diff --git a/.gitattributes b/.gitattributes index 97af60d27ea9c214269b58bc438f47dffbe680a3..c14a079015e7126cb8a88e3019ae1b7a900901e0 100644 --- a/.gitattributes +++ b/.gitattributes @@ -56,3 +56,4 @@ ultraep/tierbtrace/tierbtrace_gbs8/QWEN2_57B__ultraep__matched__seq4096__gbs8__c moonep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__moonep__matched__seq8192__gbs8__capnone/nsys/DSV2_LITE_L24__moonep__matched__seq8192__gbs8__capnone.nsys-rep filter=lfs diff=lfs merge=lfs -text moonep/tierbtrace/tierbtrace_gbs8/QWEN2_57B__moonep__matched__seq4096__gbs8__capnone/nsys/QWEN2_57B__moonep__matched__seq4096__gbs8__capnone.nsys-rep filter=lfs diff=lfs merge=lfs -text tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_235B__echo__matched__seq8192__gbs8__cap1p083984375/nsys/QWEN3_235B__echo__matched__seq8192__gbs8__cap1p083984375.nsys-rep filter=lfs diff=lfs merge=lfs -text +tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep filter=lfs diff=lfs merge=lfs -text diff --git a/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep b/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep new file mode 100644 index 0000000000000000000000000000000000000000..959ecbab30e13d70bc1830b80357b918f6c67d9c --- /dev/null +++ b/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25eaff0e2e4a38b992be3115148f03b3b72495e68bc93102a82ab27c019f0e23 +size 34284034 diff --git a/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/run.env b/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/run.env new file mode 100644 index 0000000000000000000000000000000000000000..df34a3784326edb6ffa2f547a6e6e390d629d0fd --- /dev/null +++ b/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/run.env @@ -0,0 +1,70 @@ +backend=echo +config=QWEN3_30B +family=tierB +repo=/workspace/Megatron-MoE-Echo +branch=bracing-regions-through-kernels +commit=9a236793 +num_layers=48 +hidden=2048 +ffn_hidden=6144 +moe_ffn_hidden=768 +num_experts=128 +topk=8 +experts_per_rank=16 +expert_params=4718592 +heads=32 +kv_groups=4 +kv_channels=128 +qk_layernorm=1 +shared_ffn=0 +shared_expert_overlap=0 +moe_token_dispatcher=flex +mla=0 +dense_first_layers=0 +moe_layers=48 +rotary_base=1000000 +rotary_scaling= +qkv_bias=0 +router_balancing=aux_loss +aux_loss_coeff=1e-3 +router_pre_softmax=0 +router_fusion=0 +router_dtype=fp32 +device_initiated_grouped_gemm=1 +norm_epsilon=1e-6 +seq_length=4096 +micro_batch_size=1 +global_batch_size=8 +num_microbatches=1 +assignments_per_rank=32768 +mean_tokens_per_expert=2048 +slot_policy=matched +slots_per_rank=16 +capacity_mode=static +capacity=1.11328125 +cuda_graph=1 +probe_mode=0 +async_grad_reduce=0 +branch_role=timing +memory_dump_mode=none +memory_snapshot_path= +cuda_graph_warmup_steps=3 +recompute=none +force_balance=0 +train_iters=10 +ep_size=8 +echo_dispatch_overlap=1 +echo_dispatch_recompute=0 +ultraep_quota_min_tokens= +ultraep_weight_sync_join= +profile_enable=1 +kernel_trace=1 +fp8_mode=mxfp8 +fp8_recipe=mxfp8 +training_seed=1234 +learning_rate=1e-4 +min_learning_rate=1.0e-5 +lr_warmup_iters=5 +state_dump_enable=0 +state_dump_dir= +moonep_assert_no_pending_reduce=0 diff --git a/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/train.log b/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/train.log new file mode 100644 index 0000000000000000000000000000000000000000..0fa142a7fc06bd49ce3f9d75ea7ee8c9ae47697c --- /dev/null +++ b/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/train.log @@ -0,0 +1,2805 @@ +[00:17:27] WARN: bracing-regions-through-kernels lacks --moe-router-fusion; QWEN3_30B runs unfused (timing only, routing identical) +============================================================== + QWEN3_30B / echo (slots/rank=16 policy=matched) + E=128 K=8 H'=768 m=16 + L=48 seq=4096 gbs=8 (1 microbatches) + cap_mode=static cap=1.11328125 cg=1 recompute=none + memory_dump=none snapshots=/workspace/tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/nsys/memory_snapshots/snapshot.pickle + repo=/workspace/Megatron-MoE-Echo @ 9a236793 +============================================================== +W0830 00:17:28.617000 301137 torch/distributed/run.py:766] +W0830 00:17:28.617000 301137 torch/distributed/run.py:766] ***************************************** +W0830 00:17:28.617000 301137 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +W0830 00:17:28.617000 301137 torch/distributed/run.py:766] ***************************************** +/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled + warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled") +/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled + warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled") +/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled + warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled") +/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled + warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled") +/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled + warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled") +/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled + warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled") +WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it +using world size: 8, data-parallel size: 8, context-parallel size: 1, hierarchical context-parallel sizes: None, tensor-model-parallel size: 1, pipeline-model-parallel size: 1 +WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer +Number of virtual stages per pipeline stage: None +accumulate and all-reduce gradients in fp32 for bfloat16 data type. +using torch.bfloat16 for parameters ... +/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled + warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled") +------------------------ arguments ------------------------ + account_for_embedding_in_pipeline_split ......... False + account_for_loss_in_pipeline_split .............. False + accumulate_allreduce_grads_in_fp32 .............. True + adam_beta1 ...................................... 0.9 + adam_beta2 ...................................... 0.999 + adam_eps ........................................ 1e-08 + add_bias_linear ................................. False + add_position_embedding .......................... False + add_qkv_bias .................................... False + adlr_autoresume ................................. False + adlr_autoresume_interval ........................ 1000 + align_grad_reduce ............................... True + align_param_gather .............................. False + app_tag_run_name ................................ None + app_tag_run_version ............................. 0.0.0 + apply_layernorm_1p .............................. False + apply_query_key_layer_scaling ................... False + apply_residual_connection_post_layernorm ........ False + apply_rope_fusion ............................... True + async_save ...................................... None + async_tensor_model_parallel_allreduce ........... True + attention_backend ............................... AttnBackend.auto + attention_dropout ............................... 0.0 + attention_softmax_in_fp32 ....................... False + auto_detect_ckpt_format ......................... False + barrier_with_L1_time ............................ True + bert_binary_head ................................ True + bert_embedder_type .............................. megatron + bert_load ....................................... None + bf16 ............................................ True + bias_dropout_fusion ............................. True + bias_gelu_fusion ................................ False + bias_swiglu_fusion .............................. True + biencoder_projection_dim ........................ 0 + biencoder_shared_query_context_model ............ False + block_data_path ................................. None + calc_ft_timeouts ................................ False + calculate_per_token_loss ........................ False + check_for_large_grads ........................... False + check_for_nan_in_loss_and_grad .................. False + check_for_spiky_loss ............................ False + check_weight_hash_across_dp_replicas_interval ... None + ckpt_assume_constant_structure .................. False + ckpt_convert_format ............................. None + ckpt_convert_save ............................... None + ckpt_convert_update_legacy_dist_opt_format ...... False + ckpt_format ..................................... torch_dist + ckpt_fully_parallel_load ........................ False + ckpt_fully_parallel_save ........................ True + ckpt_fully_parallel_save_deprecated ............. False + ckpt_step ....................................... None + classes_fraction ................................ 1.0 + clip_grad ....................................... 1.0 + clone_scatter_output_in_embedding ............... True + config_logger_dir ............................... + consumed_train_samples .......................... 0 + consumed_valid_samples .......................... 0 + context_parallel_size ........................... 1 + cp_comm_type .................................... ['p2p'] + create_attention_mask_in_dataloader ............. True + cross_entropy_fusion_impl ....................... native + cross_entropy_loss_fusion ....................... False + cuda_graph_scope ................................ full_iteration + cuda_graph_warmup_steps ......................... 3 + data_args_path .................................. None + data_cache_path ................................. None + data_parallel_random_init ....................... False + data_parallel_sharding_strategy ................. no_shard + data_parallel_size .............................. 8 + data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document'] + data_per_class_fraction ......................... 1.0 + data_sharding ................................... True + dataloader_type ................................. single + ddp_average_in_collective ....................... False + ddp_bucket_size ................................. None + ddp_num_buckets ................................. None + ddp_pad_buckets_for_high_nccl_busbw ............. False + decoder_first_pipeline_num_layers ............... None + decoder_last_pipeline_num_layers ................ None + decoder_num_layers .............................. None + decoder_seq_length .............................. None + decoupled_lr .................................... None + decoupled_min_lr ................................ None + decrease_batch_size_if_needed ................... False + defer_embedding_wgrad_compute ................... False + delay_wgrad_compute ............................. False + deprecated_use_mcore_models ..................... True + deterministic_mode .............................. False + dino_bottleneck_size ............................ 256 + dino_freeze_last_layer .......................... 1 + dino_head_hidden_size ........................... 2048 + dino_local_crops_number ......................... 10 + dino_local_img_size ............................. 96 + dino_norm_last_layer ............................ False + dino_teacher_temp ............................... 0.07 + dino_warmup_teacher_temp ........................ 0.04 + dino_warmup_teacher_temp_epochs ................. 30 + disable_bf16_reduced_precision_matmul ........... False + disable_mamba_mem_eff_path ...................... False + disable_straggler_on_startup .................... False + dist_ckpt_format_deprecated ..................... None + dist_ckpt_strictness ............................ assume_ok_unexpected + distribute_saved_activations .................... False + distributed_backend ............................. nccl + distributed_timeout_minutes ..................... 10 + embedding_init_method_std ....................... None + embedding_path .................................. None + empty_unused_memory_level ....................... 0 + enable_cuda_graph ............................... True + enable_experimental ............................. False + enable_ft_package ............................... False + enable_gloo_process_groups ...................... True + enable_msc ...................................... True + enable_one_logger ............................... True + encoder_num_layers .............................. 48 + encoder_seq_length .............................. 4096 + end_weight_decay ................................ 0.1 + eod_mask_loss ................................... False + error_injection_rate ............................ 0 + error_injection_type ............................ transient_error + eval_interval ................................... 1000000 + eval_iters ...................................... 0 + evidence_data_path .............................. None + exit_duration_in_mins ........................... None + exit_interval ................................... None + exit_on_missing_checkpoint ...................... False + exit_signal_handler ............................. False + exp_avg_dtype ................................... torch.float32 + exp_avg_sq_dtype ................................ torch.float32 + expert_model_parallel_size ...................... 8 + expert_tensor_parallel_size ..................... 1 + export_force_local_attention .................... False + export_kd_cfg ................................... None + export_kd_teacher_ckpt_format ................... None + export_kd_teacher_load .......................... None + export_kv_cache_quant ........................... False + export_legacy_megatron .......................... False + export_model_type ............................... GPTModel + export_moe_apply_probs_on_input ................. False + export_qk_l2_norm ............................... False + export_quant_cfg ................................ None + export_real_quant_cfg ........................... None + export_te_mcore_model ........................... False + external_cuda_graph ............................. False + ffn_hidden_size ................................. 6144 + finetune ........................................ False + finetune_data_split ............................. train + finetune_hf_dataset ............................. None + first_last_layers_bf16 .......................... False + flash_decode .................................... False + fp16 ............................................ False + fp16_lm_cross_entropy ........................... False + fp32_residual_connection ........................ False + fp8 ............................................. e4m3 + fp8_amax_compute_algo ........................... most_recent + fp8_amax_history_len ............................ 1 + fp8_interval .................................... 1 + fp8_margin ...................................... 0 + fp8_param_gather ................................ False + fp8_recipe ...................................... mxfp8 + fp8_wgrad ....................................... True + fsdp_double_buffer .............................. False + global_batch_size ............................... 8 + grad_reduce_in_bf16 ............................. False + gradient_accumulation_fusion .................... True + gradient_reduce_div_fusion ...................... True + group_query_attention ........................... True + head_lr_mult .................................... 1.0 + heterogeneous_layers_config_encoded_json ........ None + heterogeneous_layers_config_path ................ None + hidden_dropout .................................. 0.0 + hidden_size ..................................... 2048 + hierarchical_context_parallel_sizes ............. None + high_priority_stream_groups ..................... [] + hybrid_attention_ratio .......................... 0.0 + hybrid_mlp_ratio ................................ 0.0 + hybrid_override_pattern ......................... None + hysteresis ...................................... 2 + ict_head_size ................................... None + ict_load ........................................ None + img_h ........................................... 224 + img_w ........................................... 224 + indexer_batch_size .............................. 128 + indexer_log_interval ............................ 1000 + inference_batch_times_seqlen_threshold .......... -1 + inference_dynamic_batching ...................... False + inference_dynamic_batching_buffer_guaranteed_fraction 0.2 + inference_dynamic_batching_buffer_overflow_factor None + inference_dynamic_batching_buffer_size_gb ....... 40.0 + inference_dynamic_batching_chunk_size ........... 256 + inference_dynamic_batching_max_requests_override None + inference_dynamic_batching_max_tokens_override .. None + inference_dynamic_batching_num_cuda_graphs ...... 16 + inference_max_batch_size ........................ 8 + inference_max_seq_length ........................ 2560 + inference_rng_tracker ........................... False + init_method_std ................................. 0.01 + init_method_xavier_uniform ...................... False + init_model_with_meta_device ..................... False + initial_loss_scale .............................. 4294967296 + inprocess_active_world_size ..................... 8 + inprocess_barrier_timeout ....................... 120 + inprocess_completion_timeout .................... 120 + inprocess_empty_cuda_cache ...................... False + inprocess_granularity ........................... node + inprocess_hard_timeout .......................... 90 + inprocess_heartbeat_interval .................... 30 + inprocess_heartbeat_timeout ..................... 60 + inprocess_last_call_wait ........................ 1 + inprocess_max_iterations ........................ None + inprocess_monitor_process_interval .............. 1.0 + inprocess_monitor_thread_interval ............... 1.0 + inprocess_progress_watchdog_interval ............ 1.0 + inprocess_restart ............................... False + inprocess_soft_timeout .......................... 60 + inprocess_termination_grace_time ................ 1 + is_hybrid_model ................................. False + iter_per_epoch .................................. 1250 + iterations_to_skip .............................. [] + keep_fp8_transpose_cache_when_using_custom_fsdp . False + kitchen_config_file ............................. None + kitchen_recipe_number ........................... None + kv_channels ..................................... 128 + kv_lora_rank .................................... 32 + lazy_mpu_init ................................... None + load ............................................ None + load_main_params_from_ckpt ...................... None + load_model_opt_format ........................... False + local_rank ...................................... 0 + log_energy ...................................... False + log_interval .................................... 1 + log_loss_scale_to_tensorboard ................... True + log_memory_to_tensorboard ....................... False + log_num_zeros_in_grad ........................... False + log_params_norm ................................. False + log_progress .................................... False + log_straggler ................................... False + log_throughput .................................. True + log_timers_to_tensorboard ....................... False + log_validation_ppl_to_tensorboard ............... False + log_world_size_to_tensorboard ................... False + logging_level ................................... None + loss_scale ...................................... None + loss_scale_window ............................... 1000 + lr .............................................. 0.0001 + lr_decay_iters .................................. 8 + lr_decay_samples ................................ None + lr_decay_style .................................. cosine + lr_warmup_fraction .............................. None + lr_warmup_init .................................. 0.0 + lr_warmup_iters ................................. 5 + lr_warmup_samples ............................... 0 + lr_wsd_decay_iters .............................. None + lr_wsd_decay_samples ............................ None + lr_wsd_decay_style .............................. exponential + main_grads_dtype ................................ torch.float32 + main_params_dtype ............................... torch.float32 + make_vocab_size_divisible_by .................... 128 + mamba_head_dim .................................. 64 + mamba_num_groups ................................ 