Upload folder using huggingface_hub

checkpoint-4000/config.json

@@ -0,0 +1,70 @@
+{
+  "action_horizon": 50,
+  "add_pos_embed": true,
+  "apply_sincos_state_encoding": true,
+  "architectures": [
+    "Gr00tN1d6"
+  ],
+  "attn_dropout": 0.2,
+  "attn_implementation": null,
+  "backbone_embedding_dim": 2048,
+  "backbone_model_type": "eagle",
+  "backbone_trainable_params_fp32": true,
+  "collator_overwrite_image_inputs": false,
+  "color_jitter_params": {
+    "brightness": 0.1,
+    "contrast": 0.1,
+    "hue": 0.1,
+    "saturation": 0.1
+  },
+  "crop_fraction": 0.95,
+  "diffusion_model_cfg": {
+    "attention_head_dim": 48,
+    "dropout": 0.2,
+    "final_dropout": true,
+    "interleave_self_attention": true,
+    "norm_type": "ada_norm",
+    "num_attention_heads": 32,
+    "num_layers": 32,
+    "output_dim": 1024,
+    "positional_embeddings": null
+  },
+  "eagle_collator": true,
+  "formalize_language": true,
+  "gemma_collator": false,
+  "hidden_size": 1024,
+  "image_crop_size": null,
+  "image_target_size": null,
+  "input_embedding_dim": 1536,
+  "load_bf16": true,
+  "max_action_dim": 128,
+  "max_num_embodiments": 32,
+  "max_seq_len": 1024,
+  "max_state_dim": 128,
+  "model_dtype": "bfloat16",
+  "model_name": "nvidia/Eagle-Block2A-2B-v2",
+  "model_type": "Gr00tN1d6",
+  "noise_beta_alpha": 1.5,
+  "noise_beta_beta": 1.0,
+  "noise_s": 0.999,
+  "num_inference_timesteps": 4,
+  "num_timestep_buckets": 1000,
+  "random_rotation_angle": null,
+  "reproject_vision": false,
+  "select_layer": 16,
+  "shortest_image_edge": 256,
+  "state_dropout_prob": 0.0,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "tune_diffusion_model": true,
+  "tune_llm": false,
+  "tune_projector": true,
+  "tune_top_llm_layers": 4,
+  "tune_visual": false,
+  "tune_vlln": true,
+  "use_albumentations_transforms": true,
+  "use_alternate_vl_dit": true,
+  "use_flash_attention": true,
+  "use_relative_action": true,
+  "use_vlln": true
+}

checkpoint-4000/embodiment_id.json

@@ -0,0 +1,11 @@
+{
+  "robocasa_panda_omron": 13,
+  "gr1": 20,
+  "behavior_r1_pro": 24,
+  "unitree_g1": 8,
+  "oxe_google": 0,
+  "oxe_widowx": 1,
+  "libero_panda": 2,
+  "oxe_droid": 16,
+  "new_embodiment": 10
+}

