+  eval_strategy: 'no'
+  eval_steps: 500
+  eval_set_split_ratio: 0.1
+  eval_num_shards_per_epoch: 8
+  eval_batch_size: 2
+  save_best_eval_metric_name: ''
+  save_best_eval_metric_greater_is_better: true
+  deepspeed_stage: 2
+  gradient_checkpointing: false
+  resume_from_checkpoint: false
+  transformers_trust_remote_code: true
+  transformers_local_files_only: false
+  transformers_cache_dir: null
+  transformers_access_token: null
+  use_ddp: false
+  ddp_bucket_cap_mb: 100
+  num_gpus: 4
+  dataloader_num_workers: 5
+  remove_unused_columns: false
+  use_wandb: true
+  wandb_project: wam-dit4dit-robocasa-ctxpool-bmethod
+  enable_profiling: false
+  max_retries: 3
+  assert_loss_less_than: null
+  add_rl_callback: false
+  enable_open_loop_eval: false
+  open_loop_eval_traj_ids:
+  - 0
+  open_loop_eval_steps_per_traj: 100
+  open_loop_eval_plot_indices: null
+max_steps: 100000
+save_steps: 10000

avg_L3/checkpoint-40000/experiment_cfg/config.yaml

@@ -0,0 +1,396 @@
+!!python/object:gr00t.configs.base_config.Config
+data: !!python/object:gr00t.configs.data.data_config.DataConfig
+  allow_padding: true
+  dataset_mode: sharded
+  datasets:
+  - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
+    dataset_paths:
+    - /data/shared_dataset/robocasa_mg_gr00t_300
+    dataset_type: physical_embodiment
+    embodiment_tag: new_embodiment
+    mix_ratio: 1.0
+    val_dataset_path: null
+  download_cache: false
+  episode_sampling_rate: 0.1
+  image_crop_size:
+  - 244
+  - 244
+  image_target_size:
+  - 224
+  - 224
+  mock_dataset_mode: false
+  modality_configs:
+    new_embodiment:
+      action: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs:
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
+          - default
+          rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
+          - delta
+          state_key: null
+          type: &id003 !!python/object/apply:gr00t.data.types.ActionType
+          - eef
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id002
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: &id005 !!python/object/apply:gr00t.data.types.ActionRepresentation
+          - absolute
+          state_key: null
+          type: &id004 !!python/object/apply:gr00t.data.types.ActionType
+          - non_eef
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id002
+          state_key: null
+          type: *id004
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id005
+          state_key: null
+          type: *id004
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        mean_std_embedding_keys: null
+        modality_keys:
+        - end_effector_position
+        - end_effector_rotation
+        - gripper_close
+        - base_motion
+        - control_mode
+        sin_cos_embedding_keys: null
+      language: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - 0
+        mean_std_embedding_keys: null
+        modality_keys:
+        - annotation.human.action.task_description
+        sin_cos_embedding_keys: null
+      state: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - 0
+        mean_std_embedding_keys: null
+        modality_keys:
+        - end_effector_position_relative
+        - end_effector_rotation_relative
+        - gripper_qpos
+        - base_position
+        - base_rotation
+        sin_cos_embedding_keys: null
+      video: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - -24
+        - -22
+        - -20
+        - -18
+        - -16
+        - -14
+        - -12
+        - -10
+        - -8
+        - -6
+        - -4
+        - -2
+        - 0
+        - 2
+        - 4
+        - 6
+        - 8
+        - 10
+        - 12
+        - 14
+        - 16
+        mean_std_embedding_keys: null
+        modality_keys:
+        - left_view
+        - right_view
+        - wrist_view
+        sin_cos_embedding_keys: null