8 + mamba_num_heads ................................. None + mamba_state_dim ................................. 128 + manual_gc ....................................... False + manual_gc_eval .................................. True + manual_gc_interval .............................. 0 + mask_factor ..................................... 1.0 + mask_prob ....................................... 0.15 + mask_type ....................................... random + masked_softmax_fusion ........................... False + max_position_embeddings ......................... 32768 + max_tokens_to_oom ............................... 12000 + memory_snapshot_path ............................ snapshot.pickle + merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt + micro_batch_size ................................ 1 + microbatch_group_size_per_vp_stage .............. None + mid_level_dataset_surplus ....................... 0.005 + min_loss_scale .................................. 1.0 + min_lr .......................................... 1e-05 + mlp_chunks_for_prefill .......................... 1 + mmap_bin_files .................................. True + mock_data ....................................... False + moe_apply_probs_on_input ........................ False + moe_aux_loss_coeff .............................. 0.001 + moe_deepep_num_sms .............................. 20 + moe_echo_enable_random_offloading ............... False + moe_echo_expert_dispatch_overlap ................ True + moe_echo_expert_dispatcher_type ................. hybridep + moe_echo_recompute_expert_dispatch .............. False + moe_enable_deepep ............................... False + moe_enable_echo ................................. True + moe_enable_hybridep ............................. True + moe_expert_capacity_factor ...................... None + moe_extended_tp ................................. False + moe_ffn_hidden_size ............................. 768 + moe_grouped_gemm ................................ True + moe_hybridep_num_sms ............................ 16 + moe_input_jitter_eps ............................ None + moe_layer_freq .................................. 1 + moe_layer_recompute ............................. False + moe_num_echo_experts ............................ 128 + moe_pad_expert_input_to_capacity ................ False + moe_per_layer_logging ........................... False + moe_permute_fusion .............................. False + moe_received_token_capacity ..................... 1.11328125 + moe_router_bias_update_rate ..................... 0.001 + moe_router_dtype ................................ fp32 + moe_router_enable_expert_bias ................... False + moe_router_force_load_balancing ................. False + moe_router_group_topk ........................... None + moe_router_load_balancing_type .................. aux_loss + moe_router_num_groups ........................... None + moe_router_padding_for_fp8 ...................... False + moe_router_pre_softmax .......................... False + moe_router_score_function ....................... softmax + moe_router_topk ................................. 8 + moe_router_topk_scaling_factor .................. None + moe_shared_expert_intermediate_size ............. None + moe_shared_expert_overlap ....................... False + moe_token_dispatcher_type ....................... flex + moe_token_drop_policy ........................... probs + moe_upcycling_granularity ....................... 1 + moe_use_device_initiated_grouped_gemm ........... True + moe_use_legacy_grouped_gemm ..................... False + moe_use_upcycling ............................... False + moe_z_loss_coeff ................................ None + mrope_section ................................... None + mscale .......................................... 1.0 + mscale_all_dim .................................. 0.0 + mtp_loss_scaling_factor ......................... 0.1 + mtp_num_layers .................................. None + multi_latent_attention .......................... False + nccl_all_reduce_for_prefill ..................... False + nccl_communicator_config_path ................... None + nccl_ub ......................................... False + no_load_optim ................................... None + no_load_rng ..................................... None + no_persist_layer_norm ........................... False + no_rope_freq .................................... None + no_save_optim ................................... None + no_save_rng ..................................... None + non_persistent_ckpt_type ........................ None + non_persistent_global_ckpt_dir .................. None + non_persistent_local_ckpt_algo .................. fully_parallel + non_persistent_local_ckpt_dir ................... None + non_persistent_save_interval .................... None + norm_epsilon .................................... 1e-06 + normalization ................................... RMSNorm + num_attention_heads ............................. 32 + num_channels .................................... 3 + num_classes ..................................... 1000 + num_dataset_builder_threads ..................... 1 + num_distributed_optimizer_instances ............. 1 + num_experts ..................................... 128 + num_layers ...................................... 48 + num_layers_at_end_in_bf16 ....................... 1 + num_layers_at_start_in_bf16 ..................... 1 + num_layers_per_virtual_pipeline_stage ........... None + num_query_groups ................................ 4 + num_virtual_stages_per_pipeline_rank ............ None + num_workers ..................................... 2 + object_storage_cache_path ....................... None + one_logger_async ................................ False + one_logger_project .............................. megatron-lm + one_logger_run_name ............................. None + onnx_safe ....................................... None + openai_gelu ..................................... False + optimizer ....................................... adam + optimizer_cpu_offload ........................... False + optimizer_offload_fraction ...................... 1.0 + output_bert_embeddings .......................... False + overlap_cpu_optimizer_d2h_h2d ................... False + overlap_grad_reduce ............................. True + overlap_moe_expert_parallel_comm ................ False + overlap_p2p_comm ................................ False + overlap_p2p_comm_warmup_flush ................... False + overlap_param_gather ............................ True + overlap_param_gather_with_optimizer_step ........ False + override_opt_param_scheduler .................... False + params_dtype .................................... torch.bfloat16 + patch_dim ....................................... 16 + per_split_data_args_path ........................ None + perform_initialization .......................... True + pin_cpu_grads ................................... True + pin_cpu_params .................................. True + pipeline_model_parallel_comm_backend ............ None + pipeline_model_parallel_layout .................. None + pipeline_model_parallel_size .................... 1 + position_embedding_type ......................... rope + pretrained_checkpoint ........................... None + profile ......................................... True + profile_ranks ................................... [0, 1, 2, 3, 4, 5, 6, 7] + profile_step_end ................................ 8 + profile_step_start .............................. 6 + q_lora_rank ..................................... None + qk_head_dim ..................................... 128 + qk_l2_norm ...................................... False + qk_layernorm .................................... True + qk_pos_emb_head_dim ............................. 64 + query_in_block_prob ............................. 0.1 + rampup_batch_size ............................... None + rank ............................................ 0 + recompute_granularity ........................... None + recompute_method ................................ None + recompute_modules ............................... None + recompute_num_layers ............................ None + record_memory_history ........................... False + relative_attention_max_distance ................. 128 + relative_attention_num_buckets .................. 32 + replication ..................................... False + replication_factor .............................. 2 + replication_jump ................................ None + rerun_mode ...................................... validate_results + reset_attention_mask ............................ False + reset_position_ids .............................. False + result_rejected_tracker_filename ................ None + retriever_report_topk_accuracies ................ [] + retriever_score_scaling ......................... False + retriever_seq_length ............................ 256 + retro_add_retriever ............................. False + retro_attention_gate ............................ 1 + retro_cyclic_train_iters ........................ None + retro_encoder_attention_dropout ................. 0.1 + retro_encoder_hidden_dropout .................... 0.1 + retro_encoder_layers ............................ 2 + retro_num_neighbors ............................. 2 + retro_num_retrieved_chunks ...................... 2 + retro_project_dir ............................... None + retro_verify_neighbor_count ..................... True + reuse_grad_buf_for_mxfp8_param_ag ............... False + rope_scaling_factor ............................. 8.0 + rope_type ....................................... None + rotary_base ..................................... 1000000 + rotary_interleaved .............................. False + rotary_percent .................................. 1.0 + rotary_scaling_factor ........................... 1.0 + rotary_seq_len_interpolation_factor ............. None + run_workload_inspector_server ................... False + sample_rate ..................................... 1.0 + save ............................................ None + save_interval ................................... None + save_retain_interval ............................ None + scatter_gather_tensors_in_pipeline .............. True + seed ............................................ 1234 + seq_length ...................................... 4096 + sequence_parallel ............................... False + sft ............................................. False + sft_tokenizer_prompt_format ..................... nemotron-h-aligned + sgd_momentum .................................... 0.9 + short_seq_prob .................................. 0.1 + skip_train ...................................... False + skipped_train_samples ........................... 0 + spec ............................................ None + split ........................................... 100,0,0 + squared_relu .................................... False + start_weight_decay .............................. 0.1 + straggler_ctrlr_port ............................ 65535 + straggler_minmax_count .......................... 1 + suggested_communication_unit_size ............... None + swiglu .......................................... True + swin_backbone_type .............................. tiny + symmetric_ar_type ............................... None + te_rng_tracker .................................. True + tensor_model_parallel_size ...................... 1 + tensorboard_dir ................................. None + tensorboard_log_interval ........................ 1 + tensorboard_queue_size .......................... 1000 + test_data_path .................................. None + test_mode ....................................... False + tiktoken_num_special_tokens ..................... 1000 + tiktoken_pattern ................................ None + tiktoken_special_tokens ......................... None + timing_log_level ................................ 0 + timing_log_option ............................... minmax + titles_data_path ................................ None + tokenizer_model ................................. None + tokenizer_type .................................. GPT2BPETokenizer + torch_fsdp2_reshard_after_forward ............... True + tp_comm_bootstrap_backend ....................... nccl + tp_comm_bulk_dgrad .............................. True + tp_comm_bulk_wgrad .............................. True + tp_comm_overlap ................................. False + tp_comm_overlap_ag .............................. True + tp_comm_overlap_cfg ............................. None + tp_comm_overlap_rs .............................. True + tp_comm_overlap_rs_dgrad ........................ False + tp_comm_split_ag ................................ True + tp_comm_split_rs ................................ True + train_data_path ................................. None + train_iters ..................................... 10 + train_samples ................................... None + train_sync_interval ............................. None + transformer_impl ................................ transformer_engine + transformer_pipeline_model_parallel_size ........ 1 + untie_embeddings_and_output_weights ............. True + use_checkpoint_args ............................. False + use_checkpoint_opt_param_scheduler .............. False + use_cpu_initialization .......................... None + use_custom_fsdp ................................. False + use_dist_ckpt ................................... True + use_dist_ckpt_deprecated ........................ False + use_distributed_optimizer ....................... True + use_flash_attn .................................. False + use_legacy_models ............................... False + use_mp_args_from_checkpoint_args ................ False + use_one_sent_docs ............................... False + use_persistent_ckpt_worker ...................... False + use_precision_aware_optimizer ................... False + use_pytorch_profiler ............................ False + use_ring_exchange_p2p ........................... False + use_rope_scaling ................................ False + use_rotary_position_embeddings .................. False + use_sharp ....................................... False + use_tokenizer_model_from_checkpoint_args ........ True + use_torch_fsdp2 ................................. False + use_torch_optimizer_for_cpu_offload ............. False + use_tp_pp_dp_mapping ............................ False + v_head_dim ...................................... 128 + valid_data_path ................................. None + variable_seq_lengths ............................ False + virtual_pipeline_model_parallel_size ............ None + vision_backbone_type ............................ vit + vision_pretraining .............................. False + vision_pretraining_type ......................... classify + vocab_extra_ids ................................. 0 + vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json + vocab_size ...................................... None + wandb_exp_name .................................. + wandb_project ................................... + wandb_save_dir .................................. + weight_decay .................................... 0.1 + weight_decay_incr_style ......................... constant + wgrad_deferral_limit ............................ 0 + world_size ...................................... 8 + yaml_cfg ........................................ None +-------------------- end of arguments --------------------- +INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1 +> building GPT2BPETokenizer tokenizer ... +/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled + warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled") + > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304) +WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS +> initializing torch distributed ... +[Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7 +[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0 +[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0 +[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0 +> initialized tensor model parallel with size 1 +> initialized pipeline model parallel with size 1 +> setting random seeds to 1234 ... +[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0 +[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0 +[Gloo] Rank 0 is connected to 0 peer ranks. [Gloo] Rank Expected number of connected peer ranks is : 00 + is connected to 0 peer ranks. Expected number of connected peer ranks is : 0 +[Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0 +[rank3]:[W830 00:17:42.399731980 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device. +[rank6]:[W830 00:17:42.400067803 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device. +[rank5]:[W830 00:17:42.400073006 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device. +[rank1]:[W830 00:17:42.400122579 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device. +[rank2]:[W830 00:17:42.400724722 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device. +[rank7]:[W830 00:17:42.400862540 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device. +[rank4]:[W830 00:17:42.401032783 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device. +> compiling dataset index builder ... +make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets' +make: Nothing to be done for 'default'. +make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets' +>>> done with dataset index builder. Compilation time: 0.425 seconds +WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations. +> compiling and loading fused kernels ... +[rank0]:[W830 00:17:42.015150827 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device. +>>> done with compiling and loading fused kernels. Compilation time: 3.883 seconds +time to initialize megatron (seconds): 9.261 +[after megatron is initialized] datetime: 2026-08-30 00:17:48 +building GPT model ... +/usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( +/usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect. + warnings.warn( + > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 4748687360 +INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=True, overlap_grad_reduce=True, overlap_param_gather=True, align_param_gather=False, use_distributed_optimizer=True, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=40000000, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, reuse_grad_buf_for_mxfp8_param_ag=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False) +INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 21 +Params for bucket 1 (103022592 elements, 103022592 padded size): + module.output_layer.weight +Params for bucket 2 (49033984 elements, 49033984 padded size): + module.decoder.layers.45.pre_mlp_layernorm.weight + module.decoder.layers.45.self_attention.q_layernorm.weight + module.decoder.layers.47.self_attention.linear_qkv.weight + module.decoder.final_layernorm.weight + module.decoder.layers.45.self_attention.linear_qkv.weight + module.decoder.layers.47.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.46.mlp.router.weight + module.decoder.layers.46.pre_mlp_layernorm.weight + module.decoder.layers.47.mlp.router.weight + module.decoder.layers.47.self_attention.k_layernorm.weight + module.decoder.layers.46.self_attention.k_layernorm.weight + module.decoder.layers.46.self_attention.q_layernorm.weight + module.decoder.layers.47.pre_mlp_layernorm.weight + module.decoder.layers.47.self_attention.q_layernorm.weight + module.decoder.layers.47.self_attention.linear_proj.weight + module.decoder.layers.46.self_attention.linear_qkv.weight + module.decoder.layers.46.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.46.self_attention.linear_proj.weight + module.decoder.layers.45.mlp.router.weight + module.decoder.layers.45.self_attention.k_layernorm.weight +Params for bucket 3 (46672384 elements, 46672384 padded size): + module.decoder.layers.44.pre_mlp_layernorm.weight + module.decoder.layers.44.self_attention.q_layernorm.weight + module.decoder.layers.44.self_attention.linear_qkv.weight + module.decoder.layers.43.pre_mlp_layernorm.weight + module.decoder.layers.43.self_attention.q_layernorm.weight + module.decoder.layers.43.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.45.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.44.self_attention.k_layernorm.weight + module.decoder.layers.45.self_attention.linear_proj.weight + module.decoder.layers.44.mlp.router.weight + module.decoder.layers.44.