checkpoint-4000/experiment_cfg/conf.yaml

@@ -0,0 +1,252 @@
+load_config_path: null
+model:
+  model_type: Gr00tN1d6
+  model_dtype: bfloat16
+  model_name: nvidia/Eagle-Block2A-2B-v2
+  backbone_model_type: eagle
+  model_revision: null
+  tune_top_llm_layers: 4
+  backbone_embedding_dim: 2048
+  tune_llm: false
+  tune_visual: false
+  select_layer: 16
+  reproject_vision: false
+  use_flash_attention: true
+  load_bf16: false
+  collator_overwrite_image_inputs: false
+  eagle_collator: true
+  backbone_trainable_params_fp32: true
+  image_crop_size: null
+  image_target_size: null
+  shortest_image_edge: 256
+  crop_fraction: 0.95
+  random_rotation_angle: null
+  color_jitter_params:
+    brightness: 0.3
+    contrast: 0.4
+    saturation: 0.5
+    hue: 0.08
+  use_albumentations_transforms: true
+  formalize_language: true
+  apply_sincos_state_encoding: false
+  use_relative_action: true
+  max_state_dim: 29
+  max_action_dim: 29
+  action_horizon: 16
+  hidden_size: 1024
+  input_embedding_dim: 1536
+  add_pos_embed: true
+  attn_dropout: 0.2
+  use_vlln: true
+  max_seq_len: 1024
+  use_alternate_vl_dit: true
+  attend_text_every_n_blocks: 2
+  diffusion_model_cfg:
+    positional_embeddings: null
+    num_layers: 32
+    num_attention_heads: 32
+    attention_head_dim: 48
+    norm_type: ada_norm
+    dropout: 0.2
+    final_dropout: true
+    output_dim: 1024
+    interleave_self_attention: true
+  num_inference_timesteps: 4
+  noise_beta_alpha: 1.5
+  noise_beta_beta: 1.0
+  noise_s: 0.999
+  num_timestep_buckets: 1000
+  tune_projector: true
+  tune_diffusion_model: true
+  tune_vlln: true
+  state_dropout_prob: 0.0
+  state_additive_noise_scale: 0.0
+  max_num_embodiments: 32
+data:
+  datasets:
+  - dataset_paths:
+    - /data/PnPCubeLine
+    embodiment_tag: unitree_g1
+    mix_ratio: 1.0
+    dataset_type: physical_embodiment
+    val_dataset_path: null
+  modality_configs:
+    unitree_g1:
+      video:
+        delta_indices:
+        - 0
+        modality_keys:
+        - ego_view
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+      state:
+        delta_indices:
+        - 0
+        modality_keys:
+        - left_leg
+        - right_leg
+        - waist
+        - left_arm
+        - right_arm
+        - left_hand
+        - right_hand
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+      action:
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        - 24
+        - 25
+        - 26
+        - 27
+        - 28
+        - 29
+        modality_keys:
+        - left_arm
+        - right_arm
+        - left_hand
+        - right_hand
+        - waist
+        - base_height_command
+        - navigate_command
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs:
+        - rep: RELATIVE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: RELATIVE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+      language:
+        delta_indices:
+        - 0
+        modality_keys:
+        - annotation.human.task_description
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+  download_cache: false
+  shard_size: 1024
+  episode_sampling_rate: 0.1
+  num_shards_per_epoch: 100000
+  override_pretraining_statistics: false
+  mode: single_turn
+  random_chop: 0.0
+  mock_dataset_mode: false
+  shuffle: true
+  seed: 42
+  multiprocessing_context: fork
+  allow_padding: false
+  subsample_ratio: 1.0
+  image_crop_size:
+  - 244
+  - 244
+  image_target_size:
+  - 224
+  - 224
+  video_backend: torchcodec
+training:
+  output_dir: /checkpoints/g1_finetune
+  experiment_name: null
+  max_steps: 6000
+  global_batch_size: 512
+  batch_size: null
+  gradient_accumulation_steps: 1
+  learning_rate: 0.0001
+  lr_scheduler_type: cosine
+  weight_decay: 1.0e-05
+  warmup_ratio: 0.05
+  warmup_steps: 0
+  max_grad_norm: 1.0
+  optim: adamw_torch
+  start_from_checkpoint: nvidia/GR00T-N1.6-3B
+  tf32: true
+  fp16: false
+  bf16: true
+  eval_bf16: true
+  logging_steps: 10
+  save_steps: 2000
+  save_total_limit: 2
+  save_vl_model: false
+  upload_checkpoints: false
+  upload_every: 1000
+  upload_last_n_checkpoints: 5
+  max_concurrent_uploads: 2
+  eval_strategy: 'no'
+  eval_steps: 500
+  eval_set_split_ratio: 0.1
+  eval_batch_size: 2
+  save_best_eval_metric_name: ''
+  save_best_eval_metric_greater_is_better: true
+  deepspeed_stage: 2
+  gradient_checkpointing: false
+  transformers_trust_remote_code: true
+  transformers_local_files_only: false
+  transformers_cache_dir: null
+  transformers_access_token: null
+  use_ddp: false
+  ddp_bucket_cap_mb: 100
+  num_gpus: 4
+  dataloader_num_workers: 6
+  remove_unused_columns: false
+  use_wandb: false
+  wandb_project: finetune-gr00t-n1d6
+  enable_profiling: false
+  max_retries: 3
+  assert_loss_less_than: null
+  add_rl_callback: false
+  enable_open_loop_eval: false
+  open_loop_eval_traj_ids:
+  - 0
+  open_loop_eval_steps_per_traj: 100
+  open_loop_eval_plot_indices: null
+max_steps: 6000
+save_steps: 2000