+  mode: single_turn
+  multiprocessing_context: fork
+  num_shards_per_epoch: 100000
+  override_pretraining_statistics: false
+  random_chop: 0.0
+  seed: 42
+  shard_size: 512
+  shuffle: true
+  subsample_ratio: 1.0
+  video_backend: ffmpeg
+load_config_path: null
+model: !!python/object:gr00t.configs.model.wam_dit4dit.WAMDiT4DiTConfig
+  _attn_implementation_internal: null
+  _commit_hash: null
+  _name_or_path: ''
+  _output_attentions: false
+  _wam_sparse_checkpoint_config: false
+  action_dit_size: DiT-B
+  action_head_model_type: dit
+  action_text_kv: false
+  add_cross_attention: false
+  apply_sincos_state_encoding: false
+  architectures: null
+  backbone_trainable_params_fp32: true
+  backbone_type: wan22_diffusers
+  bad_words_ids: null
+  begin_suppress_tokens: null
+  bos_token_id: null
+  chunk_size_feed_forward: 0
+  color_jitter_params: null
+  conditional_frame_timestep: 0.0001
+  cross_attention_hidden_size: null
+  decoder_start_token_id: null
+  diffusion_model_cfg:
+    attention_head_dim: 64
+    dropout: 0.2
+    final_dropout: true
+    interleave_self_attention: true
+    norm_type: ada_norm
+    num_attention_heads: 12
+    num_layers: 16
+    output_dim: 2560
+    positional_embeddings: null
+  diversity_penalty: 0.0
+  do_sample: false
+  dtype: null
+  early_stopping: false
+  encoder_no_repeat_ngram_size: 0
+  eos_token_id: null
+  exponential_decay_length_penalty: null
+  fadistill_target_tau: 1.0
+  fadistill_teacher_nstep: 4
+  fadistill_teacher_traj: true
+  feature_extraction_aggregate_all: false
+  feature_extraction_aggregate_layer_window: 0
+  feature_extraction_layer: 18
+  finetuning_task: null
+  forced_bos_token_id: null
+  forced_eos_token_id: null
+  frame_downsample_factor: 2
+  freeze_modules_pattern: text_encoder,vae
+  freeze_step_weight_predictor: false
+  freeze_whole_backbone: false
+  history_num_clean_frames: 1
+  history_obs_noise_timestep: 0.0
+  id2label:
+    0: LABEL_0
+    1: LABEL_1
+  image_resize_backend: cv2_linear
+  input_embedding_dim: 768
+  is_decoder: false
+  is_encoder_decoder: false
+  label2id:
+    LABEL_0: 0
+    LABEL_1: 1
+  length_penalty: 1.0
+  load_bf16: true
+  max_length: 20
+  min_length: 0
+  model_type: WAM_DIT4DIT
+  multiview_layout: tmosaic
+  new_embodiment_train_ratio: 0
+  no_repeat_ngram_size: 0
+  nstep_agg: false
+  nstep_agg_layers: 4
+  nstep_base_residual: false
+  nstep_base_residual_gate: layerscale
+  nstep_certainty_arch: mlp
+  nstep_certainty_ckpt: ''
+  nstep_certainty_decalib: false
+  nstep_certainty_drop_cond: false
+  nstep_certainty_fusion: false
+  nstep_certainty_hidden: 1024
+  nstep_certainty_layers: 2
+  nstep_certainty_step_bias: false
+  nstep_certainty_tau: 1.0
+  nstep_compress_k: 64
+  nstep_compress_n_frame: 3
+  nstep_compress_query: false
+  nstep_latent_cond: false
+  nstep_n: 2
+  nstep_noise_ensemble: false
+  nstep_per_step_gated: false
+  nstep_query_len: 588
+  nstep_readout_exclude_step0: false
+  nstep_residual_warmup: false
+  nstep_residual_warmup_full: 20000
+  nstep_residual_warmup_start: 10000
+  nstep_step_dropout: false
+  nstep_step_embed: true
+  nstep_train_integration: false
+  num_beam_groups: 1
+  num_beams: 1
+  num_frames_in: 25
+  num_frames_out: 41
+  num_return_sequences: 1
+  output_hidden_states: false
+  output_scores: false
+  pad_token_id: null
+  prefix: null
+  problem_type: null
+  pruned_heads: {}
+  random_rotation_angle: null
+  remove_invalid_values: false
+  repeated_diffusion_steps: 4
+  repetition_penalty: 1.0
+  reproject_vision: false
+  return_dict: true
+  return_dict_in_generate: false
+  selfflow_ema_decay: 0.9999
+  selfflow_mask_ratio: 0.1
+  selfflow_rep: true
+  selfflow_rep_layer: -1
+  selfflow_rep_layer_student: -1
+  selfflow_rep_layer_teacher: -1
+  selfflow_rep_weight: 0.8
+  selfflow_teacher_noise: 0.0
+  sep_token_id: null