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.44.self_attention.linear_proj.weight + module.decoder.layers.43.mlp.router.weight + module.decoder.layers.43.self_attention.k_layernorm.weight + module.decoder.layers.43.self_attention.linear_qkv.weight + module.decoder.layers.43.self_attention.linear_proj.weight +Params for bucket 4 (49031936 elements, 49031936 padded size): + module.decoder.layers.42.mlp.router.weight + module.decoder.layers.42.self_attention.q_layernorm.weight + module.decoder.layers.42.pre_mlp_layernorm.weight + module.decoder.layers.41.pre_mlp_layernorm.weight + module.decoder.layers.41.self_attention.q_layernorm.weight + module.decoder.layers.41.self_attention.linear_proj.weight + module.decoder.layers.42.self_attention.linear_proj.weight + module.decoder.layers.42.self_attention.linear_qkv.weight + module.decoder.layers.41.self_attention.linear_qkv.weight + module.decoder.layers.40.mlp.router.weight + module.decoder.layers.40.pre_mlp_layernorm.weight + module.decoder.layers.42.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.41.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.40.self_attention.k_layernorm.weight + module.decoder.layers.40.self_attention.linear_qkv.weight + module.decoder.layers.40.self_attention.q_layernorm.weight + module.decoder.layers.42.self_attention.k_layernorm.weight + module.decoder.layers.41.mlp.router.weight + module.decoder.layers.41.self_attention.k_layernorm.weight +Params for bucket 5 (46672384 elements, 46672384 padded size): + module.decoder.layers.40.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.39.self_attention.k_layernorm.weight + module.decoder.layers.40.self_attention.linear_proj.weight + module.decoder.layers.39.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.39.self_attention.linear_proj.weight + module.decoder.layers.38.mlp.router.weight + module.decoder.layers.38.self_attention.k_layernorm.weight + module.decoder.layers.38.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.39.mlp.router.weight + module.decoder.layers.39.pre_mlp_layernorm.weight + module.decoder.layers.39.self_attention.q_layernorm.weight + module.decoder.layers.39.self_attention.linear_qkv.weight + module.decoder.layers.38.pre_mlp_layernorm.weight + module.decoder.layers.38.self_attention.q_layernorm.weight + module.decoder.layers.38.self_attention.linear_qkv.weight + module.decoder.layers.38.self_attention.linear_proj.weight +Params for bucket 6 (49031936 elements, 49031936 padded size): + module.decoder.layers.36.self_attention.k_layernorm.weight + module.decoder.layers.36.mlp.router.weight + module.decoder.layers.36.pre_mlp_layernorm.weight + module.decoder.layers.36.self_attention.q_layernorm.weight + module.decoder.layers.35.mlp.router.weight + module.decoder.layers.35.self_attention.k_layernorm.weight + module.decoder.layers.37.self_attention.q_layernorm.weight + module.decoder.layers.36.self_attention.linear_proj.weight + module.decoder.layers.35.pre_mlp_layernorm.weight + module.decoder.layers.35.self_attention.q_layernorm.weight + module.decoder.layers.37.self_attention.linear_proj.weight + module.decoder.layers.36.self_attention.linear_qkv.weight + module.decoder.layers.37.mlp.router.weight + module.decoder.layers.35.self_attention.linear_qkv.weight + module.decoder.layers.37.pre_mlp_layernorm.weight + module.decoder.layers.37.self_attention.linear_qkv.weight + module.decoder.layers.37.self_attention.k_layernorm.weight + module.decoder.layers.37.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.36.self_attention.linear_qkv.layer_norm_weight +Params for bucket 7 (46672384 elements, 46672384 padded size): + module.decoder.layers.34.mlp.router.weight + module.decoder.layers.35.self_attention.linear_proj.weight + module.decoder.layers.34.self_attention.k_layernorm.weight + module.decoder.layers.34.pre_mlp_layernorm.weight + module.decoder.layers.34.self_attention.q_layernorm.weight + module.decoder.layers.34.self_attention.linear_qkv.weight + module.decoder.layers.34.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.33.mlp.router.weight + module.decoder.layers.33.pre_mlp_layernorm.weight + module.decoder.layers.33.self_attention.q_layernorm.weight + module.decoder.layers.35.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.34.self_attention.linear_proj.weight + module.decoder.layers.33.self_attention.k_layernorm.weight + module.decoder.layers.33.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.33.self_attention.linear_proj.weight + module.decoder.layers.33.self_attention.linear_qkv.weight +Params for bucket 8 (49031936 elements, 49031936 padded size): + module.decoder.layers.32.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.32.self_attention.k_layernorm.weight + module.decoder.layers.30.pre_mlp_layernorm.weight + module.decoder.layers.30.self_attention.k_layernorm.weight + module.decoder.layers.32.pre_mlp_layernorm.weight + module.decoder.layers.32.self_attention.q_layernorm.weight + module.decoder.layers.32.self_attention.linear_proj.weight + module.decoder.layers.31.mlp.router.weight + module.decoder.layers.31.pre_mlp_layernorm.weight + module.decoder.layers.30.mlp.router.weight + module.decoder.layers.30.self_attention.q_layernorm.weight + module.decoder.layers.32.mlp.router.weight + module.decoder.layers.32.self_attention.linear_qkv.weight + module.decoder.layers.31.self_attention.k_layernorm.weight + module.decoder.layers.31.self_attention.linear_qkv.weight + module.decoder.layers.31.self_attention.q_layernorm.weight + module.decoder.layers.31.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.30.self_attention.linear_qkv.weight + module.decoder.layers.31.self_attention.linear_proj.weight +Params for bucket 9 (46672384 elements, 46672384 padded size): + module.decoder.layers.29.self_attention.k_layernorm.weight + module.decoder.layers.29.mlp.router.weight + module.decoder.layers.29.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.28.mlp.router.weight + module.decoder.layers.28.self_attention.k_layernorm.weight + module.decoder.layers.28.self_attention.linear_proj.weight + module.decoder.layers.30.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.29.pre_mlp_layernorm.weight + module.decoder.layers.30.self_attention.linear_proj.weight + module.decoder.layers.29.self_attention.q_layernorm.weight + module.decoder.layers.29.self_attention.linear_qkv.weight + module.decoder.layers.29.self_attention.linear_proj.weight + module.decoder.layers.28.pre_mlp_layernorm.weight + module.decoder.layers.28.self_attention.q_layernorm.weight + module.decoder.layers.28.self_attention.linear_qkv.weight + module.decoder.layers.28.self_attention.linear_qkv.layer_norm_weight +Params for bucket 10 (49031936 elements, 49031936 padded size): + module.decoder.layers.27.pre_mlp_layernorm.weight + module.decoder.layers.26.mlp.router.weight + module.decoder.layers.26.self_attention.k_layernorm.weight + module.decoder.layers.27.self_attention.k_layernorm.weight + module.decoder.layers.26.pre_mlp_layernorm.weight + module.decoder.layers.26.self_attention.q_layernorm.weight + module.decoder.layers.26.self_attention.linear_proj.weight + module.decoder.layers.25.mlp.router.weight + module.decoder.layers.27.self_attention.q_layernorm.weight + module.decoder.layers.25.pre_mlp_layernorm.weight + module.decoder.layers.27.self_attention.linear_proj.weight + module.decoder.layers.26.self_attention.linear_qkv.weight + module.decoder.layers.27.self_attention.linear_qkv.weight + module.decoder.layers.25.self_attention.k_layernorm.weight + module.decoder.layers.25.self_attention.linear_qkv.weight + module.decoder.layers.25.self_attention.q_layernorm.weight + module.decoder.layers.26.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.27.mlp.router.weight + module.decoder.layers.27.self_attention.linear_qkv.layer_norm_weight +Params for bucket 11 (46672384 elements, 46672384 padded size): + module.decoder.layers.24.pre_mlp_layernorm.weight + module.decoder.layers.24.self_attention.k_layernorm.weight + module.decoder.layers.24.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.24.self_attention.linear_proj.weight + module.decoder.layers.23.mlp.router.weight + module.decoder.layers.23.self_attention.k_layernorm.weight + module.decoder.layers.23.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.25.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.24.mlp.router.weight + module.decoder.layers.25.self_attention.linear_proj.weight + module.decoder.layers.24.self_attention.q_layernorm.weight + module.decoder.layers.24.self_attention.linear_qkv.weight + module.decoder.layers.23.pre_mlp_layernorm.weight + module.decoder.layers.23.self_attention.q_layernorm.weight + module.decoder.layers.23.self_attention.linear_qkv.weight + module.decoder.layers.23.self_attention.linear_proj.weight +Params for bucket 12 (49031936 elements, 49031936 padded size): + module.decoder.layers.21.self_attention.linear_qkv.weight + module.decoder.layers.20.pre_mlp_layernorm.weight + module.decoder.layers.20.self_attention.q_layernorm.weight + module.decoder.layers.22.mlp.router.weight + module.decoder.layers.22.pre_mlp_layernorm.weight + module.decoder.layers.20.self_attention.linear_qkv.weight + module.decoder.layers.21.mlp.router.weight + module.decoder.layers.21.self_attention.k_layernorm.weight + module.decoder.layers.21.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.22.self_attention.k_layernorm.weight + module.decoder.layers.22.self_attention.linear_proj.weight + module.decoder.layers.21.pre_mlp_layernorm.weight + module.decoder.layers.22.self_attention.linear_qkv.weight + module.decoder.layers.22.self_attention.q_layernorm.weight + module.decoder.layers.22.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.21.self_attention.q_layernorm.weight + module.decoder.layers.21.self_attention.linear_proj.weight + module.decoder.layers.20.mlp.router.weight + module.decoder.layers.20.self_attention.k_layernorm.weight +Params for bucket 13 (46672384 elements, 46672384 padded size): + module.decoder.layers.19.self_attention.k_layernorm.weight + module.decoder.layers.19.pre_mlp_layernorm.weight + module.decoder.layers.19.self_attention.q_layernorm.weight + module.decoder.layers.19.self_attention.linear_qkv.weight + module.decoder.layers.19.self_attention.linear_proj.weight + module.decoder.layers.18.mlp.router.weight + module.decoder.layers.18.self_attention.k_layernorm.weight + module.decoder.layers.18.self_attention.q_layernorm.weight + module.decoder.layers.18.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.20.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.20.self_attention.linear_proj.weight + module.decoder.layers.19.mlp.router.weight + module.decoder.layers.19.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.18.pre_mlp_layernorm.weight + module.decoder.layers.18.self_attention.linear_qkv.weight + module.decoder.layers.18.self_attention.linear_proj.weight +Params for bucket 14 (49031936 elements, 49031936 padded size): + module.decoder.layers.17.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.16.self_attention.k_layernorm.weight + module.decoder.layers.16.self_attention.linear_qkv.weight + module.decoder.layers.16.mlp.router.weight + module.decoder.layers.15.mlp.router.weight + module.decoder.layers.15.self_attention.k_layernorm.weight + module.decoder.layers.17.mlp.router.weight + module.decoder.layers.17.self_attention.k_layernorm.weight + module.decoder.layers.16.pre_mlp_layernorm.weight + module.decoder.layers.16.self_attention.q_layernorm.weight + module.decoder.layers.16.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.15.pre_mlp_layernorm.weight + module.decoder.layers.15.self_attention.q_layernorm.weight + module.decoder.layers.17.pre_mlp_layernorm.weight + module.decoder.layers.17.self_attention.q_layernorm.weight + module.decoder.layers.17.self_attention.linear_proj.weight + module.decoder.layers.16.self_attention.linear_proj.weight + module.decoder.layers.15.self_attention.linear_qkv.weight + module.decoder.layers.17.self_attention.linear_qkv.weight +Params for bucket 15 (46672384 elements, 46672384 padded size): + module.decoder.layers.14.self_attention.k_layernorm.weight + module.decoder.layers.15.self_attention.linear_proj.weight + module.decoder.layers.14.mlp.router.weight + module.decoder.layers.14.pre_mlp_layernorm.weight + module.decoder.layers.14.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.14.self_attention.linear_proj.weight + module.decoder.layers.13.mlp.router.weight + module.decoder.layers.13.self_attention.k_layernorm.weight + module.decoder.layers.13.self_attention.q_layernorm.weight + module.decoder.layers.13.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.15.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.14.self_attention.q_layernorm.weight + module.decoder.layers.14.self_attention.linear_qkv.weight + module.decoder.layers.13.pre_mlp_layernorm.weight + module.decoder.layers.13.self_attention.linear_qkv.weight + module.decoder.layers.13.self_attention.linear_proj.weight +Params for bucket 16 (49031936 elements, 49031936 padded size): + module.decoder.layers.12.self_attention.linear_qkv.weight + module.decoder.layers.11.self_attention.q_layernorm.weight + module.decoder.layers.12.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.11.self_attention.linear_proj.weight + module.decoder.layers.10.mlp.router.weight + module.decoder.layers.10.self_attention.k_layernorm.weight + module.decoder.layers.10.self_attention.q_layernorm.weight + module.decoder.layers.11.mlp.router.weight + module.decoder.layers.11.pre_mlp_layernorm.weight + module.decoder.layers.10.pre_mlp_layernorm.weight + module.decoder.layers.12.mlp.router.weight + module.decoder.layers.12.self_attention.k_layernorm.weight + module.decoder.layers.10.self_attention.linear_qkv.weight + module.decoder.layers.12.self_attention.q_layernorm.weight + module.decoder.layers.12.pre_mlp_layernorm.weight + module.decoder.layers.12.self_attention.linear_proj.weight + module.decoder.layers.11.self_attention.k_layernorm.weight + module.decoder.layers.11.self_attention.linear_qkv.weight + module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight +Params for bucket 17 (46672384 elements, 46672384 padded size): + module.decoder.layers.10.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.9.pre_mlp_layernorm.weight + module.decoder.layers.10.self_attention.linear_proj.weight + module.decoder.layers.9.self_attention.q_layernorm.weight + module.decoder.layers.9.self_attention.linear_qkv.weight + module.decoder.layers.9.self_attention.linear_proj.weight + module.decoder.layers.8.mlp.router.weight + module.decoder.layers.8.self_attention.k_layernorm.weight + module.decoder.layers.8.self_attention.linear_qkv.weight + module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.8.self_attention.linear_proj.weight + module.decoder.layers.9.self_attention.k_layernorm.weight + module.decoder.layers.9.mlp.router.weight + module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.8.pre_mlp_layernorm.weight + module.decoder.layers.8.self_attention.q_layernorm.weight +Params for bucket 18 (49031936 elements, 49031936 padded size): + module.decoder.layers.6.pre_mlp_layernorm.weight + module.decoder.layers.6.self_attention.q_layernorm.weight + module.decoder.layers.6.self_attention.linear_proj.weight + module.decoder.layers.7.self_attention.k_layernorm.weight + module.decoder.layers.7.mlp.router.weight + module.decoder.layers.7.pre_mlp_layernorm.weight + module.decoder.layers.7.self_attention.q_layernorm.weight + module.decoder.layers.6.self_attention.linear_qkv.weight + module.decoder.layers.7.self_attention.linear_proj.weight + module.decoder.layers.5.mlp.router.weight + module.decoder.layers.7.self_attention.linear_qkv.weight + module.decoder.layers.5.pre_mlp_layernorm.weight + module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.5.self_attention.k_layernorm.weight + module.decoder.layers.5.self_attention.linear_qkv.weight + module.decoder.layers.5.self_attention.q_layernorm.weight + module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.6.mlp.router.weight + module.decoder.layers.6.self_attention.k_layernorm.weight +Params for bucket 19 (46672384 elements, 46672384 padded size): + module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.4.mlp.router.weight + module.decoder.layers.4.self_attention.k_layernorm.weight + module.decoder.layers.5.self_attention.linear_proj.weight + module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.4.self_attention.linear_proj.weight + module.decoder.layers.3.mlp.router.weight + module.decoder.layers.3.self_attention.k_layernorm.weight + module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.4.pre_mlp_layernorm.weight + module.decoder.layers.4.self_attention.q_layernorm.weight + module.decoder.layers.4.self_attention.linear_qkv.weight + module.decoder.layers.3.pre_mlp_layernorm.weight + module.decoder.layers.3.self_attention.q_layernorm.weight + module.decoder.layers.3.self_attention.linear_qkv.weight + module.decoder.layers.3.self_attention.linear_proj.weight +Params for bucket 20 (49031936 elements, 49031936 padded size): + module.decoder.layers.1.self_attention.linear_proj.weight + module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.2.mlp.router.weight + module.decoder.layers.2.self_attention.linear_proj.weight + module.decoder.layers.2.pre_mlp_layernorm.weight + module.decoder.layers.1.mlp.router.weight + module.decoder.layers.1.self_attention.k_layernorm.weight + module.decoder.layers.0.mlp.router.weight + module.decoder.layers.2.self_attention.k_layernorm.weight + module.decoder.layers.1.pre_mlp_layernorm.weight + module.decoder.layers.0.self_attention.linear_qkv.weight + module.decoder.layers.2.self_attention.q_layernorm.weight + module.decoder.layers.2.self_attention.linear_qkv.weight + module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight + module.decoder.layers.0.pre_mlp_layernorm.weight + module.decoder.layers.0.self_attention.q_layernorm.weight + module.decoder.layers.1.self_attention.q_layernorm.weight + module.decoder.layers.1.self_attention.linear_qkv.weight + module.decoder.layers.0.self_attention.k_layernorm.weight +Params for bucket 21 (111413248 elements, 111413248 padded size): + module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight + module.embedding.word_embeddings.weight + module.decoder.layers.0.self_attention.linear_proj.weight +INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 89 +Params for bucket 1 (40894464 elements, 40894464 padded size): + module.decoder.layers.47.mlp.experts.linear_fc2.weight10 + module.decoder.layers.47.mlp.experts.linear_fc2.weight2 + module.decoder.layers.47.mlp.experts.linear_fc2.weight15 + module.decoder.layers.47.mlp.experts.linear_fc2.weight7 + module.decoder.layers.47.mlp.experts.linear_fc1.weight13 + module.decoder.layers.47.mlp.experts.linear_fc2.weight12 + module.decoder.layers.47.mlp.experts.linear_fc2.weight4 + module.decoder.layers.47.mlp.experts.linear_fc2.weight9 + module.decoder.layers.47.mlp.experts.linear_fc2.weight1 + module.decoder.layers.47.mlp.experts.linear_fc1.weight15 + module.decoder.layers.47.mlp.experts.linear_fc2.weight14 + module.decoder.layers.47.mlp.experts.linear_fc2.weight6 + module.decoder.layers.47.mlp.experts.linear_fc1.weight12 + module.decoder.layers.47.mlp.experts.linear_fc2.weight11 + module.decoder.layers.47.mlp.experts.linear_fc2.weight3 + module.decoder.layers.47.mlp.experts.linear_fc2.weight8 + module.decoder.layers.47.mlp.experts.linear_fc2.weight0 + module.decoder.layers.47.mlp.experts.linear_fc1.weight14 + module.decoder.layers.47.mlp.experts.linear_fc2.weight13 + module.decoder.layers.47.mlp.experts.linear_fc2.weight5 + module.decoder.layers.47.mlp.experts.linear_fc1.weight11 +Params for bucket 2 (40894464 elements, 40894464 padded size): + module.decoder.layers.47.mlp.experts.linear_fc1.weight10 + module.decoder.layers.47.mlp.experts.linear_fc1.weight8 + module.decoder.layers.47.mlp.experts.linear_fc1.weight6 + module.decoder.layers.47.mlp.experts.linear_fc1.weight4 + module.decoder.layers.47.mlp.experts.linear_fc1.weight2 + module.decoder.layers.47.mlp.experts.linear_fc1.weight0 + module.decoder.layers.46.mlp.experts.linear_fc2.weight14 + module.decoder.layers.46.mlp.experts.linear_fc2.weight12 + module.decoder.layers.47.mlp.experts.linear_fc1.weight9 + module.decoder.layers.47.mlp.experts.linear_fc1.weight7 + module.decoder.layers.47.mlp.experts.linear_fc1.weight5 + module.decoder.layers.47.mlp.experts.linear_fc1.weight3 + module.decoder.layers.47.mlp.experts.linear_fc1.weight1 + module.decoder.layers.46.mlp.experts.linear_fc2.weight15 + module.decoder.layers.46.mlp.experts.linear_fc2.weight13 +Params for bucket 3 (40894464 elements, 40894464 padded size): + module.decoder.layers.46.mlp.experts.linear_fc2.weight10 + module.decoder.layers.46.mlp.experts.linear_fc2.weight2 + module.decoder.layers.46.mlp.experts.linear_fc1.weight13 + module.decoder.layers.46.mlp.experts.linear_fc2.weight7 + module.decoder.layers.46.mlp.experts.linear_fc1.weight10 + module.decoder.layers.46.mlp.experts.linear_fc2.weight4 + module.decoder.layers.46.mlp.experts.linear_fc1.weight15 + module.decoder.layers.46.mlp.experts.linear_fc2.weight9 + module.decoder.layers.46.mlp.experts.linear_fc2.weight1 + module.decoder.layers.46.mlp.experts.linear_fc1.weight12 + module.decoder.layers.46.mlp.experts.linear_fc2.weight6 + module.decoder.layers.46.mlp.experts.linear_fc1.weight9 + module.decoder.layers.46.mlp.experts.linear_fc2.weight11 + module.decoder.layers.46.mlp.experts.linear_fc2.weight3 + module.decoder.layers.46.mlp.experts.linear_fc1.weight14 + module.decoder.layers.46.mlp.experts.linear_fc2.weight8 + module.decoder.layers.46.mlp.experts.linear_fc2.weight0 + module.decoder.layers.46.mlp.experts.linear_fc1.weight11 + module.decoder.layers.46.mlp.experts.linear_fc2.weight5 +Params for bucket 4 (40894464 elements, 40894464 padded size): + module.decoder.layers.46.mlp.experts.linear_fc1.weight7 + module.decoder.layers.46.mlp.experts.linear_fc1.weight5 + module.decoder.layers.46.mlp.experts.linear_fc1.weight3 + module.decoder.layers.46.mlp.experts.linear_fc1.weight1 + module.decoder.layers.45.mlp.experts.linear_fc2.weight15 + module.decoder.layers.45.mlp.experts.linear_fc2.weight13 + module.decoder.layers.45.mlp.experts.linear_fc2.weight11 + module.decoder.layers.45.mlp.experts.linear_fc2.weight9 + module.decoder.layers.46.mlp.experts.linear_fc1.weight8 + module.decoder.layers.46.mlp.experts.linear_fc1.weight4 + module.decoder.layers.46.mlp.experts.linear_fc1.weight6 + module.decoder.layers.46.mlp.experts.linear_fc1.weight2 + module.decoder.layers.46.mlp.experts.linear_fc1.weight0 + module.decoder.layers.45.mlp.experts.linear_fc2.weight14 + module.decoder.layers.45.mlp.experts.linear_fc2.weight12 + module.decoder.layers.45.mlp.experts.linear_fc2.weight10 + module.decoder.layers.45.mlp.experts.linear_fc2.weight8 +Params for bucket 5 (40894464 elements, 40894464 padded size): + module.decoder.layers.45.mlp.experts.linear_fc2.weight7 + module.decoder.layers.45.mlp.experts.linear_fc2.weight5 + module.decoder.layers.45.mlp.experts.linear_fc2.weight3 + module.decoder.layers.45.mlp.experts.linear_fc2.weight1 + module.decoder.layers.45.mlp.experts.linear_fc1.weight14 + module.decoder.layers.45.mlp.experts.linear_fc1.weight12 + module.decoder.layers.45.mlp.experts.linear_fc1.weight10 + module.decoder.layers.45.mlp.experts.linear_fc1.weight8 + module.decoder.layers.45.mlp.experts.linear_fc2.weight6 + module.decoder.layers.45.mlp.experts.linear_fc2.weight4 + module.decoder.layers.45.mlp.experts.linear_fc2.weight2 + module.decoder.layers.45.mlp.experts.linear_fc2.weight0 + module.decoder.layers.45.mlp.experts.linear_fc1.weight15 + module.decoder.layers.45.mlp.experts.linear_fc1.weight13 + module.decoder.layers.45.mlp.experts.linear_fc1.weight11 + module.decoder.layers.45.mlp.experts.linear_fc1.weight9 + module.decoder.layers.45.mlp.experts.linear_fc1.weight7 +Params for bucket 6 (40894464 elements, 40894464 padded size): + module.decoder.layers.45.mlp.experts.linear_fc1.weight0 + module.decoder.layers.44.mlp.experts.linear_fc2.weight11 + module.decoder.layers.45.mlp.experts.linear_fc1.weight5 + module.decoder.layers.44.mlp.experts.linear_fc2.weight8 + module.decoder.layers.45.mlp.experts.linear_fc1.weight2 + module.decoder.layers.44.mlp.experts.linear_fc2.weight13 + module.decoder.layers.44.mlp.experts.linear_fc2.weight5 + module.decoder.layers.44.mlp.experts.linear_fc2.weight10 + module.decoder.layers.45.mlp.experts.linear_fc1.weight4 + module.decoder.layers.44.mlp.experts.linear_fc2.weight15 + module.decoder.layers.44.mlp.experts.linear_fc2.weight7 + module.decoder.layers.45.mlp.experts.linear_fc1.weight1 + module.decoder.layers.44.mlp.experts.linear_fc2.weight12 + module.decoder.layers.44.mlp.experts.linear_fc2.weight4 + module.decoder.layers.45.mlp.experts.linear_fc1.weight6 + module.decoder.layers.44.mlp.experts.linear_fc2.weight9 + module.decoder.layers.45.mlp.experts.linear_fc1.weight3 + module.decoder.layers.44.mlp.experts.linear_fc2.weight14 + module.decoder.layers.44.mlp.experts.linear_fc2.weight6 +Params for bucket 7 (40894464 elements, 40894464 padded size): + module.decoder.layers.44.mlp.experts.linear_fc2.weight3 + module.decoder.layers.44.mlp.experts.linear_fc2.weight1 + module.decoder.layers.44.mlp.experts.linear_fc1.weight14 + module.decoder.layers.44.mlp.experts.linear_fc1.weight12 + module.decoder.layers.44.mlp.experts.linear_fc1.weight10 + module.decoder.layers.44.mlp.experts.linear_fc1.weight8 + module.decoder.layers.44.mlp.experts.linear_fc1.weight6 + module.decoder.layers.44.mlp.experts.linear_fc2.weight2 + module.decoder.layers.44.mlp.experts.linear_fc1.weight15 + module.decoder.layers.44.mlp.experts.linear_fc2.weight0 + module.decoder.layers.44.mlp.experts.linear_fc1.weight13 + module.decoder.layers.44.mlp.experts.linear_fc1.weight11 + module.decoder.layers.44.mlp.experts.linear_fc1.weight9 + module.decoder.layers.44.mlp.experts.linear_fc1.weight7 + module.decoder.layers.44.mlp.experts.linear_fc1.weight5 +Params for bucket 8 (40894464 elements, 40894464 padded size): + module.decoder.layers.44.mlp.experts.linear_fc1.weight2 + module.decoder.layers.43.mlp.experts.linear_fc2.weight12 + module.decoder.layers.43.mlp.experts.linear_fc2.weight4 + module.decoder.layers.43.mlp.experts.linear_fc2.weight9 + module.decoder.layers.43.mlp.experts.linear_fc2.weight1 + module.decoder.layers.44.mlp.experts.linear_fc1.weight4 + module.decoder.layers.43.mlp.experts.linear_fc2.weight14 + module.decoder.layers.43.mlp.experts.linear_fc2.weight6 + module.decoder.layers.44.mlp.experts.linear_fc1.weight1 + module.decoder.layers.43.mlp.experts.linear_fc2.weight11 + module.decoder.layers.43.mlp.experts.linear_fc2.weight3 + module.decoder.layers.43.mlp.experts.linear_fc2.weight8 + module.decoder.layers.43.mlp.experts.linear_fc2.weight0 + module.decoder.layers.44.mlp.experts.linear_fc1.weight3 + module.decoder.layers.43.mlp.experts.linear_fc2.weight13 + module.decoder.layers.43.mlp.experts.linear_fc2.weight5 + module.decoder.layers.44.mlp.experts.linear_fc1.weight0 + module.decoder.layers.43.mlp.experts.linear_fc2.weight10 + module.decoder.layers.43.mlp.experts.linear_fc2.weight2 + module.decoder.layers.43.mlp.experts.linear_fc2.weight15 + module.decoder.layers.43.mlp.experts.linear_fc2.weight7 +Params for bucket 9 (40894464 elements, 40894464 padded size): + module.decoder.layers.43.mlp.experts.linear_fc1.weight14 + module.decoder.layers.43.mlp.experts.linear_fc1.weight12 + module.decoder.layers.43.mlp.experts.linear_fc1.weight10 + module.decoder.layers.43.mlp.experts.linear_fc1.weight8 + module.decoder.layers.43.mlp.experts.linear_fc1.weight6 + module.decoder.layers.43.mlp.experts.linear_fc1.weight4 + module.decoder.layers.43.mlp.experts.linear_fc1.weight15 + module.decoder.layers.43.mlp.experts.linear_fc1.weight13 + module.decoder.layers.43.mlp.experts.linear_fc1.weight11 + module.decoder.layers.43.mlp.experts.linear_fc1.weight9 + module.decoder.layers.43.mlp.experts.linear_fc1.weight7 + module.decoder.layers.43.mlp.experts.linear_fc1.weight5 + module.decoder.layers.43.mlp.experts.linear_fc1.weight3 +Params for bucket 10 (40894464 elements, 40894464 padded size): + module.decoder.layers.42.mlp.experts.linear_fc2.weight5 + module.decoder.layers.42.mlp.experts.linear_fc2.weight14 + module.decoder.layers.42.mlp.experts.linear_fc2.weight2 + module.decoder.layers.43.mlp.experts.linear_fc1.weight0 + module.decoder.layers.42.mlp.experts.linear_fc2.weight11 + module.decoder.layers.42.mlp.experts.linear_fc2.weight7 + module.decoder.layers.42.mlp.experts.linear_fc2.weight4 + module.decoder.layers.42.mlp.experts.linear_fc1.weight15 + module.decoder.layers.43.mlp.experts.linear_fc1.weight2 + module.decoder.layers.42.mlp.experts.linear_fc2.weight13 + module.decoder.layers.42.mlp.experts.linear_fc2.weight9 + module.decoder.layers.42.mlp.experts.linear_fc2.weight1 + module.decoder.layers.42.mlp.experts.linear_fc2.weight10 + module.decoder.layers.42.mlp.experts.linear_fc2.weight6 + module.decoder.layers.42.mlp.experts.linear_fc2.weight15 + module.decoder.layers.42.mlp.experts.linear_fc2.weight3 + module.decoder.layers.42.mlp.experts.linear_fc1.weight14 + module.decoder.layers.43.mlp.experts.linear_fc1.weight1 + module.decoder.layers.42.mlp.experts.linear_fc2.weight12 + module.decoder.layers.42.mlp.experts.linear_fc2.weight8 + module.decoder.layers.42.mlp.experts.linear_fc2.weight0 +Params for bucket 11 (40894464 elements, 40894464 padded size): + module.decoder.layers.42.mlp.experts.linear_fc1.weight10 + module.decoder.layers.42.mlp.experts.linear_fc1.weight12 + module.decoder.layers.42.mlp.experts.linear_fc1.weight8 + module.decoder.layers.42.mlp.experts.linear_fc1.weight6 + module.decoder.layers.42.mlp.experts.linear_fc1.weight4 + module.decoder.layers.42.mlp.experts.linear_fc1.weight2 + module.decoder.layers.42.mlp.experts.linear_fc1.weight13 + module.decoder.layers.42.mlp.experts.linear_fc1.weight11 + module.decoder.layers.42.mlp.experts.linear_fc1.weight9 + module.decoder.layers.42.mlp.experts.linear_fc1.weight7 + module.decoder.layers.42.mlp.experts.linear_fc1.weight5 + module.decoder.layers.42.mlp.experts.linear_fc1.weight3 + module.decoder.layers.42.mlp.experts.linear_fc1.weight1 +Params for bucket 12 (40894464 elements, 40894464 padded size): + module.decoder.layers.42.mlp.experts.linear_fc1.weight0 + module.decoder.layers.41.mlp.experts.linear_fc2.weight12 + module.decoder.layers.41.mlp.experts.linear_fc2.weight4 + module.decoder.layers.41.mlp.experts.linear_fc1.weight15 + module.decoder.layers.41.mlp.experts.linear_fc2.weight9 + module.decoder.layers.41.mlp.experts.linear_fc2.weight1 + module.decoder.layers.41.mlp.experts.linear_fc1.weight12 + module.decoder.layers.41.mlp.experts.linear_fc2.weight14 + module.decoder.layers.41.mlp.experts.linear_fc2.weight6 + module.decoder.layers.41.mlp.experts.linear_fc2.weight11 + module.decoder.layers.41.mlp.experts.linear_fc2.weight3 + module.decoder.layers.41.mlp.experts.linear_fc1.weight14 + module.decoder.layers.41.mlp.experts.linear_fc2.weight8 + module.decoder.layers.41.mlp.experts.linear_fc2.weight0 + module.decoder.layers.41.mlp.experts.linear_fc2.weight13 + module.decoder.layers.41.mlp.experts.linear_fc2.weight5 + module.decoder.layers.41.mlp.experts.linear_fc2.weight10 + module.decoder.layers.41.mlp.experts.linear_fc2.weight2 + module.decoder.layers.41.mlp.experts.linear_fc1.weight13 + module.decoder.layers.41.mlp.experts.linear_fc2.weight15 + module.decoder.layers.41.mlp.experts.linear_fc2.weight7 +Params for bucket 13 (40894464 elements, 40894464 padded size): + module.decoder.layers.41.mlp.experts.linear_fc1.weight11 + module.decoder.layers.41.mlp.experts.linear_fc1.weight9 + module.decoder.layers.41.mlp.experts.linear_fc1.weight7 + module.decoder.layers.41.mlp.experts.linear_fc1.weight5 + module.decoder.layers.41.mlp.experts.linear_fc1.weight3 + module.decoder.layers.41.mlp.experts.linear_fc1.weight1 + module.decoder.layers.40.mlp.experts.linear_fc2.weight15 + module.decoder.layers.41.mlp.experts.linear_fc1.weight10 + module.decoder.layers.41.mlp.experts.linear_fc1.weight8 + module.decoder.layers.41.mlp.experts.linear_fc1.weight6 + module.decoder.layers.41.mlp.experts.linear_fc1.weight4 + module.decoder.layers.41.mlp.experts.linear_fc1.weight2 + module.decoder.layers.41.mlp.experts.linear_fc1.weight0 + module.decoder.layers.40.mlp.experts.linear_fc2.weight14 +Params for bucket 14 (40894464 elements, 40894464 padded size): + module.decoder.layers.40.mlp.experts.linear_fc2.weight13 + module.decoder.layers.40.mlp.experts.linear_fc2.weight5 + module.decoder.layers.40.mlp.experts.linear_fc2.weight10 + module.decoder.layers.40.mlp.experts.linear_fc2.weight2 + module.decoder.layers.40.mlp.experts.linear_fc1.weight13 + module.decoder.layers.40.mlp.experts.linear_fc2.weight7 + module.decoder.layers.40.mlp.experts.linear_fc1.weight10 + module.decoder.layers.40.mlp.experts.linear_fc2.weight12 + module.decoder.layers.40.mlp.experts.linear_fc2.weight4 + module.decoder.layers.40.mlp.experts.linear_fc1.weight15 + module.decoder.layers.40.mlp.experts.linear_fc2.weight9 + module.decoder.layers.40.mlp.experts.linear_fc2.weight1 + module.decoder.layers.40.mlp.experts.linear_fc1.weight12 + module.decoder.layers.40.mlp.experts.linear_fc2.weight6 + module.decoder.layers.40.mlp.experts.linear_fc2.weight11 + module.decoder.layers.40.mlp.experts.linear_fc2.weight3 + module.decoder.layers.40.mlp.experts.linear_fc1.weight14 + module.decoder.layers.40.mlp.experts.linear_fc2.weight8 + module.decoder.layers.40.mlp.experts.linear_fc2.weight0 + module.decoder.layers.40.mlp.experts.linear_fc1.weight11 +Params for bucket 15 (40894464 elements, 40894464 padded size): + module.decoder.layers.40.mlp.experts.linear_fc1.weight8 + module.decoder.layers.40.mlp.experts.linear_fc1.weight6 + module.decoder.layers.40.mlp.experts.linear_fc1.weight4 + module.decoder.layers.40.mlp.experts.linear_fc1.weight2 + module.decoder.layers.40.mlp.experts.linear_fc1.weight0 + module.decoder.layers.39.mlp.experts.linear_fc2.weight14 + module.decoder.layers.39.mlp.experts.linear_fc2.weight12 + module.decoder.layers.39.mlp.experts.linear_fc2.weight10 + module.decoder.layers.40.mlp.experts.linear_fc1.weight9 + module.decoder.layers.40.mlp.experts.linear_fc1.weight7 + module.decoder.layers.40.mlp.experts.linear_fc1.weight5 + module.decoder.layers.40.mlp.experts.linear_fc1.weight3 + module.decoder.layers.40.mlp.experts.linear_fc1.weight1 + module.decoder.layers.39.mlp.experts.linear_fc2.weight15 + module.decoder.layers.39.mlp.experts.linear_fc2.weight13 + module.decoder.layers.39.mlp.experts.linear_fc2.weight11 +Params for bucket 16 (40894464 elements, 40894464 padded size): + module.decoder.layers.39.mlp.experts.linear_fc2.weight8 + module.decoder.layers.39.mlp.experts.linear_fc2.weight6 + module.decoder.layers.39.mlp.experts.linear_fc2.weight4 + module.decoder.layers.39.mlp.experts.linear_fc2.weight2 + module.decoder.layers.39.mlp.experts.linear_fc2.weight0 + module.decoder.layers.39.mlp.experts.linear_fc1.weight15 + module.decoder.layers.39.mlp.experts.linear_fc1.weight13 + module.decoder.layers.39.mlp.experts.linear_fc1.weight11 + module.decoder.layers.39.mlp.experts.linear_fc1.weight9 + module.decoder.layers.39.mlp.experts.linear_fc2.weight9 + module.decoder.layers.39.mlp.experts.linear_fc2.weight7 + module.decoder.layers.39.mlp.experts.linear_fc2.weight5 + module.decoder.layers.39.mlp.experts.linear_fc2.weight3 + module.decoder.layers.39.mlp.experts.linear_fc2.weight1 + module.decoder.layers.39.mlp.experts.linear_fc1.weight14 + module.decoder.layers.39.mlp.experts.linear_fc1.weight12 + module.decoder.layers.39.mlp.experts.linear_fc1.weight10 + module.decoder.layers.39.mlp.experts.linear_fc1.weight8 +Params for bucket 17 (40894464 elements, 40894464 padded size): + module.decoder.layers.39.mlp.experts.linear_fc1.weight7 + module.decoder.layers.39.mlp.experts.linear_fc1.weight5 + module.decoder.layers.39.mlp.experts.linear_fc1.weight3 + module.decoder.layers.39.mlp.experts.linear_fc1.weight1 + module.decoder.layers.38.mlp.experts.linear_fc2.weight15 + module.decoder.layers.38.mlp.experts.linear_fc2.weight13 + module.decoder.layers.38.mlp.experts.linear_fc2.weight11 + module.decoder.layers.38.mlp.experts.linear_fc2.weight9 + module.decoder.layers.38.mlp.experts.linear_fc2.weight7 + module.decoder.layers.39.mlp.experts.linear_fc1.weight6 + module.decoder.layers.39.mlp.experts.linear_fc1.weight4 + module.decoder.layers.39.mlp.experts.linear_fc1.weight2 + module.decoder.layers.39.mlp.experts.linear_fc1.weight0 + module.decoder.layers.38.mlp.experts.linear_fc2.weight14 + module.decoder.layers.38.mlp.experts.linear_fc2.weight12 + module.decoder.layers.38.mlp.experts.linear_fc2.weight10 + module.decoder.layers.38.mlp.experts.linear_fc2.weight8 + module.decoder.layers.38.mlp.experts.linear_fc2.weight6 +Params for bucket 18 (40894464 elements, 40894464 padded size): + module.decoder.layers.38.mlp.experts.linear_fc2.weight5 + module.decoder.layers.38.mlp.experts.linear_fc2.weight3 + module.decoder.layers.38.mlp.experts.linear_fc2.weight1 + module.decoder.layers.38.mlp.experts.linear_fc1.weight14 + module.decoder.layers.38.mlp.experts.linear_fc1.weight12 + module.decoder.layers.38.mlp.experts.linear_fc1.weight10 + module.decoder.layers.38.mlp.experts.linear_fc1.weight8 + module.decoder.layers.38.mlp.experts.linear_fc1.weight6 + module.decoder.layers.38.mlp.experts.linear_fc2.weight4 + module.decoder.layers.38.mlp.experts.linear_fc2.weight2 + module.decoder.layers.38.mlp.experts.linear_fc2.weight0 + module.decoder.layers.38.mlp.experts.linear_fc1.weight15 + module.decoder.layers.38.mlp.experts.linear_fc1.weight13 + module.decoder.layers.38.mlp.experts.linear_fc1.weight11 + module.decoder.layers.38.mlp.experts.linear_fc1.weight9 + module.decoder.layers.38.mlp.experts.linear_fc1.weight7 +Params for bucket 19 (40894464 elements, 40894464 padded size): + module.decoder.layers.37.mlp.experts.linear_fc2.weight10 + module.decoder.layers.37.mlp.experts.linear_fc2.weight2 + module.decoder.layers.37.mlp.experts.linear_fc2.weight15 + module.decoder.layers.37.mlp.experts.linear_fc2.weight7 + module.decoder.layers.38.mlp.experts.linear_fc1.weight4 + module.decoder.layers.38.mlp.experts.linear_fc1.weight0 + module.decoder.layers.37.mlp.experts.linear_fc2.weight12 + module.decoder.layers.37.mlp.experts.linear_fc2.weight4 + module.decoder.layers.37.mlp.experts.linear_fc2.weight9 + module.decoder.layers.38.mlp.experts.linear_fc1.weight2 + module.decoder.layers.37.mlp.experts.linear_fc2.weight14 + module.decoder.layers.37.mlp.experts.linear_fc2.weight6 + module.decoder.layers.38.mlp.experts.linear_fc1.weight3 + module.decoder.layers.37.mlp.experts.linear_fc2.weight11 + module.decoder.layers.37.mlp.experts.linear_fc2.weight3 + module.decoder.layers.37.mlp.experts.linear_fc2.weight8 + module.decoder.layers.38.mlp.experts.linear_fc1.weight5 + module.decoder.layers.38.mlp.experts.linear_fc1.weight1 + module.decoder.layers.37.mlp.experts.linear_fc2.weight13 + module.decoder.layers.37.mlp.experts.linear_fc2.weight5 +Params for bucket 20 (40894464 elements, 40894464 padded size): + module.decoder.layers.37.mlp.experts.linear_fc1.weight13 + module.decoder.layers.37.mlp.experts.linear_fc2.weight0 + module.decoder.layers.37.mlp.experts.linear_fc1.weight15 + module.decoder.layers.37.mlp.experts.linear_fc1.weight11 + module.decoder.layers.37.mlp.experts.linear_fc1.weight9 + module.decoder.layers.37.mlp.experts.linear_fc1.weight7 + module.decoder.layers.37.mlp.experts.linear_fc1.weight5 + module.decoder.layers.37.mlp.experts.linear_fc2.weight1 + module.decoder.layers.37.mlp.experts.linear_fc1.weight14 + module.decoder.layers.37.mlp.experts.linear_fc1.weight12 + module.decoder.layers.37.mlp.experts.linear_fc1.weight10 + module.decoder.layers.37.mlp.experts.linear_fc1.weight8 + module.decoder.layers.37.mlp.experts.linear_fc1.weight6 + module.decoder.layers.37.mlp.experts.linear_fc1.weight4 +Params for bucket 21 (40894464 elements, 40894464 padded size): + module.decoder.layers.36.mlp.experts.linear_fc2.weight13 + module.decoder.layers.36.mlp.experts.linear_fc2.weight5 + module.decoder.layers.37.mlp.experts.linear_fc1.weight2 + module.decoder.layers.36.mlp.experts.linear_fc2.weight10 + module.decoder.layers.36.mlp.experts.linear_fc2.weight2 + module.decoder.layers.36.mlp.experts.linear_fc2.weight15 + module.decoder.layers.36.mlp.experts.linear_fc2.weight7 + module.decoder.layers.36.mlp.experts.linear_fc2.weight12 + module.decoder.layers.36.mlp.experts.linear_fc2.weight4 + module.decoder.layers.37.mlp.experts.linear_fc1.weight1 + module.decoder.layers.36.mlp.experts.linear_fc2.weight9 + module.decoder.layers.36.mlp.experts.linear_fc2.weight1 + module.decoder.layers.36.mlp.experts.linear_fc2.weight14 + module.decoder.layers.36.mlp.experts.linear_fc2.weight6 + module.decoder.layers.37.mlp.experts.linear_fc1.weight3 + module.decoder.layers.36.mlp.experts.linear_fc2.weight11 + module.decoder.layers.36.mlp.experts.linear_fc2.weight3 + module.decoder.layers.37.mlp.experts.linear_fc1.weight0 + module.decoder.layers.36.mlp.experts.linear_fc2.weight8 + module.decoder.layers.36.mlp.experts.linear_fc2.weight0 + module.decoder.layers.36.mlp.experts.linear_fc1.weight15 +Params for bucket 22 (40894464 elements, 40894464 padded