checkpoint-4000/experiment_cfg/config.yaml

@@ -0,0 +1,291 @@
+!!python/object:gr00t.configs.base_config.Config
+data: !!python/object:gr00t.configs.data.data_config.DataConfig
+  allow_padding: false
+  datasets:
+  - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
+    dataset_paths:
+    - /data/PnPCubeLine
+    dataset_type: physical_embodiment
+    embodiment_tag: unitree_g1
+    mix_ratio: 1.0
+    val_dataset_path: null
+  download_cache: false
+  episode_sampling_rate: 0.1
+  image_crop_size:
+  - 244
+  - 244
+  image_target_size:
+  - 224
+  - 224
+  mock_dataset_mode: false
+  modality_configs:
+    unitree_g1:
+      action: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs:
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
+          - default
+          rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
+          - relative
+          state_key: null
+          type: &id003 !!python/object/apply:gr00t.data.types.ActionType
+          - non_eef
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id002
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: &id004 !!python/object/apply:gr00t.data.types.ActionRepresentation
+          - absolute
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id004
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id004
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id004
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id004
+          state_key: null
+          type: *id003
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        - 24
+        - 25
+        - 26
+        - 27
+        - 28
+        - 29
+        mean_std_embedding_keys: null
+        modality_keys:
+        - left_arm
+        - right_arm
+        - left_hand
+        - right_hand
+        - waist
+        - base_height_command
+        - navigate_command
+        sin_cos_embedding_keys: null
+      language: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - 0
+        mean_std_embedding_keys: null
+        modality_keys:
+        - annotation.human.task_description
+        sin_cos_embedding_keys: null
+      state: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - 0
+        mean_std_embedding_keys: null
+        modality_keys:
+        - left_leg
+        - right_leg
+        - waist
+        - left_arm
+        - right_arm
+        - left_hand
+        - right_hand
+        sin_cos_embedding_keys: null
+      video: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - 0
+        mean_std_embedding_keys: null
+        modality_keys:
+        - ego_view
+        sin_cos_embedding_keys: null
+  mode: single_turn
+  multiprocessing_context: fork
+  num_shards_per_epoch: 100000
+  override_pretraining_statistics: false
+  random_chop: 0.0
+  seed: 42
+  shard_size: 1024
+  shuffle: true
+  subsample_ratio: 1.0
+  video_backend: torchcodec
+load_config_path: null
+model: !!python/object:gr00t.configs.model.gr00t_n1d6.Gr00tN1d6Config
+  _attn_implementation_autoset: false
+  _attn_implementation_internal: null
+  _commit_hash: null
+  _name_or_path: ''
+  add_cross_attention: false
+  architectures: null
+  backbone_model_type: eagle
+  backbone_trainable_params_fp32: true
+  bad_words_ids: null
+  begin_suppress_tokens: null
+  bos_token_id: null
+  chunk_size_feed_forward: 0
+  color_jitter_params:
+    brightness: 0.3
+    contrast: 0.4
+    hue: 0.08
+    saturation: 0.5
+  cross_attention_hidden_size: null
+  decoder_start_token_id: null
+  diffusion_model_cfg:
+    attention_head_dim: 48
+    dropout: 0.2
+    final_dropout: true
+    interleave_self_attention: true
+    norm_type: ada_norm
+    num_attention_heads: 32
+    num_layers: 32
+    output_dim: 1024
+    positional_embeddings: null
+  diversity_penalty: 0.0
+  do_sample: false
+  eagle_collator: true
+  early_stopping: false
+  encoder_no_repeat_ngram_size: 0
+  eos_token_id: null
+  exponential_decay_length_penalty: null
+  finetuning_task: null
+  forced_bos_token_id: null
+  forced_eos_token_id: null
+  id2label:
+    0: LABEL_0
+    1: LABEL_1
+  is_decoder: false
+  is_encoder_decoder: false
+  label2id:
+    LABEL_0: 0
+    LABEL_1: 1
+  length_penalty: 1.0
+  load_bf16: false
+  max_length: 20
+  min_length: 0
+  model_name: nvidia/Eagle-Block2A-2B-v2
+  no_repeat_ngram_size: 0
+  num_beam_groups: 1
+  num_beams: 1
+  num_return_sequences: 1
+  output_attentions: false
+  output_hidden_states: false
+  output_scores: false
+  pad_token_id: null
+  prefix: null
+  problem_type: null
+  pruned_heads: {}
+  random_rotation_angle: null
+  remove_invalid_values: false
+  repetition_penalty: 1.0
+  reproject_vision: false
+  return_dict: true
+  return_dict_in_generate: false
+  sep_token_id: null
+  state_dropout_prob: 0.0
+  suppress_tokens: null
+  task_specific_params: null
+  temperature: 1.0
+  tf_legacy_loss: false
+  tie_encoder_decoder: false
+  tie_word_embeddings: true
+  tokenizer_class: null
+  top_k: 50
+  top_p: 1.0
+  torch_dtype: null
+  torchscript: false
+  transformers_version: null
+  tune_diffusion_model: true
+  tune_llm: false
+  tune_projector: true
+  tune_visual: false
+  typical_p: 1.0
+  use_bfloat16: false
+  use_relative_action: true
+training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
+  add_rl_callback: false
+  assert_loss_less_than: null
+  batch_size: null
+  bf16: true
+  dataloader_num_workers: 6
+  ddp_bucket_cap_mb: 100
+  deepspeed_stage: 2
+  enable_open_loop_eval: false
+  enable_profiling: false
+  eval_batch_size: 2
+  eval_bf16: true
+  eval_set_split_ratio: 0.1
+  eval_steps: 500
+  eval_strategy: 'no'
+  experiment_name: null
+  fp16: false
+  global_batch_size: 512
+  gradient_accumulation_steps: 1
+  gradient_checkpointing: false
+  learning_rate: 0.0001
+  logging_steps: 10
+  lr_scheduler_type: cosine
+  max_concurrent_uploads: 2
+  max_grad_norm: 1.0
+  max_retries: 3
+  max_steps: 6000
+  num_gpus: 4
+  open_loop_eval_plot_indices: null
+  open_loop_eval_steps_per_traj: 100
+  open_loop_eval_traj_ids:
+  - 0
+  optim: adamw_torch
+  output_dir: /checkpoints/g1_finetune
+  remove_unused_columns: false
+  save_best_eval_metric_greater_is_better: true
+  save_best_eval_metric_name: ''
+  save_steps: 2000
+  save_total_limit: 2
+  save_vl_model: false
+  start_from_checkpoint: nvidia/GR00T-N1.6-3B
+  tf32: true
+  transformers_access_token: null
+  transformers_cache_dir: null
+  transformers_local_files_only: false
+  transformers_trust_remote_code: true
+  upload_checkpoints: false
+  upload_every: 1000
+  upload_last_n_checkpoints: 5
+  use_ddp: false
+  use_wandb: false
+  wandb_project: finetune-gr00t-n1d6
+  warmup_ratio: 0.05
+  warmup_steps: 0
+  weight_decay: 1.0e-05