+  state_dropout_prob: 0.0
+  step_weight_exclude_current_tokens: false
+  step_weight_predictor_ckpt: ''
+  step_weight_predictor_heads: 16
+  step_weight_predictor_layers: 4
+  step_weight_sigmas: 0.995,0.9631,0.594,0.0099
+  step_weight_target_tau: 0.7
+  suppress_tokens: null
+  task_specific_params: null
+  temperature: 1.0
+  tf_legacy_loss: false
+  tie_encoder_decoder: false
+  tie_word_embeddings: true
+  tokenizer_class: null
+  top_k: 50
+  top_p: 1.0
+  torchscript: false
+  training_mode: joint
+  transformers_version: null
+  truncate_above_feature_extraction: false
+  tune_crossattn_proj: false
+  tune_diffusion_model: true
+  tune_layers_above_feature_extraction: true
+  tune_llm: false
+  tune_patch_embed: false
+  tune_projector: true
+  tune_time_embed: false
+  tune_top_llm_layers: 4
+  tune_vgln: true
+  tune_visual: false
+  typical_p: 1.0
+  use_bfloat16: false
+  use_meta_queries: false
+  use_meta_queries_in_fm: false
+  use_noise_frames_in_extraction: true
+  use_relative_action: true
+  use_step_weight_fusion: false
+  use_vgln: false
+  use_vlm_meta: false
+  video_resolution: 256
+  video_train_crop: false
+  vl_adapter_num_layers: 0
+  wan_context_pool: avg
+  wan_context_pool_layer: 3
+  wan_context_pool_n_cond_lat: 4
+  wan_context_pool_xattn_ffn_dim: 0
+  wan_context_pool_xattn_heads: 0
+  wan_fadistill_action_from_teacher: false
+  wan_fadistill_align_weight: 0.01
+  wan_fadistill_single_step_index: -1
+  wan_fadistill_teacher_shift: 5.0
+  wan_fadistill_teacher_stepper: euler
+  wan_feature_extraction_layer: 18
+  wan_fm_train_shift: 5.0
+  wan_learnable_token_count: 64
+  wan_learnable_token_layers: 2
+  wan_learnable_token_queries: 192
+  wan_model_path: Wan-AI/Wan2.2-TI2V-5B
+  wan_text_pad_zero: true
+  wan_v2v_student: true
+  wan_v2v_teacher: true
+training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
+  adam_beta1: 0.9
+  adam_beta2: 0.95
+  adam_epsilon: 1.0e-08
+  add_rl_callback: false
+  assert_loss_less_than: null
+  batch_size: null
+  bf16: true
+  dataloader_num_workers: 5
+  ddp_bucket_cap_mb: 100
+  deepspeed_stage: 2
+  enable_open_loop_eval: false
+  enable_profiling: false
+  eval_batch_size: 2
+  eval_bf16: true
+  eval_num_shards_per_epoch: 8
+  eval_set_split_ratio: 0.1
+  eval_steps: 500
+  eval_strategy: 'no'
+  experiment_name: rc_bmethod_ctxpool_avg_L3_4latin_pd4ga2_4gpu
+  fp16: false
+  global_batch_size: 16
+  gradient_accumulation_steps: 2
+  gradient_checkpointing: false
+  learning_rate: 0.0001
+  logging_steps: 10
+  lr_action_head: 0.0001
+  lr_backbone: 1.0e-05
+  lr_base: 3.0e-05
+  lr_per_module_enabled: false
+  lr_scheduler_type: cosine_with_min_lr
+  max_concurrent_uploads: 2
+  max_grad_norm: 1.0
+  max_retries: 3
+  max_steps: 100000
+  min_lr: 5.0e-07
+  new_param_warmup_steps: 0
+  num_gpus: 4
+  open_loop_eval_plot_indices: null
+  open_loop_eval_steps_per_traj: 100
+  open_loop_eval_traj_ids:
+  - 0
+  optim: adamw_torch
+  output_dir: /data/rlwrld-unified-checkpoints/hyeonmo/checkpoints/ctxpool_bmethod/wam_dit4dit_robocasa_kitchen_ditjoint_wan22_fadistillV2V_shift5_g001_vstu_sL3_future_tf025_sl3_latin4_ctxpoolavgL3_L18_res256_tmosaic_pd4_ga2_gb16_100k_local_jnojid
+  remove_unused_columns: false
+  reset_stage2_lr_scheduler: false
+  resume_from_checkpoint: false
+  save_best_eval_metric_greater_is_better: true
+  save_best_eval_metric_name: ''
+  save_steps: 10000
+  save_total_limit: 5
+  save_vl_model: false
+  start_from_checkpoint: null
+  tf32: true
+  transformers_access_token: null
+  transformers_cache_dir: null
+  transformers_local_files_only: false
+  transformers_trust_remote_code: true
+  upload_checkpoints: false
+  upload_every: 1000
+  upload_last_n_checkpoints: 5
+  use_ddp: false
+  use_wandb: true
+  wandb_project: wam-dit4dit-robocasa-ctxpool-bmethod
+  warmup_ratio: 0.0
+  warmup_steps: 5000
+  weight_decay: 0.0