size): + module.decoder.layers.36.mlp.experts.linear_fc1.weight14 + module.decoder.layers.36.mlp.experts.linear_fc1.weight10 + module.decoder.layers.36.mlp.experts.linear_fc1.weight12 + module.decoder.layers.36.mlp.experts.linear_fc1.weight8 + module.decoder.layers.36.mlp.experts.linear_fc1.weight6 + module.decoder.layers.36.mlp.experts.linear_fc1.weight4 + module.decoder.layers.36.mlp.experts.linear_fc1.weight2 + module.decoder.layers.36.mlp.experts.linear_fc1.weight11 + module.decoder.layers.36.mlp.experts.linear_fc1.weight13 + module.decoder.layers.36.mlp.experts.linear_fc1.weight9 + module.decoder.layers.36.mlp.experts.linear_fc1.weight7 + module.decoder.layers.36.mlp.experts.linear_fc1.weight5 + module.decoder.layers.36.mlp.experts.linear_fc1.weight3 +Params for bucket 23 (40894464 elements, 40894464 padded size): + module.decoder.layers.35.mlp.experts.linear_fc2.weight8 + module.decoder.layers.35.mlp.experts.linear_fc2.weight0 + module.decoder.layers.36.mlp.experts.linear_fc1.weight1 + module.decoder.layers.35.mlp.experts.linear_fc2.weight13 + module.decoder.layers.35.mlp.experts.linear_fc2.weight5 + module.decoder.layers.35.mlp.experts.linear_fc2.weight10 + module.decoder.layers.35.mlp.experts.linear_fc2.weight2 + module.decoder.layers.35.mlp.experts.linear_fc1.weight13 + module.decoder.layers.35.mlp.experts.linear_fc2.weight15 + module.decoder.layers.35.mlp.experts.linear_fc2.weight7 + module.decoder.layers.36.mlp.experts.linear_fc1.weight0 + module.decoder.layers.35.mlp.experts.linear_fc2.weight12 + module.decoder.layers.35.mlp.experts.linear_fc2.weight4 + module.decoder.layers.35.mlp.experts.linear_fc1.weight15 + module.decoder.layers.35.mlp.experts.linear_fc2.weight9 + module.decoder.layers.35.mlp.experts.linear_fc2.weight1 + module.decoder.layers.35.mlp.experts.linear_fc2.weight14 + module.decoder.layers.35.mlp.experts.linear_fc2.weight6 + module.decoder.layers.35.mlp.experts.linear_fc2.weight11 + module.decoder.layers.35.mlp.experts.linear_fc2.weight3 + module.decoder.layers.35.mlp.experts.linear_fc1.weight14 +Params for bucket 24 (40894464 elements, 40894464 padded size): + module.decoder.layers.35.mlp.experts.linear_fc1.weight11 + module.decoder.layers.35.mlp.experts.linear_fc1.weight9 + module.decoder.layers.35.mlp.experts.linear_fc1.weight7 + module.decoder.layers.35.mlp.experts.linear_fc1.weight5 + module.decoder.layers.35.mlp.experts.linear_fc1.weight3 + module.decoder.layers.35.mlp.experts.linear_fc1.weight1 + module.decoder.layers.35.mlp.experts.linear_fc1.weight12 + module.decoder.layers.35.mlp.experts.linear_fc1.weight10 + module.decoder.layers.35.mlp.experts.linear_fc1.weight8 + module.decoder.layers.35.mlp.experts.linear_fc1.weight6 + module.decoder.layers.35.mlp.experts.linear_fc1.weight4 + module.decoder.layers.35.mlp.experts.linear_fc1.weight2 + module.decoder.layers.35.mlp.experts.linear_fc1.weight0 +Params for bucket 25 (40894464 elements, 40894464 padded size): + module.decoder.layers.34.mlp.experts.linear_fc2.weight11 + module.decoder.layers.34.mlp.experts.linear_fc2.weight3 + module.decoder.layers.34.mlp.experts.linear_fc1.weight14 + module.decoder.layers.34.mlp.experts.linear_fc2.weight8 + module.decoder.layers.34.mlp.experts.linear_fc2.weight0 + module.decoder.layers.34.mlp.experts.linear_fc1.weight11 + module.decoder.layers.34.mlp.experts.linear_fc2.weight13 + module.decoder.layers.34.mlp.experts.linear_fc2.weight5 + module.decoder.layers.34.mlp.experts.linear_fc2.weight10 + module.decoder.layers.34.mlp.experts.linear_fc2.weight2 + module.decoder.layers.34.mlp.experts.linear_fc1.weight13 + module.decoder.layers.34.mlp.experts.linear_fc2.weight15 + module.decoder.layers.34.mlp.experts.linear_fc2.weight7 + module.decoder.layers.34.mlp.experts.linear_fc2.weight12 + module.decoder.layers.34.mlp.experts.linear_fc2.weight4 + module.decoder.layers.34.mlp.experts.linear_fc1.weight15 + module.decoder.layers.34.mlp.experts.linear_fc2.weight9 + module.decoder.layers.34.mlp.experts.linear_fc2.weight1 + module.decoder.layers.34.mlp.experts.linear_fc1.weight12 + module.decoder.layers.34.mlp.experts.linear_fc2.weight14 + module.decoder.layers.34.mlp.experts.linear_fc2.weight6 +Params for bucket 26 (40894464 elements, 40894464 padded size): + module.decoder.layers.34.mlp.experts.linear_fc1.weight10 + module.decoder.layers.34.mlp.experts.linear_fc1.weight6 + module.decoder.layers.34.mlp.experts.linear_fc1.weight8 + module.decoder.layers.34.mlp.experts.linear_fc1.weight4 + module.decoder.layers.34.mlp.experts.linear_fc1.weight2 + module.decoder.layers.34.mlp.experts.linear_fc1.weight0 + module.decoder.layers.33.mlp.experts.linear_fc2.weight14 + module.decoder.layers.33.mlp.experts.linear_fc2.weight12 + module.decoder.layers.34.mlp.experts.linear_fc1.weight7 + module.decoder.layers.34.mlp.experts.linear_fc1.weight9 + module.decoder.layers.34.mlp.experts.linear_fc1.weight5 + module.decoder.layers.34.mlp.experts.linear_fc1.weight3 + module.decoder.layers.34.mlp.experts.linear_fc1.weight1 + module.decoder.layers.33.mlp.experts.linear_fc2.weight15 + module.decoder.layers.33.mlp.experts.linear_fc2.weight13 +Params for bucket 27 (40894464 elements, 40894464 padded size): + module.decoder.layers.33.mlp.experts.linear_fc2.weight10 + module.decoder.layers.33.mlp.experts.linear_fc2.weight2 + module.decoder.layers.33.mlp.experts.linear_fc2.weight7 + module.decoder.layers.33.mlp.experts.linear_fc1.weight13 + module.decoder.layers.33.mlp.experts.linear_fc2.weight4 + module.decoder.layers.33.mlp.experts.linear_fc1.weight10 + module.decoder.layers.33.mlp.experts.linear_fc2.weight9 + module.decoder.layers.33.mlp.experts.linear_fc2.weight1 + module.decoder.layers.33.mlp.experts.linear_fc1.weight15 + module.decoder.layers.33.mlp.experts.linear_fc2.weight6 + module.decoder.layers.33.mlp.experts.linear_fc1.weight12 + module.decoder.layers.33.mlp.experts.linear_fc2.weight11 + module.decoder.layers.33.mlp.experts.linear_fc2.weight3 + module.decoder.layers.33.mlp.experts.linear_fc1.weight9 + module.decoder.layers.33.mlp.experts.linear_fc2.weight8 + module.decoder.layers.33.mlp.experts.linear_fc2.weight0 + module.decoder.layers.33.mlp.experts.linear_fc1.weight14 + module.decoder.layers.33.mlp.experts.linear_fc2.weight5 + module.decoder.layers.33.mlp.experts.linear_fc1.weight11 +Params for bucket 28 (40894464 elements, 40894464 padded size): + module.decoder.layers.33.mlp.experts.linear_fc1.weight8 + module.decoder.layers.33.mlp.experts.linear_fc1.weight4 + module.decoder.layers.33.mlp.experts.linear_fc1.weight6 + module.decoder.layers.33.mlp.experts.linear_fc1.weight2 + module.decoder.layers.33.mlp.experts.linear_fc1.weight0 + module.decoder.layers.32.mlp.experts.linear_fc2.weight14 + module.decoder.layers.32.mlp.experts.linear_fc2.weight12 + module.decoder.layers.32.mlp.experts.linear_fc2.weight10 + module.decoder.layers.32.mlp.experts.linear_fc2.weight8 + module.decoder.layers.33.mlp.experts.linear_fc1.weight7 + module.decoder.layers.33.mlp.experts.linear_fc1.weight5 + module.decoder.layers.33.mlp.experts.linear_fc1.weight3 + module.decoder.layers.33.mlp.experts.linear_fc1.weight1 + module.decoder.layers.32.mlp.experts.linear_fc2.weight15 + module.decoder.layers.32.mlp.experts.linear_fc2.weight13 + module.decoder.layers.32.mlp.experts.linear_fc2.weight11 + module.decoder.layers.32.mlp.experts.linear_fc2.weight9 +Params for bucket 29 (40894464 elements, 40894464 padded size): + module.decoder.layers.32.mlp.experts.linear_fc2.weight4 + module.decoder.layers.32.mlp.experts.linear_fc2.weight6 + module.decoder.layers.32.mlp.experts.linear_fc2.weight2 + module.decoder.layers.32.mlp.experts.linear_fc2.weight0 + module.decoder.layers.32.mlp.experts.linear_fc1.weight15 + module.decoder.layers.32.mlp.experts.linear_fc1.weight13 + module.decoder.layers.32.mlp.experts.linear_fc1.weight11 + module.decoder.layers.32.mlp.experts.linear_fc1.weight9 + module.decoder.layers.32.mlp.experts.linear_fc1.weight7 + module.decoder.layers.32.mlp.experts.linear_fc2.weight7 + module.decoder.layers.32.mlp.experts.linear_fc2.weight5 + module.decoder.layers.32.mlp.experts.linear_fc2.weight3 + module.decoder.layers.32.mlp.experts.linear_fc2.weight1 + module.decoder.layers.32.mlp.experts.linear_fc1.weight14 + module.decoder.layers.32.mlp.experts.linear_fc1.weight12 + module.decoder.layers.32.mlp.experts.linear_fc1.weight10 + module.decoder.layers.32.mlp.experts.linear_fc1.weight8 +Params for bucket 30 (40894464 elements, 40894464 padded size): + module.decoder.layers.31.mlp.experts.linear_fc2.weight13 + module.decoder.layers.31.mlp.experts.linear_fc2.weight5 + module.decoder.layers.32.mlp.experts.linear_fc1.weight4 + module.decoder.layers.31.mlp.experts.linear_fc2.weight10 + module.decoder.layers.32.mlp.experts.linear_fc1.weight1 + module.decoder.layers.31.mlp.experts.linear_fc2.weight15 + module.decoder.layers.31.mlp.experts.linear_fc2.weight7 + module.decoder.layers.32.mlp.experts.linear_fc1.weight6 + module.decoder.layers.31.mlp.experts.linear_fc2.weight12 + module.decoder.layers.31.mlp.experts.linear_fc2.weight4 + module.decoder.layers.32.mlp.experts.linear_fc1.weight3 + module.decoder.layers.31.mlp.experts.linear_fc2.weight9 + module.decoder.layers.32.mlp.experts.linear_fc1.weight0 + module.decoder.layers.31.mlp.experts.linear_fc2.weight14 + module.decoder.layers.31.mlp.experts.linear_fc2.weight6 + module.decoder.layers.32.mlp.experts.linear_fc1.weight5 + module.decoder.layers.31.mlp.experts.linear_fc2.weight11 + module.decoder.layers.32.mlp.experts.linear_fc1.weight2 + module.decoder.layers.31.mlp.experts.linear_fc2.weight8 +Params for bucket 31 (40894464 elements, 40894464 padded size): + module.decoder.layers.31.mlp.experts.linear_fc2.weight3 + module.decoder.layers.31.mlp.experts.linear_fc2.weight1 + module.decoder.layers.31.mlp.experts.linear_fc1.weight14 + module.decoder.layers.31.mlp.experts.linear_fc1.weight12 + module.decoder.layers.31.mlp.experts.linear_fc1.weight10 + module.decoder.layers.31.mlp.experts.linear_fc1.weight8 + module.decoder.layers.31.mlp.experts.linear_fc1.weight6 + module.decoder.layers.31.mlp.experts.linear_fc2.weight0 + module.decoder.layers.31.mlp.experts.linear_fc2.weight2 + module.decoder.layers.31.mlp.experts.linear_fc1.weight15 + module.decoder.layers.31.mlp.experts.linear_fc1.weight13 + module.decoder.layers.31.mlp.experts.linear_fc1.weight11 + module.decoder.layers.31.mlp.experts.linear_fc1.weight9 + module.decoder.layers.31.mlp.experts.linear_fc1.weight7 + module.decoder.layers.31.mlp.experts.linear_fc1.weight5 +Params for bucket 32 (40894464 elements, 40894464 padded size): + module.decoder.layers.31.mlp.experts.linear_fc1.weight0 + module.decoder.layers.30.mlp.experts.linear_fc2.weight11 + module.decoder.layers.30.mlp.experts.linear_fc2.weight3 + module.decoder.layers.30.mlp.experts.linear_fc2.weight8 + module.decoder.layers.30.mlp.experts.linear_fc2.weight0 + module.decoder.layers.31.mlp.experts.linear_fc1.weight2 + module.decoder.layers.30.mlp.experts.linear_fc2.weight13 + module.decoder.layers.30.mlp.experts.linear_fc2.weight5 + module.decoder.layers.30.mlp.experts.linear_fc2.weight10 + module.decoder.layers.30.mlp.experts.linear_fc2.weight2 + module.decoder.layers.31.mlp.experts.linear_fc1.weight4 + module.decoder.layers.30.mlp.experts.linear_fc2.weight15 + module.decoder.layers.30.mlp.experts.linear_fc2.weight7 + module.decoder.layers.31.mlp.experts.linear_fc1.weight1 + module.decoder.layers.30.mlp.experts.linear_fc2.weight12 + module.decoder.layers.30.mlp.experts.linear_fc2.weight4 + module.decoder.layers.30.mlp.experts.linear_fc2.weight9 + module.decoder.layers.30.mlp.experts.linear_fc2.weight1 + module.decoder.layers.31.mlp.experts.linear_fc1.weight3 + module.decoder.layers.30.mlp.experts.linear_fc2.weight14 + module.decoder.layers.30.mlp.experts.linear_fc2.weight6 +Params for bucket 33 (40894464 elements, 40894464 padded size): + module.decoder.layers.30.mlp.experts.linear_fc1.weight14 + module.decoder.layers.30.mlp.experts.linear_fc1.weight12 + module.decoder.layers.30.mlp.experts.linear_fc1.weight10 + module.decoder.layers.30.mlp.experts.linear_fc1.weight8 + module.decoder.layers.30.mlp.experts.linear_fc1.weight6 + module.decoder.layers.30.mlp.experts.linear_fc1.weight4 + module.decoder.layers.30.mlp.experts.linear_fc1.weight15 + module.decoder.layers.30.mlp.experts.linear_fc1.weight11 + module.decoder.layers.30.mlp.experts.linear_fc1.weight13 + module.decoder.layers.30.mlp.experts.linear_fc1.weight9 + module.decoder.layers.30.mlp.experts.linear_fc1.weight7 + module.decoder.layers.30.mlp.experts.linear_fc1.weight5 + module.decoder.layers.30.mlp.experts.linear_fc1.weight3 +Params for bucket 34 (40894464 elements, 40894464 padded size): + module.decoder.layers.30.mlp.experts.linear_fc1.weight2 + module.decoder.layers.29.mlp.experts.linear_fc2.weight12 + module.decoder.layers.29.mlp.experts.linear_fc2.weight4 + module.decoder.layers.29.mlp.experts.linear_fc1.weight14 + module.decoder.layers.29.mlp.experts.linear_fc2.weight9 + module.decoder.layers.29.mlp.experts.linear_fc2.weight1 + module.decoder.layers.29.mlp.experts.linear_fc2.weight14 + module.decoder.layers.29.mlp.experts.linear_fc2.weight6 + module.decoder.layers.30.mlp.experts.linear_fc1.weight1 + module.decoder.layers.29.mlp.experts.linear_fc2.weight11 + module.decoder.layers.29.mlp.experts.linear_fc2.weight3 + module.decoder.layers.29.mlp.experts.linear_fc2.weight8 + module.decoder.layers.29.mlp.experts.linear_fc2.weight0 + module.decoder.layers.29.mlp.experts.linear_fc2.weight13 + module.decoder.layers.29.mlp.experts.linear_fc2.weight5 + module.decoder.layers.29.mlp.experts.linear_fc1.weight15 + module.decoder.layers.30.mlp.experts.linear_fc1.weight0 + module.decoder.layers.29.mlp.experts.linear_fc2.weight10 + module.decoder.layers.29.mlp.experts.linear_fc2.weight2 + module.decoder.layers.29.mlp.experts.linear_fc2.weight15 + module.decoder.layers.29.mlp.experts.linear_fc2.weight7 +Params for bucket 35 (40894464 elements, 40894464 padded size): + module.decoder.layers.29.mlp.experts.linear_fc1.weight12 + module.decoder.layers.29.mlp.experts.linear_fc1.weight10 + module.decoder.layers.29.mlp.experts.linear_fc1.weight8 + module.decoder.layers.29.mlp.experts.linear_fc1.weight6 + module.decoder.layers.29.mlp.experts.linear_fc1.weight4 + module.decoder.layers.29.mlp.experts.linear_fc1.weight2 + module.decoder.layers.29.mlp.experts.linear_fc1.weight13 + module.decoder.layers.29.mlp.experts.linear_fc1.weight11 + module.decoder.layers.29.mlp.experts.linear_fc1.weight9 + module.decoder.layers.29.mlp.experts.linear_fc1.weight7 + module.decoder.layers.29.mlp.experts.linear_fc1.weight5 + module.decoder.layers.29.mlp.experts.linear_fc1.weight3 + module.decoder.layers.29.mlp.experts.linear_fc1.weight1 +Params for bucket 36 (40894464 elements, 40894464 padded size): + module.decoder.layers.28.mlp.experts.linear_fc2.weight14 + module.decoder.layers.28.mlp.experts.linear_fc2.weight6 + module.decoder.layers.28.mlp.experts.linear_fc2.weight11 + module.decoder.layers.28.mlp.experts.linear_fc2.weight3 + module.decoder.layers.28.mlp.experts.linear_fc1.weight14 + module.decoder.layers.29.mlp.experts.linear_fc1.weight0 + module.decoder.layers.28.mlp.experts.linear_fc2.weight8 + module.decoder.layers.28.mlp.experts.linear_fc2.weight0 + module.decoder.layers.28.mlp.experts.linear_fc2.weight13 + module.decoder.layers.28.mlp.experts.linear_fc2.weight5 + module.decoder.layers.28.mlp.experts.linear_fc2.weight10 + module.decoder.layers.28.mlp.experts.linear_fc2.weight2 + module.decoder.layers.28.mlp.experts.linear_fc1.weight13 + module.decoder.layers.28.mlp.experts.linear_fc2.weight15 + module.decoder.layers.28.mlp.experts.linear_fc2.weight7 + module.decoder.layers.28.mlp.experts.linear_fc2.weight12 + module.decoder.layers.28.mlp.experts.linear_fc2.weight4 + module.decoder.layers.28.mlp.experts.linear_fc1.weight15 + module.decoder.layers.28.mlp.experts.linear_fc2.weight9 + module.decoder.layers.28.mlp.experts.linear_fc2.weight1 + module.decoder.layers.28.mlp.experts.linear_fc1.weight12 +Params for bucket 37 (40894464 elements, 40894464 padded size): + module.decoder.layers.28.mlp.experts.linear_fc1.weight11 + module.decoder.layers.28.mlp.experts.linear_fc1.weight9 + module.decoder.layers.28.mlp.experts.linear_fc1.weight7 + module.decoder.layers.28.mlp.experts.linear_fc1.weight5 + module.decoder.layers.28.mlp.experts.linear_fc1.weight3 + module.decoder.layers.28.mlp.experts.linear_fc1.weight1 + module.decoder.layers.27.mlp.experts.linear_fc2.weight14 + module.decoder.layers.28.mlp.experts.linear_fc1.weight10 + module.decoder.layers.28.mlp.experts.linear_fc1.weight8 + module.decoder.layers.28.mlp.experts.linear_fc1.weight6 + module.decoder.layers.28.mlp.experts.linear_fc1.weight4 + module.decoder.layers.28.mlp.experts.linear_fc1.weight2 + module.decoder.layers.28.mlp.experts.linear_fc1.weight0 + module.decoder.layers.27.mlp.experts.linear_fc2.weight15 +Params for bucket 38 (40894464 elements, 40894464 padded size): + module.decoder.layers.27.mlp.experts.linear_fc2.weight12 + module.decoder.layers.27.mlp.experts.linear_fc2.weight4 + module.decoder.layers.27.mlp.experts.linear_fc1.weight11 + module.decoder.layers.27.mlp.experts.linear_fc2.weight9 + module.decoder.layers.27.mlp.experts.linear_fc2.weight1 + module.decoder.layers.27.mlp.experts.linear_fc2.weight6 + module.decoder.layers.27.mlp.experts.linear_fc1.weight13 + module.decoder.layers.27.mlp.experts.linear_fc2.weight11 + module.decoder.layers.27.mlp.experts.linear_fc2.weight3 + module.decoder.layers.27.mlp.experts.linear_fc1.weight10 + module.decoder.layers.27.mlp.experts.linear_fc2.weight8 + module.decoder.layers.27.mlp.experts.linear_fc2.weight0 + module.decoder.layers.27.mlp.experts.linear_fc1.weight15 + module.decoder.layers.27.mlp.experts.linear_fc2.weight13 + module.decoder.layers.27.mlp.experts.linear_fc2.weight5 + module.decoder.layers.27.mlp.experts.linear_fc1.weight12 + module.decoder.layers.27.mlp.experts.linear_fc2.weight10 + module.decoder.layers.27.mlp.experts.linear_fc2.weight2 + module.decoder.layers.27.mlp.experts.linear_fc2.weight7 + module.decoder.layers.27.mlp.experts.linear_fc1.weight14 +Params for bucket 39 (40894464 elements, 40894464 padded size): + module.decoder.layers.27.mlp.experts.linear_fc1.weight9 + module.decoder.layers.27.mlp.experts.linear_fc1.weight7 + module.decoder.layers.27.mlp.experts.linear_fc1.weight5 + module.decoder.layers.27.mlp.experts.linear_fc1.weight3 + module.decoder.layers.27.mlp.experts.linear_fc1.weight1 + module.decoder.layers.26.mlp.experts.linear_fc2.weight15 + module.decoder.layers.26.mlp.experts.linear_fc2.weight13 + module.decoder.layers.26.mlp.experts.linear_fc2.weight11 + module.decoder.layers.27.mlp.experts.linear_fc1.weight8 + module.decoder.layers.27.mlp.experts.linear_fc1.weight6 + module.decoder.layers.27.mlp.experts.linear_fc1.weight4 + module.decoder.layers.27.mlp.experts.linear_fc1.weight2 + module.decoder.layers.27.mlp.experts.linear_fc1.weight0 + module.decoder.layers.26.mlp.experts.linear_fc2.weight14 + module.decoder.layers.26.mlp.experts.linear_fc2.weight12 + module.decoder.layers.26.mlp.experts.linear_fc2.weight10 +Params for bucket 40 (40894464 elements, 40894464 padded size): + module.decoder.layers.26.mlp.experts.linear_fc2.weight9 + module.decoder.layers.26.mlp.experts.linear_fc2.weight7 + module.decoder.layers.26.mlp.experts.linear_fc2.weight5 + module.decoder.layers.26.mlp.experts.linear_fc2.weight3 + module.decoder.layers.26.mlp.experts.linear_fc2.weight1 + module.decoder.layers.26.mlp.experts.linear_fc1.weight15 + module.decoder.layers.26.mlp.experts.linear_fc1.weight13 + module.decoder.layers.26.mlp.experts.linear_fc1.weight11 + module.decoder.layers.26.mlp.experts.linear_fc1.weight9 + module.decoder.layers.26.mlp.experts.linear_fc2.weight8 + module.decoder.layers.26.mlp.experts.linear_fc2.weight6 + module.decoder.layers.26.mlp.experts.linear_fc2.weight4 + module.decoder.layers.26.mlp.experts.linear_fc2.weight2 + module.decoder.layers.26.mlp.experts.linear_fc2.weight0 + module.decoder.layers.26.mlp.experts.linear_fc1.weight14 + module.decoder.layers.26.mlp.experts.linear_fc1.weight12 + module.decoder.layers.26.mlp.experts.linear_fc1.weight10 + module.decoder.layers.26.mlp.experts.linear_fc1.weight8 +Params for bucket 41 (40894464 elements, 40894464 padded size): + module.decoder.layers.26.mlp.experts.linear_fc1.weight7 + module.decoder.layers.26.mlp.experts.linear_fc1.weight3 + module.decoder.layers.26.mlp.experts.linear_fc1.weight5 + module.decoder.layers.26.mlp.experts.linear_fc1.weight1 + module.decoder.layers.25.mlp.experts.linear_fc2.weight14 + module.decoder.layers.25.mlp.experts.linear_fc2.weight12 + module.decoder.layers.25.mlp.experts.linear_fc2.weight10 + module.decoder.layers.25.mlp.experts.linear_fc2.weight8 + module.decoder.layers.25.mlp.experts.linear_fc2.weight6 + module.decoder.layers.26.mlp.experts.linear_fc1.weight6 + module.decoder.layers.26.mlp.experts.linear_fc1.weight4 + module.decoder.layers.26.mlp.experts.linear_fc1.weight2 + module.decoder.layers.26.mlp.experts.linear_fc1.weight0 + module.decoder.layers.25.mlp.experts.linear_fc2.weight15 + module.decoder.layers.25.mlp.experts.linear_fc2.weight13 + module.decoder.layers.25.mlp.experts.linear_fc2.weight11 + module.decoder.layers.25.mlp.experts.linear_fc2.weight9 + module.decoder.layers.25.mlp.experts.linear_fc2.weight7 +Params for bucket 42 (40894464 elements, 40894464 padded size): + module.decoder.layers.25.mlp.experts.linear_fc2.weight4 + module.decoder.layers.25.mlp.experts.linear_fc2.weight2 + module.decoder.layers.25.mlp.experts.linear_fc2.weight0 + module.decoder.layers.25.mlp.experts.linear_fc1.weight15 + module.decoder.layers.25.mlp.experts.linear_fc1.weight13 + module.decoder.layers.25.mlp.experts.linear_fc1.weight11 + module.decoder.layers.25.mlp.experts.linear_fc1.weight9 + module.decoder.layers.25.mlp.experts.linear_fc1.weight7 + module.decoder.layers.25.mlp.experts.linear_fc2.weight5 + module.decoder.layers.25.mlp.experts.linear_fc2.weight1 + module.decoder.layers.25.mlp.experts.linear_fc2.weight3 + module.decoder.layers.25.mlp.experts.linear_fc1.weight14 + module.decoder.layers.25.mlp.experts.linear_fc1.weight12 + module.decoder.layers.25.mlp.experts.linear_fc1.weight10 + module.decoder.layers.25.mlp.experts.linear_fc1.weight8 + module.decoder.layers.25.mlp.experts.linear_fc1.weight6 +Params for bucket 43 (40894464 elements, 40894464 padded size): + module.decoder.layers.25.mlp.experts.linear_fc1.weight3 + module.decoder.layers.24.mlp.experts.linear_fc2.weight15 + module.decoder.layers.24.mlp.experts.linear_fc2.weight7 + module.decoder.layers.25.mlp.experts.linear_fc1.weight0 + module.decoder.layers.24.mlp.experts.linear_fc2.weight12 + module.decoder.layers.24.mlp.experts.linear_fc2.weight4 + module.decoder.layers.25.mlp.experts.linear_fc1.weight5 + module.decoder.layers.24.mlp.experts.linear_fc2.weight9 + module.decoder.layers.25.mlp.experts.linear_fc1.weight2 + module.decoder.layers.24.mlp.experts.linear_fc2.weight14 + module.decoder.layers.24.mlp.experts.linear_fc2.weight6 + module.decoder.layers.24.mlp.experts.linear_fc2.weight11 + module.decoder.layers.24.mlp.experts.linear_fc2.weight3 + module.decoder.layers.25.mlp.experts.linear_fc1.weight4 + module.decoder.layers.24.mlp.experts.linear_fc2.weight8 + module.decoder.layers.25.mlp.experts.linear_fc1.weight1 + module.decoder.layers.24.mlp.experts.linear_fc2.weight13 + module.decoder.layers.24.mlp.experts.linear_fc2.weight5 + module.decoder.layers.24.mlp.experts.linear_fc2.weight10 + module.decoder.layers.24.mlp.experts.linear_fc2.weight2 +Params for bucket 44 (40894464 elements, 40894464 padded size): + module.decoder.layers.24.mlp.experts.linear_fc2.weight1 + module.decoder.layers.24.mlp.experts.linear_fc1.weight14 + module.decoder.layers.24.mlp.experts.linear_fc1.weight12 + module.decoder.layers.24.mlp.experts.linear_fc1.weight10 + module.decoder.layers.24.mlp.experts.linear_fc1.weight8 + module.decoder.layers.24.mlp.experts.linear_fc1.weight6 + module.decoder.layers.24.mlp.experts.linear_fc1.weight4 + module.decoder.layers.24.mlp.experts.linear_fc1.weight15 + module.decoder.layers.24.mlp.experts.linear_fc1.weight13 + module.decoder.layers.24.mlp.experts.linear_fc2.weight0 + module.decoder.layers.24.mlp.experts.linear_fc1.weight11 + module.decoder.layers.24.mlp.experts.linear_fc1.weight9 + module.decoder.layers.24.mlp.experts.linear_fc1.weight7 + module.decoder.layers.24.mlp.experts.linear_fc1.weight5 +Params for bucket 45 (40894464 elements, 40894464 padded size): + module.decoder.layers.23.mlp.experts.linear_fc2.weight10 + module.decoder.layers.23.mlp.experts.linear_fc2.weight2 + module.decoder.layers.24.mlp.experts.linear_fc1.weight3 + module.decoder.layers.23.mlp.experts.linear_fc2.weight15 + module.decoder.layers.23.mlp.experts.linear_fc2.weight7 + module.decoder.layers.24.mlp.experts.linear_fc1.weight0 + module.decoder.layers.23.mlp.experts.linear_fc2.weight12 + module.decoder.layers.23.mlp.experts.linear_fc2.weight4 + module.decoder.layers.23.mlp.experts.linear_fc1.weight15 + module.decoder.layers.23.mlp.experts.linear_fc2.weight9 + module.decoder.layers.23.mlp.experts.linear_fc2.weight1 + module.decoder.layers.24.mlp.experts.linear_fc1.weight2 + module.decoder.layers.23.mlp.experts.linear_fc2.weight14 + module.decoder.layers.23.mlp.experts.linear_fc2.weight6 + module.decoder.layers.23.mlp.experts.linear_fc2.weight11 + module.decoder.layers.23.mlp.experts.linear_fc2.weight3 + module.decoder.layers.23.mlp.experts.linear_fc2.weight8 + module.decoder.layers.23.mlp.experts.linear_fc2.weight0 + module.decoder.layers.24.mlp.experts.linear_fc1.weight1 + module.decoder.layers.23.mlp.experts.linear_fc2.weight13 + module.decoder.layers.23.mlp.experts.linear_fc2.weight5 +Params for bucket 46 (40894464 elements, 40894464 padded size): + module.decoder.layers.23.mlp.experts.linear_fc1.weight11 + module.decoder.layers.23.mlp.experts.linear_fc1.weight13 + module.decoder.layers.23.mlp.experts.linear_fc1.weight9 + module.decoder.layers.23.mlp.experts.linear_fc1.weight7 + module.decoder.layers.23.mlp.experts.linear_fc1.weight5 + module.decoder.layers.23.mlp.experts.linear_fc1.weight3 + module.decoder.layers.23.mlp.experts.linear_fc1.weight14 + module.decoder.layers.23.mlp.experts.linear_fc1.weight12 + module.decoder.layers.23.mlp.experts.linear_fc1.weight10 + module.decoder.layers.23.mlp.experts.linear_fc1.weight8 + module.decoder.layers.23.mlp.experts.linear_fc1.weight6 + module.decoder.layers.23.mlp.experts.linear_fc1.weight4 + module.decoder.layers.23.mlp.experts.linear_fc1.weight2 +Params for bucket 47 (40894464 elements, 40894464 padded size): + module.decoder.layers.22.mlp.experts.linear_fc2.weight12 + module.decoder.layers.22.mlp.experts.linear_fc2.weight4 + module.decoder.layers.22.mlp.experts.linear_fc1.weight15 + module.decoder.layers.22.mlp.experts.linear_fc2.weight9 + module.decoder.layers.22.mlp.experts.linear_fc2.weight1 + module.decoder.layers.22.mlp.experts.linear_fc2.weight14 + module.decoder.layers.22.mlp.experts.linear_fc2.weight6 + module.decoder.layers.22.mlp.experts.linear_fc2.weight11 + module.decoder.layers.22.mlp.experts.linear_fc2.weight3 + module.decoder.layers.22.mlp.experts.linear_fc1.weight14 + module.decoder.layers.23.mlp.experts.linear_fc1.weight1 + module.decoder.layers.22.mlp.experts.linear_fc2.weight8 + module.decoder.layers.22.mlp.experts.linear_fc2.weight0 + module.decoder.layers.22.mlp.experts.linear_fc2.weight13 + module.decoder.layers.22.mlp.experts.linear_fc2.weight5 + module.decoder.layers.22.mlp.experts.linear_fc2.weight10 + module.decoder.layers.22.mlp.experts.linear_fc2.weight2 + module.decoder.layers.22.mlp.experts.linear_fc1.weight13 + module.decoder.layers.23.mlp.experts.linear_fc1.weight0 + module.decoder.layers.22.mlp.experts.linear_fc2.weight15 + module.decoder.layers.22.mlp.experts.linear_fc2.weight7 +Params for bucket 48 (40894464 elements, 40894464 padded size): + module.decoder.layers.22.mlp.experts.linear_fc1.weight9 + module.decoder.layers.22.mlp.experts.linear_fc1.weight11 + module.decoder.layers.22.mlp.experts.linear_fc1.weight7 + module.decoder.layers.22.mlp.experts.linear_fc1.weight5 + module.decoder.layers.22.mlp.experts.linear_fc1.weight3 + module.decoder.layers.22.mlp.experts.linear_fc1.weight1 + module.decoder.layers.22.mlp.experts.linear_fc1.weight12 + module.decoder.layers.22.mlp.experts.linear_fc1.weight8 + module.decoder.layers.22.mlp.experts.linear_fc1.weight10 + module.decoder.layers.22.mlp.experts.linear_fc1.weight6 + module.decoder.layers.22.mlp.experts.linear_fc1.weight4 + module.decoder.layers.22.mlp.experts.linear_fc1.weight2 + module.decoder.layers.22.mlp.experts.linear_fc1.weight0 +Params for bucket 49 (40894464 elements, 40894464 padded size): + module.decoder.layers.21.mlp.experts.linear_fc2.weight11 + module.decoder.layers.21.mlp.experts.linear_fc2.weight3 + module.decoder.layers.21.mlp.experts.linear_fc2.weight8 + module.decoder.layers.21.mlp.experts.linear_fc2.weight0 + module.decoder.layers.21.mlp.experts.linear_fc1.weight14 + module.decoder.layers.21.mlp.experts.linear_fc2.weight13 + module.decoder.layers.21.mlp.experts.linear_fc2.weight5 + module.decoder.layers.21.mlp.experts.linear_fc1.weight11 + module.decoder.layers.21.mlp.experts.linear_fc2.weight10 + module.decoder.layers.21.mlp.experts.linear_fc2.weight2 + module.decoder.layers.21.mlp.experts.linear_fc2.weight15 + module.decoder.layers.21.mlp.experts.linear_fc2.weight7 + module.decoder.layers.21.mlp.experts.linear_fc1.weight13 + module.decoder.layers.21.mlp.experts.linear_fc2.weight12 + module.decoder.layers.21.mlp.experts.linear_fc2.weight4 + module.decoder.layers.21.mlp.experts.linear_fc2.weight9 + module.decoder.layers.21.mlp.experts.linear_fc2.weight1 + module.decoder.layers.21.mlp.experts.linear_fc1.weight15 + module.decoder.layers.21.mlp.experts.linear_fc2.weight14 + module.decoder.layers.21.mlp.experts.linear_fc2.weight6 + module.decoder.layers.21.mlp.experts.linear_fc1.weight12 +Params for bucket 50 (40894464 elements, 40894464 padded size): + module.decoder.layers.21.mlp.experts.linear_fc1.weight9 + module.decoder.layers.21.mlp.experts.linear_fc1.weight7 + module.decoder.layers.21.mlp.experts.linear_fc1.weight5 + module.decoder.layers.21.mlp.experts.linear_fc1.weight3 + module.decoder.layers.21.mlp.experts.linear_fc1.weight1 + module.decoder.layers.20.mlp.experts.linear_fc2.weight14 + module.decoder.layers.20.mlp.experts.linear_fc2.weight12 + module.decoder.layers.21.mlp.experts.linear_fc1.weight10 + module.decoder.layers.21.mlp.experts.linear_fc1.weight6 + module.decoder.layers.21.mlp.experts.linear_fc1.weight8 + module.decoder.layers.21.mlp.experts.linear_fc1.weight4 + module.decoder.layers.21.mlp.experts.linear_fc1.weight2 + module.decoder.layers.21.mlp.experts.linear_fc1.weight0 + module.decoder.layers.20.mlp.experts.linear_fc2.weight15 + module.decoder.layers.20.mlp.experts.linear_fc2.weight13 +Params for bucket 51 (40894464 elements, 40894464 padded size): + module.decoder.layers.20.mlp.experts.linear_fc2.weight6 + module.decoder.layers.20.mlp.experts.linear_fc2.weight11 + module.decoder.layers.20.mlp.experts.linear_fc2.weight3 + module.decoder.layers.20.mlp.experts.linear_fc1.weight13 + module.decoder.layers.20.mlp.experts.linear_fc2.weight8 + module.decoder.layers.20.mlp.experts.linear_fc2.weight0 + module.decoder.layers.20.mlp.experts.linear_fc1.weight10 + module.decoder.layers.20.mlp.experts.linear_fc2.weight5 + module.decoder.layers.20.mlp.experts.linear_fc1.weight15 + module.decoder.layers.20.mlp.experts.linear_fc2.weight10 + module.decoder.layers.20.mlp.experts.linear_fc2.weight2 + module.decoder.layers.20.mlp.experts.linear_fc1.weight12 + module.decoder.layers.20.mlp.experts.linear_fc2.weight7 + module.decoder.layers.20.mlp.experts.linear_fc1.weight9 + module.decoder.layers.20.mlp.experts.linear_fc2.weight4 + module.decoder.layers.20.mlp.experts.linear_fc1.weight14 + module.decoder.layers.20.mlp.experts.linear_fc2.weight9 + module.decoder.layers.20.mlp.experts.linear_fc2.weight1 + module.decoder.layers.20.mlp.experts.linear_fc1.weight11 +Params for bucket 52 (40894464 elements, 40894464 padded size): + module.decoder.layers.20.mlp.experts.linear_fc1.weight8 + module.decoder.layers.20.mlp.experts.linear_fc1.weight6 + module.decoder.layers.20.mlp.experts.linear_fc1.weight4 + module.decoder.layers.20.mlp.experts.linear_fc1.weight2 + module.decoder.layers.20.mlp.experts.linear_fc1.weight0 + module.decoder.layers.19.mlp.experts.linear_fc2.weight14 + module.decoder.layers.19.mlp.experts.linear_fc2.weight12 + module.decoder.layers.19.mlp.experts.linear_fc2.weight10 + module.decoder.layers.19.mlp.experts.linear_fc2.weight8 + module.decoder.layers.20.mlp.experts.linear_fc1.weight7 + module.decoder.layers.20.mlp.experts.linear_fc1.weight5 + module.decoder.layers.20.mlp.experts.linear_fc1.weight3 + module.decoder.layers.20.mlp.experts.linear_fc1.weight1 + module.decoder.layers.19.mlp.experts.linear_fc2.weight15 + module.decoder.layers.19.mlp.experts.linear_fc2.weight13 + module.decoder.layers.19.mlp.experts.linear_fc2.weight11 + module.decoder.layers.19.mlp.experts.linear_fc2.weight9 +Params for bucket 53 (40894464 elements, 40894464 padded size): + module.decoder.layers.19.mlp.experts.linear_fc2.weight6 + module.decoder.layers.19.mlp.experts.linear_fc2.weight4 + module.decoder.layers.19.mlp.experts.linear_fc2.weight2 + module.decoder.layers.19.mlp.experts.linear_fc2.weight0 + module.decoder.layers.19.mlp.experts.linear_fc1.weight14 + module.decoder.layers.19.mlp.experts.linear_fc1.weight12 + module.decoder.layers.19.mlp.experts.linear_fc1.weight10 + module.decoder.layers.19.mlp.experts.linear_fc1.weight8 + module.decoder.layers.19.mlp.experts.linear_fc2.weight7 + module.decoder.layers.19.mlp.experts.linear_fc2.weight3 + module.decoder.layers.19.mlp.experts.linear_fc2.weight5 + module.decoder.layers.19.mlp.experts.linear_fc2.weight1 + module.decoder.layers.19.mlp.experts.linear_fc1.weight15 + module.decoder.layers.19.mlp.experts.linear_fc1.weight13 + module.decoder.layers.19.mlp.experts.linear_fc1.weight11 + module.decoder.layers.19.mlp.experts.linear_fc1.weight9 + module.decoder.layers.19.mlp.experts.linear_fc1.weight7 +Params for bucket 54 (40894464 elements, 40894464 padded size): + module.decoder.layers.19.mlp.experts.linear_fc1.weight0 + module.decoder.layers.18.mlp.experts.linear_fc2.weight11 + module.decoder.layers.19.mlp.experts.linear_fc1.weight5 + module.decoder.layers.18.mlp.experts.linear_fc2.weight8 + module.decoder.layers.19.mlp.experts.linear_fc1.weight2 + module.decoder.layers.18.mlp.experts.linear_fc2.weight13 + module.decoder.layers.18.mlp.experts.linear_fc2.weight5 + module.decoder.layers.18.mlp.experts.linear_fc2.weight10 + module.decoder.layers.19.mlp.experts.linear_fc1.weight4 + module.decoder.layers.18.mlp.experts.linear_fc2.weight15 + module.decoder.layers.18.mlp.experts.linear_fc2.weight7 + module.decoder.layers.19.mlp.experts.linear_fc1.weight1 + module.decoder.layers.18.mlp.experts.linear_fc2.weight12 + module.decoder.layers.18.mlp.experts.linear_fc2.weight4 + module.decoder.layers.19.mlp.experts.linear_fc1.weight6 + module.decoder.layers.18.mlp.experts.linear_fc2.weight9 + module.decoder.layers.19.mlp.experts.linear_fc1.weight3 + module.decoder.layers.18.mlp.experts.linear_fc2.weight14 + module.decoder.layers.18.mlp.experts.linear_fc2.weight6 +Params for bucket 55 (40894464 elements, 40894464 padded size): + module.decoder.layers.18.mlp.experts.linear_fc2.weight3 + module.decoder.layers.18.mlp.experts.linear_fc2.weight1 + module.decoder.layers.18.mlp.experts.linear_fc1.weight14 + module.decoder.layers.18.mlp.experts.linear_fc1.weight12 + module.decoder.layers.18.mlp.experts.linear_fc1.weight10 + module.decoder.layers.18.mlp.experts.linear_fc1.weight8 + module.decoder.layers.18.mlp.experts.linear_fc1.weight6 + module.decoder.layers.18.mlp.experts.linear_fc1.weight15 + module.decoder.layers.18.mlp.experts.linear_fc2.weight2 + module.decoder.layers.18.mlp.experts.linear_fc2.weight0 + module.decoder.layers.18.mlp.experts.linear_fc1.weight13 + module.decoder.layers.18.mlp.experts.linear_fc1.weight11 + module.decoder.layers.18.mlp.experts.linear_fc1.weight9 + module.decoder.layers.18.mlp.experts.linear_fc1.weight7 + module.decoder.layers.18.mlp.experts.linear_fc1.weight5 +Params for bucket 56 (40894464 elements, 40894464 padded size): + module.decoder.layers.18.mlp.experts.linear_fc1.weight2 + module.decoder.layers.17.mlp.experts.linear_fc2.weight14 + module.decoder.layers.17.mlp.experts.linear_fc2.weight6 + module.decoder.layers.17.mlp.experts.linear_fc2.weight11 + module.decoder.layers.17.mlp.experts.linear_fc2.weight3 + module.decoder.layers.18.mlp.experts.linear_fc1.weight4 + module.decoder.layers.17.mlp.experts.linear_fc2.weight8 + module.decoder.layers.17.mlp.experts.linear_fc2.weight0 + module.decoder.layers.18.mlp.experts.linear_fc1.weight1 + module.decoder.layers.17.mlp.experts.linear_fc2.weight13 + module.decoder.layers.17.mlp.experts.linear_fc2.weight5 + module.decoder.layers.17.mlp.experts.linear_fc2.weight10 + module.decoder.layers.17.mlp.experts.linear_fc2.weight2 + module.decoder.layers.18.mlp.experts.linear_fc1.weight3 + module.decoder.layers.17.mlp.experts.linear_fc2.weight15 + module.decoder.layers.17.mlp.experts.linear_fc2.weight7 + module.decoder.layers.18.mlp.experts.linear_fc1.weight0 + module.decoder.layers.17.mlp.experts.linear_fc2.weight12 + module.decoder.layers.17.mlp.experts.linear_fc2.weight4 + module.decoder.layers.17.mlp.experts.linear_fc2.weight9 + module.decoder.layers.17.mlp.experts.linear_fc2.weight1 +Params for bucket 57 (40894464 elements, 40894464 padded size): + module.decoder.layers.17.mlp.experts.linear_fc1.weight15 + module.decoder.layers.17.mlp.experts.linear_fc1.weight13 + module.decoder.layers.17.mlp.experts.linear_fc1.weight11 + module.decoder.layers.17.mlp.experts.linear_fc1.weight9 + module.decoder.layers.17.mlp.experts.linear_fc1.weight7 + module.decoder.layers.17.mlp.experts.linear_fc1.weight5 + module.decoder.layers.17.mlp.experts.linear_fc1.weight3 + module.decoder.layers.17.mlp.experts.linear_fc1.weight12 + module.decoder.layers.17.mlp.experts.linear_fc1.weight14 + module.decoder.layers.17.mlp.experts.linear_fc1.weight10 + module.decoder.layers.17.mlp.experts.linear_fc1.weight8 + module.decoder.layers.17.mlp.experts.linear_fc1.weight6 + module.decoder.layers.17.mlp.experts.linear_fc1.weight4 +Params for bucket 58 (40894464 elements, 40894464 padded size): + module.decoder.layers.17.mlp.experts.linear_fc1.weight1 + module.decoder.layers.16.mlp.experts.linear_fc2.weight8 + module.decoder.layers.16.mlp.experts.linear_fc2.weight0 + module.decoder.layers.16.mlp.experts.linear_fc2.weight13 + module.decoder.layers.16.mlp.experts.linear_fc2.weight5 + module.decoder.layers.16.mlp.experts.linear_fc2.weight10 + module.decoder.layers.16.mlp.experts.linear_fc2.weight2 + module.decoder.layers.17.mlp.experts.linear_fc1.weight0 + module.decoder.layers.16.mlp.experts.linear_fc2.weight15 + module.decoder.layers.16.mlp.experts.linear_fc2.weight7 + module.decoder.layers.16.mlp.experts.linear_fc2.weight12 + module.decoder.layers.16.mlp.experts.linear_fc2.weight4 + module.decoder.layers.16.mlp.experts.linear_fc1.weight15 + module.decoder.layers.17.mlp.experts.linear_fc1.weight2 + module.decoder.layers.16.mlp.experts.linear_fc2.weight9 + module.decoder.layers.16.mlp.experts.linear_fc2.weight1 + module.decoder.layers.16.mlp.experts.linear_fc2.weight14 + module.decoder.layers.16.mlp.experts.linear_fc2.weight6 + module.decoder.layers.16.mlp.experts.linear_fc2.weight11 + module.decoder.layers.16.mlp.experts.linear_fc2.weight3 + module.decoder.layers.16.mlp.experts.linear_fc1.weight14 +Params for bucket 59 (40894464 elements, 40894464 padded size): + module.decoder.layers.16.mlp.experts.linear_fc1.weight13 + module.decoder.layers.16.mlp.experts.linear_fc1.weight9 + module.decoder.layers.16.mlp.experts.linear_fc1.weight11 + module.decoder.layers.16.mlp.experts.linear_fc1.weight7 + module.decoder.layers.16.mlp.experts.linear_fc1.weight5 + module.decoder.layers.16.mlp.experts.linear_fc1.weight3 + module.decoder.layers.16.mlp.experts.linear_fc1.weight1 + module.decoder.layers.16.mlp.experts.linear_fc1.weight10 + module.decoder.layers.16.mlp.experts.linear_fc1.weight12 + module.decoder.layers.16.mlp.experts.linear_fc1.weight8 + module.decoder.layers.16.mlp.experts.linear_fc1.weight6 + module.decoder.layers.16.mlp.experts.linear_fc1.weight4 + module.decoder.layers.16.mlp.experts.linear_fc1.weight2 +Params for bucket 60 (40894464 elements, 40894464 padded size): + module.decoder.layers.15.mlp.experts.linear_fc2.weight15 + module.decoder.layers.15.mlp.experts.linear_fc2.weight7 + module.decoder.layers.15.mlp.experts.linear_fc1.weight14 + module.decoder.layers.16.mlp.experts.linear_fc1.weight0 + module.decoder.layers.15.mlp.experts.linear_fc2.weight12 + module.decoder.layers.15.mlp.experts.linear_fc2.weight4 + module.decoder.layers.15.mlp.experts.linear_fc2.weight9 + module.decoder.layers.15.mlp.experts.linear_fc2.weight1 + module.decoder.layers.15.mlp.experts.linear_fc2.weight14 + module.decoder.layers.15.mlp.experts.linear_fc2.weight6 + module.decoder.layers.15.mlp.experts.linear_fc1.weight13 + module.decoder.layers.15.mlp.experts.linear_fc2.weight11 + module.decoder.layers.15.mlp.experts.linear_fc2.weight3 + module.decoder.layers.15.mlp.experts.linear_fc2.weight8 + module.decoder.layers.15.mlp.experts.linear_fc2.weight0 + module.decoder.layers.15.mlp.experts.linear_fc1.weight15 + module.decoder.layers.15.mlp.experts.linear_fc2.weight13 + module.decoder.layers.15.mlp.experts.linear_fc2.weight5 + module.decoder.layers.15.mlp.experts.linear_fc1.weight12 + module.decoder.layers.15.mlp.experts.linear_fc2.weight10 + module.decoder.layers.15.mlp.experts.linear_fc2.weight2 +Params for bucket 61 (40894464 elements, 40894464 padded size): + module.decoder.layers.15.mlp.experts.linear_fc1.weight8 + module.decoder.layers.15.mlp.experts.linear_fc1.weight10 + module.decoder.layers.15.mlp.experts.linear_fc1.weight6 + module.decoder.layers.15.mlp.experts.linear_fc1.weight4 + module.decoder.layers.15.mlp.experts.linear_fc1.weight2 + module.decoder.layers.15.mlp.experts.linear_fc1.weight0 + module.decoder.layers.14.mlp.experts.linear_fc2.weight14 + module.decoder.layers.15.mlp.experts.linear_fc1.weight11 + module.decoder.layers.15.mlp.experts.linear_fc1.weight7 + module.decoder.layers.15.mlp.experts.linear_fc1.weight9 + module.decoder.layers.15.mlp.experts.linear_fc1.weight5 + module.decoder.layers.15.mlp.experts.linear_fc1.weight3 + module.decoder.layers.15.mlp.experts.linear_fc1.weight1 + module.decoder.layers.14.mlp.experts.linear_fc2.weight15 +Params for bucket 62 (40894464 elements, 40894464 padded size): + module.decoder.layers.14.mlp.experts.linear_fc2.weight10 + module.decoder.layers.14.mlp.experts.linear_fc2.weight2 + module.decoder.layers.14.mlp.experts.linear_fc1.weight13 + module.decoder.layers.14.mlp.experts.linear_fc2.weight7 + module.decoder.layers.14.mlp.experts.linear_fc1.weight10 + module.decoder.layers.14.mlp.experts.linear_fc2.weight12 + module.decoder.layers.14.mlp.experts.linear_fc2.weight4 + module.decoder.layers.14.mlp.experts.linear_fc1.weight15 + module.decoder.layers.14.mlp.experts.linear_fc2.weight9 + module.decoder.layers.14.mlp.experts.linear_fc2.weight1 + module.decoder.layers.14.mlp.experts.linear_fc1.weight12 + module.decoder.layers.14.mlp.experts.linear_fc2.weight6 + module.decoder.layers.14.mlp.experts.linear_fc2.weight11 + module.decoder.layers.14.mlp.experts.linear_fc2.weight3 + module.decoder.layers.14.mlp.experts.linear_fc1.weight14 + module.decoder.layers.14.mlp.experts.linear_fc2.weight8 + module.decoder.layers.14.mlp.experts.linear_fc2.weight0 + module.decoder.layers.14.mlp.experts.linear_fc1.weight11 + module.decoder.layers.14.mlp.experts.linear_fc2.weight13 + module.decoder.layers.14.mlp.experts.linear_fc2.weight5 +Params for bucket 63 (40894464 elements, 40894464 padded size): + module.decoder.layers.14.mlp.experts.linear_fc1.weight9 + module.decoder.layers.14.mlp.experts.linear_fc1.weight5 + module.decoder.layers.14.mlp.experts.linear_fc1.weight7 + module.decoder.layers.14.mlp.experts.linear_fc1.weight3 + module.decoder.layers.14.mlp.experts.linear_fc1.weight1 + module.decoder.layers.13.mlp.experts.linear_fc2.weight14 + module.decoder.layers.13.mlp.experts.linear_fc2.weight12 + module.decoder.layers.13.mlp.experts.linear_fc2.weight10 + module.decoder.layers.14.mlp.experts.linear_fc1.weight6 + module.decoder.layers.14.mlp.experts.linear_fc1.weight8 + module.decoder.layers.14.mlp.experts.linear_fc1.weight4 + module.decoder.layers.14.mlp.experts.linear_fc1.weight2 + module.decoder.layers.14.mlp.experts.linear_fc1.weight0 + module.decoder.layers.13.mlp.experts.linear_fc2.weight15 + module.decoder.layers.13.mlp.experts.linear_fc2.weight13 + module.decoder.layers.13.mlp.experts.linear_fc2.weight11 +Params for bucket 64 (40894464 elements, 40894464 padded size): + module.decoder.layers.13.mlp.experts.linear_fc2.weight8 + module.decoder.layers.13.mlp.experts.linear_fc2.weight6 + module.decoder.layers.13.mlp.experts.linear_fc2.weight4 + module.decoder.layers.13.mlp.experts.linear_fc2.weight2 + module.decoder.layers.13.mlp.experts.linear_fc2.weight0 + module.decoder.layers.13.mlp.experts.linear_fc1.weight15 + module.decoder.layers.13.mlp.experts.linear_fc1.weight13 + module.decoder.layers.13.mlp.experts.linear_fc1.weight11 + module.decoder.layers.13.mlp.experts.linear_fc1.weight9 + module.decoder.layers.13.mlp.experts.linear_fc2.weight9 + module.decoder.layers.13.mlp.experts.linear_fc2.weight7 + module.decoder.layers.13.mlp.experts.linear_fc2.weight5 + module.decoder.layers.13.mlp.experts.linear_fc2.weight3 + module.decoder.layers.13.mlp.experts.linear_fc2.weight1 + module.decoder.layers.13.mlp.experts.linear_fc1.weight14 + module.decoder.layers.13.mlp.experts.linear_fc1.weight12 + module.decoder.layers.13.mlp.experts.linear_fc1.weight10 + module.decoder.layers.13.mlp.experts.linear_fc1.weight8 +Params for bucket 65 (40894464 elements, 40894464 padded size): + module.decoder.layers.13.mlp.experts.linear_fc1.weight7 + module.decoder.layers.13.mlp.experts.linear_fc1.weight3 + module.decoder.layers.13.mlp.experts.linear_fc1.weight5 + module.decoder.layers.13.mlp.experts.linear_fc1.weight1 + module.decoder.layers.12.mlp.experts.linear_fc2.weight14 + module.decoder.layers.12.mlp.experts.linear_fc2.weight12 + module.decoder.layers.12.mlp.experts.linear_fc2.weight10 + module.decoder.layers.12.mlp.experts.linear_fc2.weight8 + module.decoder.layers.12.mlp.experts.linear_fc2.weight6 + module.decoder.layers.13.mlp.experts.linear_fc1.weight4 + module.decoder.layers.13.mlp.experts.linear_fc1.weight6 + module.decoder.layers.13.mlp.experts.linear_fc1.weight2 + module.decoder.layers.13.mlp.experts.linear_fc1.weight0 + module.decoder.layers.12.mlp.experts.linear_fc2.weight15 + module.decoder.layers.12.mlp.experts.linear_fc2.weight13 + module.decoder.layers.12.mlp.experts.linear_fc2.weight11 + module.decoder.layers.12.mlp.experts.linear_fc2.weight9 + module.decoder.layers.12.mlp.experts.linear_fc2.weight7 +Params for bucket 66 (40894464 elements, 40894464 padded size): + module.decoder.layers.12.mlp.experts.linear_fc2.weight4 + module.decoder.layers.12.mlp.experts.linear_fc2.weight2 + module.decoder.layers.12.mlp.experts.linear_fc2.weight0 + module.decoder.layers.12.mlp.experts.linear_fc1.weight15 + module.decoder.layers.12.mlp.experts.linear_fc1.weight13 + module.decoder.layers.12.mlp.experts.linear_fc1.weight11 + module.decoder.layers.12.mlp.experts.linear_fc1.weight9 + module.decoder.layers.12.mlp.experts.linear_fc1.weight7 + module.decoder.layers.12.mlp.experts.linear_fc2.weight5 + module.decoder.layers.12.mlp.experts.linear_fc2.weight3 + module.decoder.layers.12.mlp.experts.linear_fc2.weight1 + module.decoder.layers.12.mlp.experts.linear_fc1.weight14 + module.decoder.layers.12.mlp.experts.linear_fc1.weight12 + module.decoder.layers.12.mlp.experts.linear_fc1.weight10 + module.decoder.layers.12.mlp.experts.linear_fc1.weight8 + module.decoder.layers.12.mlp.experts.linear_fc1.weight6 +Params for bucket 67 (40894464 elements, 40894464 padded size): + module.decoder.layers.11.mlp.experts.linear_fc2.weight10 + module.decoder.layers.11.mlp.experts.linear_fc2.weight2 + module.decoder.layers.12.mlp.experts.linear_fc1.weight4 + module.decoder.layers.11.mlp.experts.linear_fc2.weight15 + module.decoder.layers.11.mlp.experts.linear_fc2.weight7 + module.decoder.layers.12.mlp.experts.linear_fc1.weight1 + module.decoder.layers.11.mlp.experts.linear_fc2.weight12 + module.decoder.layers.11.mlp.experts.linear_fc2.weight4 + module.decoder.layers.11.mlp.experts.linear_fc2.weight9 + module.decoder.layers.12.mlp.experts.linear_fc1.weight3 + module.decoder.layers.11.mlp.experts.linear_fc2.weight14 + module.decoder.layers.11.mlp.experts.linear_fc2.weight6 + module.decoder.layers.12.mlp.experts.linear_fc1.weight0 + module.decoder.layers.11.mlp.experts.linear_fc2.weight11 + module.decoder.layers.11.mlp.experts.linear_fc2.weight3 + module.decoder.layers.12.mlp.experts.linear_fc1.weight5 + module.decoder.layers.11.mlp.experts.linear_fc2.weight8 + module.decoder.layers.12.mlp.experts.linear_fc1.weight2 + module.decoder.layers.11.mlp.experts.linear_fc2.weight13 + module.decoder.layers.11.mlp.experts.linear_fc2.weight5 +Params for bucket 68 (40894464 elements, 40894464 padded size): + module.decoder.layers.11.mlp.experts.linear_fc1.weight13 + module.decoder.layers.11.mlp.experts.linear_fc2.weight0 + module.decoder.layers.11.mlp.experts.linear_fc1.weight15 + module.decoder.layers.11.mlp.experts.linear_fc1.weight11 + module.decoder.layers.11.mlp.experts.linear_fc1.weight9 + module.decoder.layers.11.mlp.experts.linear_fc1.weight7 + module.decoder.layers.11.mlp.experts.linear_fc1.weight5 + module.decoder.layers.11.mlp.experts.linear_fc2.weight1 + module.decoder.layers.11.mlp.experts.linear_fc1.weight14 + module.decoder.layers.11.mlp.experts.linear_fc1.weight12 + module.decoder.layers.11.mlp.experts.linear_fc1.weight10 + module.decoder.layers.11.mlp.experts.linear_fc1.weight8 + module.decoder.layers.11.mlp.experts.linear_fc1.weight6 + module.decoder.layers.11.mlp.experts.linear_fc1.weight4 +Params for bucket 69 (40894464 elements, 40894464 padded size): + module.decoder.layers.10.mlp.experts.linear_fc2.weight13 + module.decoder.layers.10.mlp.experts.linear_fc2.weight5 + module.decoder.layers.10.mlp.experts.linear_fc1.weight15 + module.decoder.layers.11.mlp.experts.linear_fc1.weight2 + module.decoder.layers.10.mlp.experts.linear_fc2.weight10 + module.decoder.layers.10.mlp.experts.linear_fc2.weight2 + module.decoder.layers.10.mlp.experts.linear_fc2.weight15 + module.decoder.layers.10.mlp.experts.linear_fc2.weight7 + module.decoder.layers.10.mlp.experts.linear_fc2.weight12 + module.decoder.layers.10.mlp.experts.linear_fc2.weight4 + module.decoder.layers.11.mlp.experts.linear_fc1.weight1 + module.decoder.layers.10.mlp.experts.linear_fc2.weight9 + module.decoder.layers.10.mlp.experts.linear_fc2.weight1 + module.decoder.layers.10.mlp.experts.linear_fc2.weight14 + module.decoder.layers.10.mlp.experts.linear_fc2.weight6 + module.decoder.layers.11.mlp.experts.linear_fc1.weight3 + module.decoder.layers.10.mlp.experts.linear_fc2.weight11 + module.decoder.layers.10.mlp.experts.linear_fc2.weight3 + module.decoder.layers.11.mlp.experts.linear_fc1.weight0 + module.decoder.layers.10.mlp.experts.linear_fc2.weight8 + module.decoder.layers.10.mlp.experts.linear_fc2.weight0 +Params for bucket 70 (40894464 elements, 40894464 padded size): + module.decoder.layers.10.mlp.experts.linear_fc1.weight11 + module.decoder.layers.10.mlp.experts.linear_fc1.weight13 + module.decoder.layers.10.mlp.experts.linear_fc1.weight9 + module.decoder.layers.10.mlp.experts.linear_fc1.weight7 + module.decoder.layers.10.mlp.experts.linear_fc1.weight5 + module.decoder.layers.10.mlp.experts.linear_fc1.weight3 + module.decoder.layers.10.mlp.experts.linear_fc1.weight14 + module.decoder.layers.10.mlp.experts.linear_fc1.weight12 + module.decoder.layers.10.mlp.experts.linear_fc1.weight10 + module.decoder.layers.10.mlp.experts.linear_fc1.weight8 + module.decoder.layers.10.mlp.experts.linear_fc1.weight6 + module.decoder.layers.10.mlp.experts.linear_fc1.weight4 + module.decoder.layers.10.mlp.experts.linear_fc1.weight2 +Params for bucket 71 (40894464 elements, 40894464 padded size): + module.decoder.layers.9.mlp.experts.linear_fc2.weight9 + module.decoder.layers.9.mlp.experts.linear_fc2.weight1 + module.decoder.layers.9.mlp.experts.linear_fc2.weight14 + module.decoder.layers.9.mlp.experts.linear_fc2.weight6 + module.decoder.layers.10.mlp.experts.linear_fc1.weight1 + module.decoder.layers.9.mlp.experts.linear_fc2.weight11 + module.decoder.layers.9.mlp.experts.linear_fc2.weight3 + module.decoder.layers.9.mlp.experts.linear_fc1.weight14 + module.decoder.layers.9.mlp.experts.linear_fc2.weight8 + module.decoder.layers.9.mlp.experts.linear_fc2.weight0 + module.decoder.layers.9.mlp.experts.linear_fc2.weight13 + module.decoder.layers.9.mlp.experts.linear_fc2.weight5 + module.decoder.layers.10.mlp.experts.linear_fc1.weight0 + module.decoder.layers.9.mlp.experts.linear_fc2.weight10 + module.decoder.layers.9.mlp.experts.linear_fc2.weight2 + module.decoder.layers.9.mlp.experts.linear_fc1.weight13 + module.decoder.layers.9.mlp.experts.linear_fc2.weight15 + module.decoder.layers.9.mlp.experts.linear_fc2.weight7 + module.decoder.layers.9.mlp.experts.linear_fc2.weight12 + module.decoder.layers.9.mlp.experts.linear_fc2.weight4 + module.decoder.layers.9.mlp.experts.linear_fc1.weight15 +Params for bucket 72 (40894464 elements, 40894464 padded size): + module.decoder.layers.9.mlp.experts.linear_fc1.weight12 + module.decoder.layers.9.mlp.experts.linear_fc1.weight10 + module.decoder.layers.9.mlp.experts.linear_fc1.weight8 + module.decoder.layers.9.mlp.experts.linear_fc1.weight6 + module.decoder.layers.9.mlp.experts.linear_fc1.weight4 + module.decoder.layers.9.mlp.experts.linear_fc1.weight2 + module.decoder.layers.9.mlp.experts.linear_fc1.weight0 + module.decoder.layers.9.mlp.experts.linear_fc1.weight11 + module.decoder.layers.9.mlp.experts.linear_fc1.weight9 + module.decoder.layers.9.mlp.experts.linear_fc1.weight7 + module.decoder.layers.9.mlp.experts.linear_fc1.weight5 + module.decoder.layers.9.mlp.experts.linear_fc1.weight3 + module.decoder.layers.9.mlp.experts.linear_fc1.weight1 +Params for bucket 73 (40894464 elements, 40894464 padded size): + module.decoder.layers.8.mlp.experts.linear_fc2.weight10 + module.decoder.layers.8.mlp.experts.linear_fc2.weight2 + module.decoder.layers.8.mlp.experts.linear_fc1.weight13 + module.decoder.layers.8.mlp.experts.linear_fc2.weight15 + module.decoder.layers.8.mlp.experts.linear_fc2.weight7 + module.decoder.layers.8.mlp.experts.linear_fc2.weight12 + module.decoder.layers.8.mlp.experts.linear_fc2.weight4 + module.decoder.layers.8.mlp.experts.linear_fc1.weight15 + module.decoder.layers.8.mlp.experts.linear_fc2.weight9 + module.decoder.layers.8.mlp.experts.linear_fc2.weight1 + module.decoder.layers.8.mlp.experts.linear_fc1.weight12 + module.decoder.layers.8.mlp.experts.linear_fc2.weight14 + module.decoder.layers.8.mlp.experts.linear_fc2.weight6 + module.decoder.layers.8.mlp.experts.linear_fc2.weight11 + module.decoder.layers.8.mlp.experts.linear_fc2.weight3 + module.decoder.layers.8.mlp.experts.linear_fc1.weight14 + module.decoder.layers.8.mlp.experts.linear_fc2.weight8 + module.decoder.layers.8.mlp.experts.linear_fc2.weight0 + module.decoder.layers.8.mlp.experts.linear_fc1.weight11 + module.decoder.layers.8.mlp.experts.linear_fc2.weight13 + module.decoder.layers.8.mlp.experts.linear_fc2.weight5 +Params for bucket 74 (40894464 elements, 40894464 padded size): + module.decoder.layers.8.mlp.experts.linear_fc1.weight9 + module.decoder.layers.8.mlp.experts.linear_fc1.weight7 + module.decoder.layers.8.mlp.experts.linear_fc1.weight5 + module.decoder.layers.8.mlp.experts.linear_fc1.weight3 + module.decoder.layers.8.mlp.experts.linear_fc1.weight1 + module.decoder.layers.7.mlp.experts.linear_fc2.weight15 + module.decoder.layers.7.mlp.experts.linear_fc2.weight13 + module.decoder.layers.8.mlp.experts.linear_fc1.weight10 + module.decoder.layers.8.mlp.experts.linear_fc1.weight6 + module.decoder.layers.8.mlp.experts.linear_fc1.weight8 + module.decoder.layers.8.mlp.experts.linear_fc1.weight4 + module.decoder.layers.8.mlp.experts.linear_fc1.weight2 + module.decoder.layers.8.mlp.experts.linear_fc1.weight0 + module.decoder.layers.7.mlp.experts.linear_fc2.weight14 + module.decoder.layers.7.mlp.experts.linear_fc2.weight12 +Params for bucket 75 (40894464 elements, 40894464 padded size): + module.decoder.layers.7.mlp.experts.linear_fc2.weight3 + module.decoder.layers.7.mlp.experts.linear_fc1.weight14 + module.decoder.layers.7.mlp.experts.linear_fc2.weight0 + module.decoder.layers.7.mlp.experts.linear_fc1.weight11 + module.decoder.layers.7.mlp.experts.linear_fc2.weight9 + module.decoder.layers.7.mlp.experts.linear_fc2.weight5 + module.decoder.layers.7.mlp.experts.linear_fc2.weight2 + module.decoder.layers.7.mlp.experts.linear_fc1.weight13 + module.decoder.layers.7.mlp.experts.linear_fc2.weight11 + module.decoder.layers.7.mlp.experts.linear_fc2.weight7 + module.decoder.layers.7.mlp.experts.linear_fc1.weight10 + module.decoder.layers.7.mlp.experts.linear_fc2.weight8 + module.decoder.layers.7.mlp.experts.linear_fc2.weight4 + module.decoder.layers.7.mlp.experts.linear_fc1.weight15 + module.decoder.layers.7.mlp.experts.linear_fc2.weight1 + module.decoder.layers.7.mlp.experts.linear_fc1.weight12 + module.decoder.layers.7.mlp.experts.linear_fc2.weight10 + module.decoder.layers.7.mlp.experts.linear_fc2.weight6 + module.decoder.layers.7.mlp.experts.linear_fc1.weight9 +Params for bucket 76 (40894464 elements, 40894464 padded size): + module.decoder.layers.7.mlp.experts.linear_fc1.weight8 + module.decoder.layers.7.mlp.experts.linear_fc1.weight4 + module.decoder.layers.7.mlp.experts.linear_fc1.weight6 + module.decoder.layers.7.mlp.experts.linear_fc1.weight2 + module.decoder.layers.7.mlp.experts.linear_fc1.weight0 + module.decoder.layers.6.mlp.experts.linear_fc2.weight14 + module.decoder.layers.6.mlp.experts.linear_fc2.weight12 + module.decoder.layers.6.mlp.experts.linear_fc2.weight10 + module.decoder.layers.6.mlp.experts.linear_fc2.weight8 + module.decoder.layers.7.mlp.experts.linear_fc1.weight5 + module.decoder.layers.7.mlp.experts.linear_fc1.weight7 + module.decoder.layers.7.mlp.experts.linear_fc1.weight3 + module.decoder.layers.7.mlp.experts.linear_fc1.weight1 + module.decoder.layers.6.mlp.experts.linear_fc2.weight15 + module.decoder.layers.6.mlp.experts.linear_fc2.weight13 + module.decoder.layers.6.mlp.experts.linear_fc2.weight11 + module.decoder.layers.6.mlp.experts.linear_fc2.weight9 +Params for bucket 77 (40894464 elements, 40894464 padded size): + module.decoder.layers.6.mlp.experts.linear_fc2.weight6 + module.decoder.layers.6.mlp.experts.linear_fc2.weight4 + module.decoder.layers.6.mlp.experts.linear_fc2.weight2 + module.decoder.layers.6.mlp.experts.linear_fc2.weight0 + module.decoder.layers.6.mlp.experts.linear_fc1.weight15 + module.decoder.layers.6.mlp.experts.linear_fc1.weight13 + module.decoder.layers.6.mlp.experts.linear_fc1.weight11 + module.decoder.layers.6.mlp.experts.linear_fc1.weight9 + module.decoder.layers.6.mlp.experts.linear_fc1.weight7 + module.decoder.layers.6.mlp.experts.linear_fc2.weight7 + module.decoder.layers.6.mlp.experts.linear_fc2.weight3 + module.decoder.layers.6.mlp.experts.linear_fc2.weight5 + module.decoder.layers.6.mlp.experts.linear_fc2.weight1 + module.decoder.layers.6.mlp.experts.linear_fc1.weight14 + module.decoder.layers.6.mlp.experts.linear_fc1.weight12 + module.decoder.layers.6.mlp.experts.linear_fc1.weight10 + module.decoder.layers.6.mlp.experts.linear_fc1.weight8 +Params for bucket 78 (40894464 elements, 40894464 padded size): + module.decoder.layers.6.mlp.experts.linear_fc1.weight1 + module.decoder.layers.5.mlp.experts.linear_fc2.weight12 + module.decoder.layers.5.mlp.experts.linear_fc2.weight4 + module.decoder.layers.5.mlp.experts.linear_fc2.weight9 + module.decoder.layers.6.mlp.experts.linear_fc1.weight3 + module.decoder.layers.5.mlp.experts.linear_fc2.weight14 + module.decoder.layers.5.mlp.experts.linear_fc2.weight6 + module.decoder.layers.6.mlp.experts.linear_fc1.weight0 + module.decoder.layers.5.mlp.experts.linear_fc2.weight11 + module.decoder.layers.6.mlp.experts.linear_fc1.weight5 + module.decoder.layers.5.mlp.experts.linear_fc2.weight8 + module.decoder.layers.6.mlp.experts.linear_fc1.weight6 + module.decoder.layers.6.mlp.experts.linear_fc1.weight2 + module.decoder.layers.5.mlp.experts.linear_fc2.weight13 + module.decoder.layers.5.mlp.experts.linear_fc2.weight5 + module.decoder.layers.5.mlp.experts.linear_fc2.weight10 + module.decoder.layers.6.mlp.experts.linear_fc1.weight4 + module.decoder.layers.5.mlp.experts.linear_fc2.weight15 + module.decoder.layers.5.mlp.experts.linear_fc2.weight7 +Params for bucket 79 (40894464 elements, 40894464 padded size): + module.decoder.layers.5.mlp.experts.linear_fc1.weight15 + module.decoder.layers.5.mlp.experts.linear_fc2.weight0 + module.decoder.layers.5.mlp.experts.linear_fc2.weight2 + module.decoder.layers.5.mlp.experts.linear_fc1.weight13 + module.decoder.layers.5.mlp.experts.linear_fc1.weight11 + module.decoder.layers.5.mlp.experts.linear_fc1.weight9 + module.decoder.layers.5.mlp.experts.linear_fc1.weight7 + module.decoder.layers.5.mlp.experts.linear_fc1.weight5 + module.decoder.layers.5.mlp.experts.linear_fc2.weight3 + module.decoder.layers.5.mlp.experts.linear_fc2.weight1 + module.decoder.layers.5.mlp.experts.linear_fc1.weight14 + module.decoder.layers.5.mlp.experts.linear_fc1.weight12 + module.decoder.layers.5.mlp.experts.linear_fc1.weight10 + module.decoder.layers.5.mlp.experts.linear_fc1.weight8 + module.decoder.layers.5.mlp.experts.linear_fc1.weight6 +Params for bucket 80 (40894464 elements, 40894464 padded size): + module.decoder.layers.4.mlp.experts.linear_fc2.weight12 + module.decoder.layers.4.mlp.experts.linear_fc2.weight0 + module.decoder.layers.5.mlp.experts.linear_fc1.weight4 + module.decoder.layers.4.mlp.experts.linear_fc2.weight9 + module.decoder.layers.5.mlp.experts.linear_fc1.weight1 + module.decoder.layers.4.mlp.experts.linear_fc2.weight14 + module.decoder.layers.4.mlp.experts.linear_fc2.weight6 + module.decoder.layers.4.mlp.experts.linear_fc2.weight2 + module.decoder.layers.4.mlp.experts.linear_fc2.weight11 + module.decoder.layers.5.mlp.experts.linear_fc1.weight3 + module.decoder.layers.4.mlp.experts.linear_fc2.weight8 + module.decoder.layers.4.mlp.experts.linear_fc2.weight4 + module.decoder.layers.5.mlp.experts.linear_fc1.weight0 + module.decoder.layers.4.mlp.experts.linear_fc2.weight13 + module.decoder.layers.4.mlp.experts.linear_fc2.weight5 + module.decoder.layers.4.mlp.experts.linear_fc2.weight1 + module.decoder.layers.4.mlp.experts.linear_fc2.weight10 + module.decoder.layers.5.mlp.experts.linear_fc1.weight2 + module.decoder.layers.4.mlp.experts.linear_fc2.weight15 + module.decoder.layers.4.mlp.experts.linear_fc2.weight7 + module.decoder.layers.4.mlp.experts.linear_fc2.weight3 +Params for bucket 81 (40894464 elements, 40894464 padded size): + module.decoder.layers.4.mlp.experts.linear_fc1.weight15 + module.decoder.layers.4.mlp.experts.linear_fc1.weight11 + module.decoder.layers.4.mlp.experts.linear_fc1.weight13 + module.decoder.layers.4.mlp.experts.linear_fc1.weight9 + module.decoder.layers.4.mlp.experts.linear_fc1.weight7 + module.decoder.layers.4.mlp.experts.linear_fc1.weight5 + module.decoder.layers.4.mlp.experts.linear_fc1.weight3 + module.decoder.layers.4.mlp.experts.linear_fc1.weight12 + module.decoder.layers.4.mlp.experts.linear_fc1.weight14 + module.decoder.layers.4.mlp.experts.linear_fc1.weight10 + module.decoder.layers.4.mlp.experts.linear_fc1.weight8 + module.decoder.layers.4.mlp.experts.linear_fc1.weight6 + module.decoder.layers.4.mlp.experts.linear_fc1.weight4 +Params for bucket 82 (40894464 elements, 40894464 padded size): + module.decoder.layers.3.mlp.experts.linear_fc2.weight14 + module.decoder.layers.3.mlp.experts.linear_fc2.weight6 + module.decoder.layers.4.mlp.experts.linear_fc1.weight0 + module.decoder.layers.3.mlp.experts.linear_fc2.weight11 + module.decoder.layers.3.mlp.experts.linear_fc2.weight3 + module.decoder.layers.3.mlp.experts.linear_fc2.weight8 + module.decoder.layers.3.mlp.experts.linear_fc2.weight0 + module.decoder.layers.4.mlp.experts.linear_fc1.weight2 + module.decoder.layers.3.mlp.experts.linear_fc2.weight13 + module.decoder.layers.3.mlp.experts.linear_fc2.weight5 + module.decoder.layers.3.mlp.experts.linear_fc1.weight15 + module.decoder.layers.3.mlp.experts.linear_fc2.weight10 + module.decoder.layers.3.mlp.experts.linear_fc2.weight2 + module.decoder.layers.3.mlp.experts.linear_fc2.weight15 + module.decoder.layers.3.mlp.experts.linear_fc2.weight7 + module.decoder.layers.4.mlp.experts.linear_fc1.weight1 + module.decoder.layers.3.mlp.experts.linear_fc2.weight12 + module.decoder.layers.3.mlp.experts.linear_fc2.weight4 + module.decoder.layers.3.mlp.experts.linear_fc1.weight14 + module.decoder.layers.3.mlp.experts.linear_fc2.weight9 + module.decoder.layers.3.mlp.experts.linear_fc2.weight1 +Params for bucket 83 (40894464 elements, 40894464 padded size): + module.decoder.layers.3.mlp.experts.linear_fc1.weight10 + module.decoder.layers.3.mlp.experts.linear_fc1.weight12 + module.decoder.layers.3.mlp.experts.linear_fc1.weight8 + module.decoder.layers.3.mlp.experts.linear_fc1.weight6 + module.decoder.layers.3.mlp.experts.linear_fc1.weight4 + module.decoder.layers.3.mlp.experts.linear_fc1.weight2 + module.decoder.layers.3.mlp.experts.linear_fc1.weight13 + module.decoder.layers.3.mlp.experts.linear_fc1.weight11 + module.decoder.layers.3.mlp.experts.linear_fc1.weight9 + module.decoder.layers.3.mlp.experts.linear_fc1.weight7 + module.decoder.layers.3.mlp.experts.linear_fc1.weight5 + module.decoder.layers.3.mlp.experts.linear_fc1.weight3 + module.decoder.layers.3.mlp.experts.linear_fc1.weight1 +Params for bucket 84 (40894464 elements, 40894464 padded size): + module.decoder.layers.2.mlp.experts.linear_fc2.weight8 + module.decoder.layers.2.mlp.experts.linear_fc2.weight0 + module.decoder.layers.2.mlp.experts.linear_fc2.weight13 + module.decoder.layers.2.mlp.experts.linear_fc2.weight5 + module.decoder.layers.2.mlp.experts.linear_fc2.weight10 + module.decoder.layers.2.mlp.experts.linear_fc2.weight2 + module.decoder.layers.2.mlp.experts.linear_fc1.weight13 + module.decoder.layers.2.mlp.experts.linear_fc2.weight15 + module.decoder.layers.2.mlp.experts.linear_fc2.weight7 + module.decoder.layers.3.mlp.experts.linear_fc1.weight0 + module.decoder.layers.2.mlp.experts.linear_fc2.weight12 + module.decoder.layers.2.mlp.experts.linear_fc2.weight4 + module.decoder.layers.2.mlp.experts.linear_fc1.weight15 + module.decoder.layers.2.mlp.experts.linear_fc2.weight9 + module.decoder.layers.2.mlp.experts.linear_fc2.weight1 + module.decoder.layers.2.mlp.experts.linear_fc1.weight12 + module.decoder.layers.2.mlp.experts.linear_fc2.weight14 + module.decoder.layers.2.mlp.experts.linear_fc2.weight6 + module.decoder.layers.2.mlp.experts.linear_fc2.weight11 + module.decoder.layers.2.mlp.experts.linear_fc2.weight3 + module.decoder.layers.2.mlp.experts.linear_fc1.weight14 +Params for bucket 85 (40894464 elements, 40894464 padded size): + module.decoder.layers.2.mlp.experts.linear_fc1.weight11 + module.decoder.layers.2.mlp.experts.linear_fc1.weight9 + module.decoder.layers.2.mlp.experts.linear_fc1.weight7 + module.decoder.layers.2.mlp.experts.linear_fc1.weight5 + module.decoder.layers.2.mlp.experts.linear_fc1.weight3 + module.decoder.layers.2.mlp.experts.linear_fc1.weight1 + module.decoder.layers.1.mlp.experts.linear_fc2.weight15 + module.decoder.layers.2.mlp.experts.linear_fc1.weight8 + module.decoder.layers.2.mlp.experts.linear_fc1.weight10 + module.decoder.layers.2.mlp.experts.linear_fc1.weight6 + module.decoder.layers.2.mlp.experts.linear_fc1.weight4 + module.decoder.layers.2.mlp.experts.linear_fc1.weight2 + module.decoder.layers.2.mlp.experts.linear_fc1.weight0 + module.decoder.layers.1.mlp.experts.linear_fc2.weight14 +Params for bucket 86 (40894464 elements, 40894464 padded size): + module.decoder.layers.1.mlp.experts.linear_fc2.weight7 + module.decoder.layers.1.mlp.experts.linear_fc1.weight14 + module.decoder.layers.1.mlp.experts.linear_fc2.weight12 + module.decoder.layers.1.mlp.experts.linear_fc2.weight4 + module.decoder.layers.1.mlp.experts.linear_fc1.weight11 + module.decoder.layers.1.mlp.experts.linear_fc2.weight9 + module.decoder.layers.1.mlp.experts.linear_fc2.weight1 + module.decoder.layers.1.mlp.experts.linear_fc2.weight6 + module.decoder.layers.1.mlp.experts.linear_fc1.weight13 + module.decoder.layers.1.mlp.experts.linear_fc2.weight11 + module.decoder.layers.1.mlp.experts.linear_fc2.weight3 + module.decoder.layers.1.mlp.experts.linear_fc1.weight10 + module.decoder.layers.1.mlp.experts.linear_fc2.weight8 + module.decoder.layers.1.mlp.experts.linear_fc2.weight0 + module.decoder.layers.1.mlp.experts.linear_fc1.weight15 + module.decoder.layers.1.mlp.experts.linear_fc2.weight13 + module.decoder.layers.1.mlp.experts.linear_fc2.weight5 + module.decoder.layers.1.mlp.experts.linear_fc1.weight12 + module.decoder.layers.1.mlp.experts.linear_fc2.weight10 + module.decoder.layers.1.mlp.experts.linear_fc2.weight2 +Params for bucket 87 (40894464 elements, 40894464 padded size): + module.decoder.layers.1.mlp.experts.linear_fc1.weight6 + module.decoder.layers.1.mlp.experts.linear_fc1.weight8 + module.decoder.layers.1.mlp.experts.linear_fc1.weight4 + module.decoder.layers.1.mlp.experts.linear_fc1.weight2 + module.decoder.layers.1.mlp.experts.linear_fc1.weight0 + module.decoder.layers.0.mlp.experts.linear_fc2.weight15 + module.decoder.layers.0.mlp.experts.linear_fc2.weight13 + module.decoder.layers.0.mlp.experts.linear_fc2.weight11 + module.decoder.layers.1.mlp.experts.linear_fc1.weight9 + module.decoder.layers.1.mlp.experts.linear_fc1.weight5 + module.decoder.layers.1.mlp.experts.linear_fc1.weight7 + module.decoder.layers.1.mlp.experts.linear_fc1.weight3 + module.decoder.layers.1.mlp.experts.linear_fc1.weight1 + module.decoder.layers.0.mlp.experts.linear_fc2.weight14 + module.decoder.layers.0.mlp.experts.linear_fc2.weight12 + module.decoder.layers.0.mlp.experts.linear_fc2.weight10 +Params for bucket 88 (40894464 elements, 40894464 padded size): + module.decoder.layers.0.mlp.experts.linear_fc2.weight9 + module.decoder.layers.0.mlp.experts.linear_fc2.weight7 + module.decoder.layers.0.mlp.experts.linear_fc2.weight5 + module.decoder.layers.0.mlp.experts.linear_fc2.weight3 + module.decoder.layers.0.mlp.experts.linear_fc2.weight1 + module.decoder.layers.0.mlp.experts.linear_fc1.weight15 + module.decoder.layers.0.mlp.experts.linear_fc1.weight13 + module.decoder.layers.0.mlp.experts.linear_fc1.weight11 + module.decoder.layers.0.mlp.experts.linear_fc1.weight8 + module.decoder.layers.0.mlp.experts.linear_fc2.weight8 + module.decoder.layers.0.mlp.experts.linear_fc2.weight6 + module.decoder.layers.0.mlp.experts.linear_fc2.weight4 + module.decoder.layers.0.mlp.experts.linear_fc2.weight2 + module.decoder.layers.0.mlp.experts.linear_fc2.weight0 + module.decoder.layers.0.mlp.experts.linear_fc1.weight14 + module.decoder.layers.0.mlp.experts.linear_fc1.weight12 + module.decoder.layers.0.mlp.experts.linear_fc1.weight10 + module.decoder.layers.0.mlp.experts.linear_fc1.weight9 +Params for bucket 89 (25165824 elements, 25165824 padded size): + module.decoder.layers.0.mlp.experts.linear_fc1.weight4 + module.decoder.layers.0.mlp.experts.linear_fc1.weight6 + module.decoder.layers.0.mlp.experts.linear_fc1.weight2 + module.decoder.layers.0.mlp.experts.linear_fc1.weight1 + module.decoder.layers.0.mlp.experts.linear_fc1.weight0 + module.decoder.layers.0.mlp.experts.linear_fc1.weight7 + module.decoder.layers.0.mlp.experts.linear_fc1.weight5 + module.decoder.layers.0.mlp.experts.linear_fc1.weight3 +INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0001, min_lr=1e-05, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp8_recipe='mxfp8', fp16=False, bf16=True, reuse_grad_buf_for_mxfp8_param_ag=False, params_dtype=torch.bfloat16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=True, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=, config_logger_dir='') +INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine +[after model, optimizer, and learning rate scheduler are built] datetime: 2026-08-30 00:17:50 +> building train, validation, and test datasets ... + > datasets target sizes (minimum size): + train: 80 + validation: 0 + test: 0 +INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None] +> building train, validation, and test datasets for GPT ... +INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(80, 0, 0) and config=GPTDatasetConfig(random_seed=1234, sequence_length=4096, blend=(['/workspace/data/wikitext/wikitext103_gpt2_text_document'], None), blend_per_split=None, split='100,0,0', split_matrix=[(0, 1.0), None, None], num_dataset_builder_threads=1, path_to_cache=None, mmap_bin_files=True, mock=False, tokenizer=, mid_level_dataset_surplus=0.005, reset_position_ids=False, reset_attention_mask=False, eod_mask_loss=False, create_attention_mask=True, drop_last_partial_validation_sequence=True, add_extra_token_to_sequence=True, object_storage_cache_path=None) +INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx +INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths +INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers +INFO:megatron.core.datasets.indexed_dataset: Extract the document indices +INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029 +INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350 +INFO:megatron.core.datasets.gpt_dataset:Build and save the GPTDataset train indices +INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073 +INFO:megatron.core.datasets.gpt_dataset:> total number of epochs: 1 +> finished creating GPT datasets ... +[after dataloaders are built] datetime: 2026-08-30 00:17:50 +done with setup ... +(min, max) time across ranks (ms): + model-and-optimizer-setup ......................: (1049.83, 1296.05) + train/valid/test-data-iterators-setup ..........: (144.74, 163.13) +training ... +Overwriting rerun_state_machine.current_iteration from -1 to 0... +[before the start of training step] datetime: 2026-08-30 00:17:50 +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future. + warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.") + [2026-08-30 00:18:36] iteration 1/ 10 | consumed samples: 8 | elapsed time per iteration (ms): 45858.3 | throughput per GPU (TFLOP/s/GPU): 1.9 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.095255E+01 | loss scale: 1.0 | grad norm: 54.339 | number of skipped iterations: 0 | number of nan iterations: 0 | +Number of parameters in transformer block in billions: 29.90 +Number of parameters in embedding layers in billions: 0.21 +Total number of parameters in billions: 30.10 +Number of parameters in most loaded shard in billions: 30.1034 +Theoretical memory footprints: weight and optimizer=215316.59 MB +[Rank 0] (after 1 iterations) memory (MB) | allocated: 80701.08984375 | max allocated: 87236.40185546875 | reserved: 88812.0 | max reserved: 88812.0 +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") +/workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future. + warnings.warn("The use_host_meta is deprecated, it will be removed in the future.") + [2026-08-30 00:18:37] iteration 2/ 10 | consumed samples: 16 | elapsed time per iteration (ms): 1250.6 | throughput per GPU (TFLOP/s/GPU): 71.3 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.095489E+01 | loss scale: 1.0 | grad norm: 56.497 | number of skipped iterations: 0 | number of nan iterations: 0 | + [2026-08-30 00:18:38] iteration 3/ 10 | consumed samples: 24 | elapsed time per iteration (ms): 1225.6 | throughput per GPU (TFLOP/s/GPU): 72.7 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 1.008554E+01 | loss scale: 1.0 | grad norm: 179.717 | number of skipped iterations: 0 | number of nan iterations: 0 | +Capture CUDA graph for training!!! +CUDA graph capture done!!! + [2026-08-30 00:18:42] iteration 4/ 10 | consumed samples: 32 | elapsed time per iteration (ms): 3308.7 | throughput per GPU (TFLOP/s/GPU): 26.9 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 9.603868E+00 | loss scale: 1.0 | grad norm: 8.801 | number of skipped iterations: 0 | number of nan iterations: 0 | + [2026-08-30 00:18:42] iteration 5/ 10 | consumed samples: 40 | elapsed time per iteration (ms): 401.6 | throughput per GPU (TFLOP/s/GPU): 221.9 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 9.309317E+00 | loss scale: 1.0 | grad norm: 4.144 | number of skipped iterations: 0 | number of nan iterations: 0 | + [2026-08-30 00:18:42] iteration 6/ 10 | consumed samples: 48 | elapsed time per iteration (ms): 398.8 | throughput per GPU (TFLOP/s/GPU): 223.5 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 9.279056E+00 | loss scale: 1.0 | grad norm: 6.541 | number of skipped iterations: 0 | number of nan iterations: 0 | + [2026-08-30 00:18:46] iteration 7/ 10 | consumed samples: 56 | elapsed time per iteration (ms): 3269.4 | throughput per GPU (TFLOP/s/GPU): 27.3 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 9.102093E+00 | loss scale: 1.0 | grad norm: 3.525 | number of skipped iterations: 0 | number of nan iterations: 0 | + [2026-08-30 00:18:46] iteration 8/ 10 | consumed samples: 64 | elapsed time per iteration (ms): 458.5 | throughput per GPU (TFLOP/s/GPU): 194.4 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.907867E+00 | loss scale: 1.0 | grad norm: 3.404 | number of skipped iterations: 0 | number of nan iterations: 0 | +Capture range started in the application. +Capture range ended in the application. +Generating '/tmp/nsys-report-b003.qdstrm' + [1/1] [0% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-repProcessing events... + [1/1] [0% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [5% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [0% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [5% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [6% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [7% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [8% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [9% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [10% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [11% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [12% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [13% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [14% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=15% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=16% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=17% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==18% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==19% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==20% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==21% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [===22% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [===23% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [===24% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [====25% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [====26% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [====27% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [====28% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=====29% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=====30% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=====31% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=====32% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [======33% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [======34% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [======35% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=======36% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=======37% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=======38% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=======39% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [========40% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [========41% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [========42% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=========43% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=========44% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=========45% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=========46% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==========47% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==========48% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==========49% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [===========50% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [===========51% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [===========52% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [===========53% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [============54% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [============55% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [============56% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [============57% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=============58% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=============59% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [=============60% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==============61% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==============62% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [==============63% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [================70% ] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [========================100%] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep [1/1] [========================100%] QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125.nsys-rep diff --git a/tierb-fixed-moonep-full/hf_watcher.log b/tierb-fixed-moonep-full/hf_watcher.log index 2863c669e5b9b97bec08bd6f5d42e3efee2cb98f..fa2a228ba5d5d19661aed45a4d71b18834e6e950 100644 --- a/tierb-fixed-moonep-full/hf_watcher.log +++ b/tierb-fixed-moonep-full/hf_watcher.log @@ -27,3 +27,4 @@ Watching /workspace/tierb-fixed-moonep-full -> HF every 600s (kill to stop) [2026-08-30T00:10:16+00:00] backup pass uploaded: /workspace/tierb-fixed-moonep-full -> MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/tierb-fixed-moonep-full https://huggingface.co/MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/tree/main +[2026-08-30T00:20:20+00:00] backup pass