checkpoint-4000/experiment_cfg/final_model_config.json

@@ -0,0 +1,53 @@
+{
+  "model_type": "Gr00tN1d6",
+  "model_dtype": "bfloat16",
+  "model_name": "nvidia/Eagle-Block2A-2B-v2",
+  "backbone_model_type": "eagle",
+  "model_revision": null,
+  "tune_top_llm_layers": 4,
+  "backbone_embedding_dim": 2048,
+  "tune_llm": false,
+  "tune_visual": false,
+  "select_layer": 16,
+  "reproject_vision": false,
+  "use_flash_attention": true,
+  "load_bf16": true,
+  "collator_overwrite_image_inputs": false,
+  "eagle_collator": true,
+  "backbone_trainable_params_fp32": true,
+  "apply_sincos_state_encoding": true,
+  "use_relative_action": true,
+  "max_state_dim": 128,
+  "max_action_dim": 128,
+  "action_horizon": 50,
+  "hidden_size": 1024,
+  "input_embedding_dim": 1536,
+  "add_pos_embed": true,
+  "attn_dropout": 0.2,
+  "use_vlln": true,
+  "max_seq_len": 1024,
+  "use_alternate_vl_dit": true,
+  "attend_text_every_n_blocks": 2,
+  "diffusion_model_cfg": {
+    "attention_head_dim": 48,
+    "dropout": 0.2,
+    "final_dropout": true,
+    "interleave_self_attention": true,
+    "norm_type": "ada_norm",
+    "num_attention_heads": 32,
+    "num_layers": 32,
+    "output_dim": 1024,
+    "positional_embeddings": null
+  },
+  "num_inference_timesteps": 4,
+  "noise_beta_alpha": 1.5,
+  "noise_beta_beta": 1.0,
+  "noise_s": 0.999,
+  "num_timestep_buckets": 1000,
+  "tune_projector": true,
+  "tune_diffusion_model": true,
+  "tune_vlln": true,
+  "state_dropout_prob": 0.0,
+  "state_additive_noise_scale": 0.0,
+  "max_num_embodiments": 32
+}

checkpoint-4000/global_step4000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3119b9ba23e36428ae55e9e1d0c3900eb58683bbae2020edd8e8f658b9c45b2e
+size 4859923697

checkpoint-4000/global_step4000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3be7efbf34e0c90d79fa285371ed108c7f6d8b2113ad5e0f58be74287d943b0
+size 4859923761

checkpoint-4000/global_step4000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b74ef7944e50282f18551511fab67660dc5962c0c57125cf1b88896ae639787f
+size 4859923569

checkpoint-4000/global_step4000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90c5e1afc967bfb361e0d9f2fb9a42945690fad8e48269432ac62d35b7729589
+size 4859917873

checkpoint-4000/global_step4000/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0167eeda5fa81617a214708b8358332c5beb9ddccd03d63194b62acc7ad08f0f
+size 9907202435

checkpoint-4000/latest

@@ -0,0 +1 @@
+global_step4000

checkpoint-4000/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b21864df58a76a4372a0049731ae1c343137e01644798e208523907ede4c8cc
+size 4991091456

checkpoint-4000/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a01e9468a42fcf825c860edcff7be66f5bde3924216e42b3a99ac9b122d63d3
+size 1582283096