avg_L3/checkpoint-40000/experiment_cfg/dataset_statistics.json

@@ -0,0 +1,317 @@
+{
+  "new_embodiment": {
+    "state": {
+      "end_effector_position_relative": {
+        "min": [
+          -0.46594083309173584,
+          -0.8278615474700928,
+          0.011948669329285622
+        ],
+        "max": [
+          0.8734620809555054,
+          0.8441421985626221,
+          1.1348456144332886
+        ],
+        "mean": [
+          0.266306072473526,
+          -0.04002884030342102,
+          0.5537278652191162
+        ],
+        "std": [
+          0.19139817357063293,
+          0.2582354247570038,
+          0.2461009919643402
+        ],
+        "q01": [
+          -0.251964715719223,
+          -0.6177875232696534,
+          0.07541620314121246
+        ],
+        "q99": [
+          0.736240336894989,
+          0.6068667078018188,
+          0.9772176861763
+        ]
+      },
+      "end_effector_rotation_relative": {
+        "min": [
+          -0.9999996423721313,
+          -0.9910850524902344,
+          -0.9571061134338379,
+          6.351022108219695e-08
+        ],
+        "max": [
+          0.9999980926513672,
+          0.990567147731781,
+          0.964046835899353,
+          0.9740113019943237
+        ],
+        "mean": [
+          -0.14963096380233765,
+          0.1248803362250328,
+          0.0027880165725946426,
+          0.2082664668560028
+        ],
+        "std": [
+          0.7683019042015076,
+          0.36296239495277405,
+          0.37635329365730286,
+          0.2322801947593689
+        ],
+        "q01": [
+          -0.9973158955574035,
+          -0.6837371301651001,
+          -0.7814695286750794,
+          0.0013687612861394884
+        ],
+        "q99": [
+          0.9957762026786804,
+          0.8491807365417481,
+          0.7687161755561829,
+          0.802422981262207
+        ]
+      },
+      "gripper_qpos": {
+        "min": [
+          -0.027591193094849586,
+          -0.05319468677043915
+        ],
+        "max": [
+          0.07896614819765091,
+          0.10181591659784317
+        ],
+        "mean": [
+          0.030854083597660065,
+          -0.03072655387222767
+        ],
+        "std": [
+          0.013360796496272087,
+          0.013391095213592052
+        ],
+        "q01": [
+          -4.0135159069905056e-05,
+          -0.04032127052545548
+        ],
+        "q99": [
+          0.04066328033804893,
+          -0.00040660659200511887
+        ]
+      },
+      "base_position": {
+        "min": [
+          0.31250813603401184,
+          -5.100343704223633,
+          0.6963803172111511
+        ],
+        "max": [
+          5.601607322692871,
+          -0.4872598946094513,
+          0.7495372295379639
+        ],
+        "mean": [
+          2.5315651893615723,
+          -1.5020122528076172,
+          0.7073471546173096
+        ],
+        "std": [
+          1.4670116901397705,
+          0.8559004664421082,
+          0.006971539929508914
+        ],
+        "q01": [
+          0.5000457167625427,
+          -4.100119590759277,
+          0.6999446749687195
+        ],
+        "q99": [
+          5.5999836921691895,
+          -0.778001925945282,
+          0.7071865797042847
+        ]
+      },
+      "base_rotation": {
+        "min": [
+          0.0,
+          0.0,
+          -1.0,
+          0.0
+        ],
+        "max": [
+          0.0,
+          0.0,
+          1.0,
+          1.0
+        ],
+        "mean": [
+          0.0,
+          0.0,
+          0.2765239179134369,
+          0.6414101719856262
+        ],
+        "std": [
+          0.0,
+          0.0,
+          0.6192871332168579,
+          0.35383468866348267
+        ],
+        "q01": [
+          0.0,
+          0.0,
+          -1.0,
+          5.569559434661642e-06
+        ],
+        "q99": [
+          0.0,
+          0.0,
+          1.0,
+          1.0
+        ]
+      }
+    },
+    "action": {
+      "end_effector_position": {
+        "min": [
+          -1.0,
+          -1.0,
+          -1.0
+        ],
+        "max": [
+          1.0,
+          1.0,
+          1.0
+        ],
+        "mean": [
+          0.029150759801268578,
+          0.007607606239616871,
+          -0.037477411329746246
+        ],
+        "std": [
+          0.4175023138523102,
+          0.41343387961387634,
+          0.36864909529685974
+        ],
+        "q01": [
+          -1.0,
+          -1.0,
+          -1.0
+        ],
+        "q99": [
+          1.0,
+          1.0,
+          1.0
+        ]
+      },
+      "end_effector_rotation": {
+        "min": [
+          -1.0,
+          -1.0,
+          -1.0
+        ],
+        "max": [
+          1.0,
+          1.0,
+          1.0
+        ],
+        "mean": [
+          0.007235987111926079,
+          -0.023321982473134995,
+          -0.0024119403678923845
+        ],
+        "std": [
+          0.147710919380188,
+          0.15399158000946045,
+          0.13929523527622223
+        ],
+        "q01": [
+          -0.42321433901786804,
+          -0.5219977021217346,
+          -0.4101182746887207
+        ],
+        "q99": [
+          0.47503985762596074,
+          0.38744416117668146,
+          0.4242807936668394
+        ]
+      },
+      "gripper_close": {
+        "min": [
+          0.0
+        ],
+        "max": [
+          1.0
+        ],
+        "mean": [
+          0.34521719813346863
+        ],
+        "std": [
+          0.47949448227882385
+        ],
+        "q01": [
+          0.0
+        ],
+        "q99": [
+          1.0
+        ]
+      },
+      "base_motion": {
+        "min": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "max": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "mean": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "std": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "q01": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "q99": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ]
+      },
+      "control_mode": {
+        "min": [
+          0.0
+        ],
+        "max": [
+          0.0
+        ],
+        "mean": [
+          0.0
+        ],
+        "std": [
+          0.0
+        ],
+        "q01": [
+          0.0
+        ],
+        "q99": [
+          0.0
+        ]
+      }
+    },
+    "relative_action": {}
+  }
+}