checkpoint-4000/processor_config.json

@@ -0,0 +1,508 @@
+{
+  "processor_class": "Gr00tN1d6Processor",
+  "processor_kwargs": {
+    "modality_configs": {
+      "behavior_r1_pro": {
+        "video": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "observation.images.rgb.head_256_256",
+            "observation.images.rgb.left_wrist_256_256",
+            "observation.images.rgb.right_wrist_256_256"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "state": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "robot_pos",
+            "robot_ori_cos",
+            "robot_ori_sin",
+            "robot_2d_ori",
+            "robot_2d_ori_cos",
+            "robot_2d_ori_sin",
+            "robot_lin_vel",
+            "robot_ang_vel",
+            "arm_left_qpos",
+            "arm_left_qpos_sin",
+            "arm_left_qpos_cos",
+            "eef_left_pos",
+            "eef_left_quat",
+            "gripper_left_qpos",
+            "arm_right_qpos",
+            "arm_right_qpos_sin",
+            "arm_right_qpos_cos",
+            "eef_right_pos",
+            "eef_right_quat",
+            "gripper_right_qpos",
+            "trunk_qpos"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "action": {
+          "delta_indices": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+            12,
+            13,
+            14,
+            15,
+            16,
+            17,
+            18,
+            19,
+            20,
+            21,
+            22,
+            23,
+            24,
+            25,
+            26,
+            27,
+            28,
+            29,
+            30,
+            31
+          ],
+          "modality_keys": [
+            "base",
+            "torso",
+            "left_arm",
+            "left_gripper",
+            "right_arm",
+            "right_gripper"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": [
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": "trunk_qpos"
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": "arm_left_qpos"
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": "arm_right_qpos"
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            }
+          ]
+        },
+        "language": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "annotation.human.coarse_action"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        }
+      },
+      "gr1": {
+        "video": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "ego_view_bg_crop_pad_res256_freq20"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "state": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand",
+            "waist"
+          ],
+          "sin_cos_embedding_keys": [
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand",
+            "waist"
+          ],
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "action": {
+          "delta_indices": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+            12,
+            13,
+            14,
+            15
+          ],
+          "modality_keys": [
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand",
+            "waist"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": [
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            }
+          ]
+        },
+        "language": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "task"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        }
+      },
+      "robocasa_panda_omron": {
+        "video": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "res256_image_side_0",
+            "res256_image_side_1",
+            "res256_image_wrist_0"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "state": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "end_effector_position_relative",
+            "end_effector_rotation_relative",
+            "gripper_qpos",
+            "base_position",
+            "base_rotation"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "action": {
+          "delta_indices": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+            12,
+            13,
+            14,
+            15
+          ],
+          "modality_keys": [
+            "end_effector_position",
+            "end_effector_rotation",
+            "gripper_close",
+            "base_motion",
+            "control_mode"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": [
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            }
+          ]
+        },
+        "language": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "annotation.human.action.task_description"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        }
+      },
+      "unitree_g1": {
+        "video": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "ego_view"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "state": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "left_leg",
+            "right_leg",
+            "waist",
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "action": {
+          "delta_indices": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+            12,
+            13,
+            14,
+            15,
+            16,
+            17,
+            18,
+            19,
+            20,
+            21,
+            22,
+            23,
+            24,
+            25,
+            26,
+            27,
+            28,
+            29
+          ],
+          "modality_keys": [
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand",
+            "waist",
+            "base_height_command",
+            "navigate_command"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": [
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            }
+          ]
+        },
+        "language": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "annotation.human.task_description"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        }
+      }
+    },
+    "image_crop_size": null,
+    "image_target_size": null,
+    "use_albumentations": true,
+    "random_rotation_angle": null,
+    "color_jitter_params": {
+      "brightness": 0.3,
+      "contrast": 0.4,
+      "saturation": 0.5,
+      "hue": 0.08
+    },
+    "shortest_image_edge": 256,
+    "crop_fraction": 0.95,
+    "model_name": "nvidia/Eagle-Block2A-2B-v2",
+    "model_type": "eagle",
+    "formalize_language": true,
+    "max_state_dim": 128,
+    "max_action_dim": 128,
+    "max_action_horizon": 50,
+    "use_percentiles": false,
+    "clip_outliers": true,
+    "apply_sincos_state_encoding": true,
+    "use_relative_action": true
+  }
+}

checkpoint-4000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec6e49911abd2fc83f053e1132fac39188a0cb116e9bcc50d5838ef806a074aa
+size 15429

checkpoint-4000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9cb4df49d6d89c24b739036627a9d60424a477c08a177a460a5101db99cac40
+size 15429

checkpoint-4000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6eedcf354541f2bd34492c7e662db5818f5d506c0fe3a8ba2b4801c0787c351d
+size 15429

checkpoint-4000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76ff614b9d971d68080f9f1a9a41d0e375160e7d554e294957d190ce405f6763
+size 15429

checkpoint-4000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93727dfaa883ba52dca9378688334804bd7142a7af39e64fe8ca4bf76f6b302c
+size 1465