avg_L3/checkpoint-40000/experiment_cfg/final_model_config.json

@@ -0,0 +1,155 @@
+{
+  "model_type": "WAM_DIT4DIT",
+  "model_dtype": "bfloat16",
+  "backbone_type": "wan22_diffusers",
+  "wan_model_path": "Wan-AI/Wan2.2-TI2V-5B",
+  "wan_diffusers_model_path": "Wan-AI/Wan2.2-TI2V-5B-Diffusers",
+  "wan_tokenizer_model_id": "Wan-AI/Wan2.1-T2V-1.3B",
+  "wan_feature_extraction_layer": 18,
+  "cosmos_model_path": "nvidia/Cosmos-Predict2.5-2B",
+  "cosmos_model_revision": "diffusers/base/pre-trained",
+  "backbone_embedding_dim": 3072,
+  "num_frames_in": 25,
+  "num_frames_out": 41,
+  "frame_downsample_factor": 2,
+  "video_resolution": 256,
+  "feature_extraction_layer": 18,
+  "feature_extraction_aggregate_all": false,
+  "feature_extraction_aggregate_layer_window": 0,
+  "truncate_above_feature_extraction": false,
+  "vl_adapter_num_layers": 0,
+  "tune_layers_above_feature_extraction": true,
+  "n_tune_layers_below_feature_extraction": 0,
+  "tune_patch_embed": false,
+  "tune_time_embed": false,
+  "tune_crossattn_proj": false,
+  "freeze_modules_pattern": "text_encoder,vae",
+  "freeze_whole_backbone": false,
+  "fm_loss_weight": 1.0,
+  "action_loss_weight": 1.0,
+  "conditional_frame_timestep": 0.0001,
+  "history_obs_noise_timestep": 0.0,
+  "history_num_clean_frames": 1,
+  "selfflow_rep": true,
+  "selfflow_rep_layer": -1,
+  "selfflow_rep_layer_student": -1,
+  "selfflow_rep_layer_teacher": -1,
+  "selfflow_mask_ratio": 0.1,
+  "selfflow_ema_decay": 0.9999,
+  "selfflow_rep_weight": 0.8,
+  "selfflow_teacher_noise": 0.0,
+  "fadistill_teacher_traj": true,
+  "fadistill_target_tau": 1.0,
+  "fadistill_teacher_nstep": 4,
+  "wan_fadistill_teacher_shift": 5.0,
+  "wan_fadistill_single_step_index": -1,
+  "wan_fadistill_align_weight": 0.01,
+  "wan_fadistill_teacher_stepper": "euler",
+  "wan_fadistill_action_from_teacher": false,
+  "wan_v2v_teacher": true,
+  "wan_v2v_student": true,
+  "wan_v2v_fm_mix": false,
+  "wan_block_causal": false,
+  "wan_learnable_token": false,
+  "wan_learnable_token_count": 64,
+  "wan_learnable_token_queries": 192,
+  "wan_learnable_token_layers": 2,
+  "wan_context_pool": "avg",
+  "wan_context_pool_layer": 3,
+  "wan_context_pool_n_cond_lat": 4,
+  "wan_context_pool_xattn_heads": 0,
+  "wan_context_pool_xattn_ffn_dim": 0,
+  "wan_fm_train_shift": 5.0,
+  "wan_text_pad_zero": true,
+  "nstep_agg": false,
+  "nstep_n": 2,
+  "nstep_query_len": 588,
+  "nstep_step_embed": true,
+  "nstep_agg_layers": 4,
+  "nstep_train_integration": false,
+  "nstep_noise_ensemble": false,
+  "nstep_base_residual": false,
+  "nstep_base_residual_gate": "layerscale",
+  "nstep_compress_query": false,
+  "nstep_compress_k": 64,
+  "nstep_compress_n_frame": 3,
+  "nstep_readout_exclude_step0": false,
+  "nstep_per_step_gated": false,
+  "nstep_step_dropout": false,
+  "nstep_residual_warmup": false,
+  "nstep_residual_warmup_start": 10000,
+  "nstep_residual_warmup_full": 20000,
+  "nstep_certainty_fusion": false,
+  "nstep_certainty_ckpt": "",
+  "nstep_certainty_tau": 1.0,
+  "nstep_certainty_drop_cond": false,
+  "nstep_certainty_arch": "mlp",
+  "nstep_certainty_hidden": 1024,
+  "nstep_certainty_layers": 2,
+  "nstep_certainty_decalib": false,
+  "nstep_certainty_step_bias": false,
+  "nstep_latent_cond": false,
+  "use_step_weight_fusion": false,
+  "step_weight_target_tau": 0.7,
+  "step_weight_sigmas": "0.995,0.9631,0.594,0.0099",
+  "step_weight_predictor_layers": 4,
+  "step_weight_predictor_heads": 16,
+  "step_weight_predictor_ckpt": "",
+  "freeze_step_weight_predictor": false,
+  "step_weight_exclude_current_tokens": false,
+  "multiview_layout": "tmosaic",
+  "fm_time_distribution": "uniform",
+  "fm_high_sigma_ratio": null,
+  "fm_high_sigma_min": 0.98,
+  "future_loss_type": "flow_matching",
+  "future_num_inference_steps": 1,
+  "use_noise_frames_in_extraction": true,
+  "use_meta_queries": false,
+  "n_meta_queries": 8,
+  "meta_queries_mode": "meta_only",
+  "use_meta_queries_in_fm": false,
+  "use_vlm_meta": false,
+  "use_vgln": false,
+  "tune_vgln": true,
+  "action_text_kv": false,
+  "state_dropout_prob": 0.0,
+  "state_additive_noise_scale": 0.0,
+  "training_mode": "joint",
+  "action_head_model_type": "dit",
+  "max_state_dim": 64,
+  "apply_sincos_state_encoding": false,
+  "use_percentiles": false,
+  "video_train_crop": false,
+  "image_resize_backend": "cv2_linear",
+  "max_action_dim": 32,
+  "action_horizon": 16,
+  "video_context_exceeds_action_horizon": false,
+  "action_future_slots": null,
+  "hidden_size": 2560,
+  "input_embedding_dim": 768,
+  "repeated_diffusion_steps": 4,
+  "add_pos_embed": true,
+  "max_seq_len": 1024,
+  "attn_dropout": 0.2,
+  "action_dit_size": "DiT-B",
+  "diffusion_model_cfg": {
+    "num_attention_heads": 12,
+    "attention_head_dim": 64,
+    "num_layers": 16,
+    "output_dim": 2560,
+    "dropout": 0.2,
+    "final_dropout": true,
+    "interleave_self_attention": true,
+    "norm_type": "ada_norm",
+    "positional_embeddings": null
+  },
+  "num_inference_timesteps": 4,
+  "noise_beta_alpha": 1.5,
+  "noise_beta_beta": 1.0,
+  "noise_s": 0.999,
+  "num_timestep_buckets": 1000,
+  "tune_projector": true,
+  "tune_diffusion_model": true,
+  "max_num_embodiments": 36,
+  "new_embodiment_train_ratio": 0
+}

avg_L3/checkpoint-40000/model-00001-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c1f889831c4e697ba28ed4dabf7e452c422391ece9284b6b66e00dd1fddd3e7
+size 4912561816

avg_L3/checkpoint-40000/model-00002-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f4492b08bd696f51302e510d37968ebaeff4d42cfcb7749951bfa68af8b9b84
+size 4997866232

avg_L3/checkpoint-40000/model-00003-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03395a0f315201aa2563a43fb25d294cbf621b1cf522922a0054d96341da32b4
+size 4953653648

avg_L3/checkpoint-40000/model-00004-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7453a1cac7b1e8f39d80d29466ef2b1eac0bfe1c08f0e4b713bfc5d3eca0a936
+size 4985255480

avg_L3/checkpoint-40000/model-00005-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aadd56338297f0f220d0dcaafb6514680ef663a5ccedfed57c66d0b68021b44b
+size 4834206416

avg_L3/checkpoint-40000/model-00006-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:504d3d882d2cae78237ad3370baf25088cf062ee869f516533c6e0451002bb1d
+size 4985255656

avg_L3/checkpoint-40000/model-00007-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f28aaadd21b827c10962a27f875f0ebf175cf37f7e9f04e7332ff20c6ee0b937
+size 2867243424

avg_L3/checkpoint-40000/processor_config.json

@@ -0,0 +1,67 @@
+{
+  "processor_class": "WAMDiT4DiTProcessor",
+  "downsampled_num_frames_in": 13,
+  "downsampled_num_frames_out": 21,
+  "frame_downsample_factor": 2,
+  "max_state_dim": 64,
+  "max_action_dim": 32,
+  "max_action_horizon": 16,
+  "formalize_language": true,
+  "new_embodiment_train_ratio": 0,
+  "use_percentiles": false,
+  "clip_outliers": true,
+  "apply_sincos_state_encoding": false,
+  "video_train_crop": false,
+  "image_resize_backend": "cv2_linear",
+  "multiview_layout": "tmosaic",
+  "video_resolution": 256,
+  "use_relative_action": false,
+  "embodiment_id_mapping": {
+    "new_embodiment": 0,
+    "fractal20220817_data": 1,
+    "kuka": 2,
+    "bridge_orig": 3,
+    "taco_play": 4,
+    "jaco_play": 5,
+    "berkeley_cable_routing": 6,
+    "roboturk": 7,
+    "viola": 8,
+    "berkeley_autolab_ur5": 9,
+    "toto": 10,
+    "language_table": 11,
+    "stanford_hydra_dataset_converted_externally_to_rlds": 12,
+    "austin_buds_dataset_converted_externally_to_rlds": 13,
+    "nyu_franka_play_dataset_converted_externally_to_rlds": 14,
+    "furniture_bench_dataset_converted_externally_to_rlds": 15,
+    "ucsd_kitchen_dataset_converted_externally_to_rlds": 16,
+    "austin_sailor_dataset_converted_externally_to_rlds": 17,
+    "austin_sirius_dataset_converted_externally_to_rlds": 18,
+    "dlr_edan_shared_control_converted_externally_to_rlds": 19,
+    "iamlab_cmu_pickup_insert_converted_externally_to_rlds": 20,
+    "utaustin_mutex": 21,
+    "berkeley_fanuc_manipulation": 22,
+    "cmu_stretch": 23,
+    "bc_z": 24,
+    "fmb_dataset": 25,
+    "dobbe": 26,
+    "droid": 27,
+    "agibot_dexhand": 28,
+    "agibot_gripper": 29,
+    "galaxea": 30,
+    "humanoid_everyday_g1": 31,
+    "humanoid_everyday_h1": 32,
+    "action_net": 33,
+    "neural_gr1": 34,
+    "simulation_gr1": 35,
+    "libero_panda": 2
+  },
+  "use_vlm_meta": false,
+  "vlm_model_name": "huiwon/alinvlm_v1_3",
+  "vlm_backbone_model_type": "qwen3_vl",
+  "conversation_image_first": false,
+  "image_aug": true,
+  "image_aug_stronger": false,
+  "image_aug_moderate": true,
+  "image_aug_moderate_allview": true,
+  "text_dropout": 0.1
+}