checkpoint-4000/trainer_state.json

@@ -0,0 +1,2434 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.6666666666666666,
+  "eval_steps": 500,
+  "global_step": 4000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "grad_norm": 1.3613678216934204,
+      "learning_rate": 3e-06,
+      "loss": 1.4557,
+      "step": 10
+    },
+    {
+      "grad_norm": 0.8573107719421387,
+      "learning_rate": 6.333333333333334e-06,
+      "loss": 1.424,
+      "step": 20
+    },
+    {
+      "grad_norm": 0.34049543738365173,
+      "learning_rate": 9.666666666666667e-06,
+      "loss": 1.3391,
+      "step": 30
+    },
+    {
+      "grad_norm": 0.44315752387046814,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 1.2758,
+      "step": 40
+    },
+    {
+      "grad_norm": 0.49191826581954956,
+      "learning_rate": 1.6333333333333335e-05,
+      "loss": 1.175,
+      "step": 50
+    },
+    {
+      "grad_norm": 0.3820122480392456,
+      "learning_rate": 1.9666666666666666e-05,
+      "loss": 1.1102,
+      "step": 60
+    },
+    {
+      "grad_norm": 0.36782893538475037,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 1.0885,
+      "step": 70
+    },
+    {
+      "grad_norm": 0.28842034935951233,
+      "learning_rate": 2.633333333333333e-05,
+      "loss": 1.0746,
+      "step": 80
+    },
+    {
+      "grad_norm": 0.43426793813705444,
+      "learning_rate": 2.9666666666666672e-05,
+      "loss": 1.0604,
+      "step": 90
+    },
+    {
+      "grad_norm": 0.34185388684272766,
+      "learning_rate": 3.3e-05,
+      "loss": 1.0523,
+      "step": 100
+    },
+    {
+      "grad_norm": 0.29589465260505676,
+      "learning_rate": 3.633333333333333e-05,
+      "loss": 1.0355,
+      "step": 110
+    },
+    {
+      "grad_norm": 0.42089858651161194,
+      "learning_rate": 3.966666666666667e-05,
+      "loss": 1.0121,
+      "step": 120
+    },
+    {
+      "grad_norm": 0.4485030472278595,
+      "learning_rate": 4.3e-05,
+      "loss": 0.98,
+      "step": 130
+    },
+    {
+      "grad_norm": 0.3440701365470886,
+      "learning_rate": 4.633333333333333e-05,
+      "loss": 0.9491,
+      "step": 140
+    },
+    {
+      "grad_norm": 0.7382627725601196,
+      "learning_rate": 4.966666666666667e-05,
+      "loss": 0.9054,
+      "step": 150
+    },
+    {
+      "grad_norm": 0.6498928070068359,
+      "learning_rate": 5.300000000000001e-05,
+      "loss": 0.8479,
+      "step": 160
+    },
+    {
+      "grad_norm": 0.6362508535385132,
+      "learning_rate": 5.633333333333334e-05,
+      "loss": 0.7839,
+      "step": 170
+    },
+    {
+      "grad_norm": 0.8518823385238647,
+      "learning_rate": 5.966666666666667e-05,
+      "loss": 0.7336,
+      "step": 180
+    },
+    {
+      "grad_norm": 1.1828296184539795,
+      "learning_rate": 6.3e-05,
+      "loss": 0.6629,
+      "step": 190
+    },
+    {
+      "grad_norm": 0.944200336933136,
+      "learning_rate": 6.633333333333334e-05,
+      "loss": 0.6056,
+      "step": 200
+    },
+    {
+      "grad_norm": 0.758030354976654,
+      "learning_rate": 6.966666666666668e-05,
+      "loss": 0.5545,
+      "step": 210
+    },
+    {
+      "grad_norm": 1.2042676210403442,
+      "learning_rate": 7.3e-05,
+      "loss": 0.5044,
+      "step": 220
+    },
+    {
+      "grad_norm": 1.0285093784332275,
+      "learning_rate": 7.633333333333334e-05,
+      "loss": 0.4607,
+      "step": 230
+    },
+    {
+      "grad_norm": 1.126552700996399,
+      "learning_rate": 7.966666666666666e-05,
+      "loss": 0.4219,
+      "step": 240
+    },
+    {
+      "grad_norm": 0.7760715484619141,
+      "learning_rate": 8.3e-05,
+      "loss": 0.3931,
+      "step": 250
+    },
+    {
+      "grad_norm": 1.1023366451263428,
+      "learning_rate": 8.633333333333334e-05,
+      "loss": 0.362,
+      "step": 260
+    },
+    {
+      "grad_norm": 0.9510546326637268,
+      "learning_rate": 8.966666666666666e-05,
+      "loss": 0.3313,