avg_L3/checkpoint-40000/statistics.json

@@ -0,0 +1,317 @@
+{
+  "new_embodiment": {
+    "state": {
+      "end_effector_position_relative": {
+        "min": [
+          -0.46594083309173584,
+          -0.8278615474700928,
+          0.011948669329285622
+        ],
+        "max": [
+          0.8734620809555054,
+          0.8441421985626221,
+          1.1348456144332886
+        ],
+        "mean": [
+          0.266306072473526,
+          -0.04002884030342102,
+          0.5537278652191162
+        ],
+        "std": [
+          0.19139817357063293,
+          0.2582354247570038,
+          0.2461009919643402
+        ],
+        "q01": [
+          -0.251964715719223,
+          -0.6177875232696534,
+          0.07541620314121246
+        ],
+        "q99": [
+          0.736240336894989,
+          0.6068667078018188,
+          0.9772176861763
+        ]
+      },
+      "end_effector_rotation_relative": {
+        "min": [
+          -0.9999996423721313,
+          -0.9910850524902344,
+          -0.9571061134338379,
+          6.351022108219695e-08
+        ],
+        "max": [
+          0.9999980926513672,
+          0.990567147731781,
+          0.964046835899353,
+          0.9740113019943237
+        ],
+        "mean": [
+          -0.14963096380233765,
+          0.1248803362250328,
+          0.0027880165725946426,
+          0.2082664668560028
+        ],
+        "std": [
+          0.7683019042015076,
+          0.36296239495277405,
+          0.37635329365730286,
+          0.2322801947593689
+        ],
+        "q01": [
+          -0.9973158955574035,
+          -0.6837371301651001,
+          -0.7814695286750794,
+          0.0013687612861394884
+        ],
+        "q99": [
+          0.9957762026786804,
+          0.8491807365417481,
+          0.7687161755561829,
+          0.802422981262207
+        ]
+      },
+      "gripper_qpos": {
+        "min": [
+          -0.027591193094849586,
+          -0.05319468677043915
+        ],
+        "max": [
+          0.07896614819765091,
+          0.10181591659784317
+        ],
+        "mean": [
+          0.030854083597660065,
+          -0.03072655387222767
+        ],
+        "std": [
+          0.013360796496272087,
+          0.013391095213592052
+        ],
+        "q01": [
+          -4.0135159069905056e-05,
+          -0.04032127052545548
+        ],
+        "q99": [
+          0.04066328033804893,
+          -0.00040660659200511887
+        ]
+      },
+      "base_position": {
+        "min": [
+          0.31250813603401184,
+          -5.100343704223633,
+          0.6963803172111511
+        ],
+        "max": [
+          5.601607322692871,
+          -0.4872598946094513,
+          0.7495372295379639
+        ],
+        "mean": [
+          2.5315651893615723,
+          -1.5020122528076172,
+          0.7073471546173096
+        ],
+        "std": [
+          1.4670116901397705,
+          0.8559004664421082,
+          0.006971539929508914
+        ],
+        "q01": [
+          0.5000457167625427,
+          -4.100119590759277,
+          0.6999446749687195
+        ],
+        "q99": [
+          5.5999836921691895,
+          -0.778001925945282,
+          0.7071865797042847
+        ]
+      },
+      "base_rotation": {
+        "min": [
+          0.0,
+          0.0,
+          -1.0,
+          0.0
+        ],
+        "max": [
+          0.0,
+          0.0,
+          1.0,
+          1.0
+        ],
+        "mean": [
+          0.0,
+          0.0,
+          0.2765239179134369,
+          0.6414101719856262
+        ],
+        "std": [
+          0.0,
+          0.0,
+          0.6192871332168579,
+          0.35383468866348267
+        ],
+        "q01": [
+          0.0,
+          0.0,
+          -1.0,
+          5.569559434661642e-06
+        ],
+        "q99": [
+          0.0,
+          0.0,
+          1.0,
+          1.0
+        ]
+      }
+    },
+    "action": {
+      "end_effector_position": {
+        "min": [
+          -1.0,
+          -1.0,
+          -1.0
+        ],
+        "max": [
+          1.0,
+          1.0,
+          1.0
+        ],
+        "mean": [
+          0.029150759801268578,
+          0.007607606239616871,
+          -0.037477411329746246
+        ],
+        "std": [
+          0.4175023138523102,
+          0.41343387961387634,
+          0.36864909529685974
+        ],
+        "q01": [
+          -1.0,
+          -1.0,
+          -1.0
+        ],
+        "q99": [
+          1.0,
+          1.0,
+          1.0
+        ]
+      },
+      "end_effector_rotation": {
+        "min": [
+          -1.0,
+          -1.0,
+          -1.0
+        ],
+        "max": [
+          1.0,
+          1.0,
+          1.0
+        ],
+        "mean": [
+          0.007235987111926079,
+          -0.023321982473134995,
+          -0.0024119403678923845
+        ],
+        "std": [
+          0.147710919380188,
+          0.15399158000946045,
+          0.13929523527622223
+        ],
+        "q01": [
+          -0.42321433901786804,
+          -0.5219977021217346,
+          -0.4101182746887207
+        ],
+        "q99": [
+          0.47503985762596074,
+          0.38744416117668146,
+          0.4242807936668394
+        ]
+      },
+      "gripper_close": {
+        "min": [
+          0.0
+        ],
+        "max": [
+          1.0
+        ],
+        "mean": [
+          0.34521719813346863
+        ],
+        "std": [
+          0.47949448227882385
+        ],
+        "q01": [
+          0.0
+        ],
+        "q99": [
+          1.0
+        ]
+      },
+      "base_motion": {
+        "min": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "max": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "mean": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "std": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "q01": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ],
+        "q99": [
+          0.0,
+          0.0,
+          0.0,
+          0.0
+        ]
+      },
+      "control_mode": {
+        "min": [
+          0.0
+        ],
+        "max": [
+          0.0
+        ],
+        "mean": [
+          0.0
+        ],
+        "std": [
+          0.0
+        ],
+        "q01": [
+          0.0
+        ],
+        "q99": [
+          0.0
+        ]
+      }
+    },
+    "relative_action": {}
+  }
+}