+      "step": 270
+    },
+    {
+      "grad_norm": 0.8212385177612305,
+      "learning_rate": 9.300000000000001e-05,
+      "loss": 0.2998,
+      "step": 280
+    },
+    {
+      "grad_norm": 0.8118603229522705,
+      "learning_rate": 9.633333333333335e-05,
+      "loss": 0.28,
+      "step": 290
+    },
+    {
+      "grad_norm": 1.409279704093933,
+      "learning_rate": 9.966666666666667e-05,
+      "loss": 0.2516,
+      "step": 300
+    },
+    {
+      "grad_norm": 2.358229398727417,
+      "learning_rate": 9.999938485971279e-05,
+      "loss": 0.2319,
+      "step": 310
+    },
+    {
+      "grad_norm": 0.9309172630310059,
+      "learning_rate": 9.999725846827562e-05,
+      "loss": 0.2257,
+      "step": 320
+    },
+    {
+      "grad_norm": 0.9242991805076599,
+      "learning_rate": 9.999361329594254e-05,
+      "loss": 0.209,
+      "step": 330
+    },
+    {
+      "grad_norm": 0.9545067548751831,
+      "learning_rate": 9.998844945344405e-05,
+      "loss": 0.1922,
+      "step": 340
+    },
+    {
+      "grad_norm": 0.9418713450431824,
+      "learning_rate": 9.99817670976436e-05,
+      "loss": 0.175,
+      "step": 350
+    },
+    {
+      "grad_norm": 0.8053094744682312,
+      "learning_rate": 9.997356643153303e-05,
+      "loss": 0.161,
+      "step": 360
+    },
+    {
+      "grad_norm": 1.237301230430603,
+      "learning_rate": 9.996384770422629e-05,
+      "loss": 0.1531,
+      "step": 370
+    },
+    {
+      "grad_norm": 1.0992234945297241,
+      "learning_rate": 9.995261121095194e-05,
+      "loss": 0.1391,
+      "step": 380
+    },
+    {
+      "grad_norm": 1.061476707458496,
+      "learning_rate": 9.993985729304408e-05,
+      "loss": 0.1302,
+      "step": 390
+    },
+    {
+      "grad_norm": 0.8854357004165649,
+      "learning_rate": 9.992558633793212e-05,
+      "loss": 0.1156,
+      "step": 400
+    },
+    {
+      "grad_norm": 0.7982974648475647,
+      "learning_rate": 9.990979877912891e-05,
+      "loss": 0.108,
+      "step": 410
+    },
+    {
+      "grad_norm": 0.8326540589332581,
+      "learning_rate": 9.989249509621759e-05,
+      "loss": 0.1122,
+      "step": 420
+    },
+    {
+      "grad_norm": 0.7788134813308716,
+      "learning_rate": 9.987367581483705e-05,
+      "loss": 0.1114,
+      "step": 430
+    },
+    {
+      "grad_norm": 0.8146352767944336,
+      "learning_rate": 9.985334150666592e-05,
+      "loss": 0.1074,
+      "step": 440
+    },
+    {
+      "grad_norm": 0.745625913143158,
+      "learning_rate": 9.983149278940526e-05,
+      "loss": 0.105,
+      "step": 450
+    },
+    {
+      "grad_norm": 0.9119800329208374,
+      "learning_rate": 9.980813032675974e-05,
+      "loss": 0.099,
+      "step": 460
+    },
+    {
+      "grad_norm": 0.6874001622200012,
+      "learning_rate": 9.978325482841753e-05,
+      "loss": 0.0983,
+      "step": 470
+    },
+    {
+      "grad_norm": 1.0723556280136108,
+      "learning_rate": 9.975686705002867e-05,
+      "loss": 0.0953,
+      "step": 480
+    },
+    {
+      "grad_norm": 0.7956628203392029,
+      "learning_rate": 9.972896779318219e-05,
+      "loss": 0.0813,
+      "step": 490
+    },
+    {
+      "grad_norm": 0.8455336093902588,
+      "learning_rate": 9.969955790538175e-05,
+      "loss": 0.0781,
+      "step": 500
+    },
+    {
+      "grad_norm": 0.8610921502113342,
+      "learning_rate": 9.966863828001982e-05,
+      "loss": 0.0751,
+      "step": 510
+    },
+    {
+      "grad_norm": 0.7537756562232971,
+      "learning_rate": 9.963620985635065e-05,
+      "loss": 0.0692,
+      "step": 520
+    },
+    {
+      "grad_norm": 0.7432633638381958,
+      "learning_rate": 9.960227361946164e-05,
+      "loss": 0.0664,
+      "step": 530
+    },
+    {
+      "grad_norm": 0.9131788611412048,
+      "learning_rate": 9.95668306002435e-05,
+      "loss": 0.0691,
+      "step": 540
+    },
+    {
+      "grad_norm": 0.913440465927124,