avg_L3/checkpoint-40000/wandb_config.json

@@ -0,0 +1 @@
+{"project": "wam-dit4dit-robocasa-ctxpool-bmethod", "run_id": "rc_bmethod_ctxpool_avg_L3_4latin_pd4ga2_4gpu"}

avg_L3/checkpoint-40000/zero_to_fp32.py

@@ -0,0 +1,790 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS, AUTOEP_LAYERS_KEY,
+                                            AUTOEP_LAYERS_KEY_LEGACY, AUTOEP_ZERO3_EXPERT_STATE_FORMAT_KEY,
+                                            AUTOEP_ZERO3_PARTITIONED_EXPERT_STATE_FORMAT)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def _has_autoep_zero3_partitioned_metadata(state_dict):
+    autoep_layers = state_dict.get(AUTOEP_LAYERS_KEY)
+    if autoep_layers is None:
+        autoep_layers = state_dict.get(AUTOEP_LAYERS_KEY_LEGACY)
+    if not isinstance(autoep_layers, list):
+        return False
+    return any(
+        isinstance(entry, dict)
+        and entry.get(AUTOEP_ZERO3_EXPERT_STATE_FORMAT_KEY) == AUTOEP_ZERO3_PARTITIONED_EXPERT_STATE_FORMAT
+        for entry in autoep_layers)
+
+
+def _raise_if_autoep_zero3_partitioned_state(state_dict):
+    if _has_autoep_zero3_partitioned_metadata(state_dict):
+        raise NotImplementedError("zero_to_fp32 does not support AutoEP ZeRO-3 partition-native checkpoints. "
+                                  "AutoEP expert parameters are partitioned over expert replica groups, so "
+                                  "global data-parallel consolidation would produce incomplete expert tensors. "
+                                  "Use ds_to_universal.py for expert-aware conversion.")
+
+
+def _raise_if_autoep_zero3_partitioned_checkpoint(model_files):
+    for file in model_files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+        _raise_if_autoep_zero3_partitioned_state(state_dict)
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+        _raise_if_autoep_zero3_partitioned_state(state_dict)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if ZERO_STAGE not in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    # parse_model_states rejects AutoEP ZeRO-3 partition-native checkpoints
+    # before the expensive optimizer-shard load below.
+    model_files = get_model_state_files(ds_checkpoint_dir)
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info("Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info("Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)