+      "learning_rate": 9.952988187535886e-05,
+      "loss": 0.0665,
+      "step": 550
+    },
+    {
+      "grad_norm": 0.7507103681564331,
+      "learning_rate": 9.949142856720961e-05,
+      "loss": 0.0641,
+      "step": 560
+    },
+    {
+      "grad_norm": 0.7820024490356445,
+      "learning_rate": 9.945147184390278e-05,
+      "loss": 0.0695,
+      "step": 570
+    },
+    {
+      "grad_norm": 0.7565670609474182,
+      "learning_rate": 9.941001291921512e-05,
+      "loss": 0.0663,
+      "step": 580
+    },
+    {
+      "grad_norm": 0.6481784582138062,
+      "learning_rate": 9.936705305255612e-05,
+      "loss": 0.0661,
+      "step": 590
+    },
+    {
+      "grad_norm": 0.6356178522109985,
+      "learning_rate": 9.932259354892984e-05,
+      "loss": 0.0641,
+      "step": 600
+    },
+    {
+      "grad_norm": 0.7873478531837463,
+      "learning_rate": 9.927663575889521e-05,
+      "loss": 0.062,
+      "step": 610
+    },
+    {
+      "grad_norm": 0.7071146368980408,
+      "learning_rate": 9.922918107852504e-05,
+      "loss": 0.0639,
+      "step": 620
+    },
+    {
+      "grad_norm": 0.8344459533691406,
+      "learning_rate": 9.918023094936363e-05,
+      "loss": 0.0607,
+      "step": 630
+    },
+    {
+      "grad_norm": 0.7432113885879517,
+      "learning_rate": 9.912978685838294e-05,
+      "loss": 0.0622,
+      "step": 640
+    },
+    {
+      "grad_norm": 0.7232807278633118,
+      "learning_rate": 9.90778503379374e-05,
+      "loss": 0.0595,
+      "step": 650
+    },
+    {
+      "grad_norm": 0.705303966999054,
+      "learning_rate": 9.902442296571743e-05,
+      "loss": 0.0571,
+      "step": 660
+    },
+    {
+      "grad_norm": 0.7458016872406006,
+      "learning_rate": 9.896950636470147e-05,
+      "loss": 0.0652,
+      "step": 670
+    },
+    {
+      "grad_norm": 0.7948721647262573,
+      "learning_rate": 9.891310220310666e-05,
+      "loss": 0.0596,
+      "step": 680
+    },
+    {
+      "grad_norm": 0.78054279088974,
+      "learning_rate": 9.885521219433823e-05,
+      "loss": 0.0564,
+      "step": 690
+    },
+    {
+      "grad_norm": 0.7607908248901367,
+      "learning_rate": 9.879583809693738e-05,
+      "loss": 0.0565,
+      "step": 700
+    },
+    {
+      "grad_norm": 0.6699449419975281,
+      "learning_rate": 9.873498171452789e-05,
+      "loss": 0.0553,
+      "step": 710
+    },
+    {
+      "grad_norm": 0.7117671370506287,
+      "learning_rate": 9.867264489576135e-05,
+      "loss": 0.0555,
+      "step": 720
+    },
+    {
+      "grad_norm": 0.8966276049613953,
+      "learning_rate": 9.860882953426099e-05,
+      "loss": 0.0576,
+      "step": 730
+    },
+    {
+      "grad_norm": 0.8119339346885681,
+      "learning_rate": 9.854353756856412e-05,
+      "loss": 0.0571,
+      "step": 740
+    },
+    {
+      "grad_norm": 0.8719547390937805,
+      "learning_rate": 9.847677098206332e-05,
+      "loss": 0.0554,
+      "step": 750
+    },
+    {
+      "grad_norm": 0.6648759245872498,
+      "learning_rate": 9.840853180294608e-05,
+      "loss": 0.0551,
+      "step": 760
+    },
+    {
+      "grad_norm": 0.8022528290748596,
+      "learning_rate": 9.833882210413332e-05,
+      "loss": 0.0562,
+      "step": 770
+    },
+    {
+      "grad_norm": 0.6207597255706787,
+      "learning_rate": 9.826764400321633e-05,
+      "loss": 0.0553,
+      "step": 780
+    },
+    {
+      "grad_norm": 0.7227312922477722,
+      "learning_rate": 9.819499966239243e-05,
+      "loss": 0.0544,
+      "step": 790
+    },
+    {
+      "grad_norm": 0.5820620059967041,
+      "learning_rate": 9.812089128839938e-05,
+      "loss": 0.0544,
+      "step": 800
+    },
+    {
+      "grad_norm": 0.6752761602401733,
+      "learning_rate": 9.804532113244828e-05,
+      "loss": 0.0511,
+      "step": 810
+    },
+    {
+      "grad_norm": 0.650572657585144,
+      "learning_rate": 9.796829149015517e-05,
+      "loss": 0.0529,
+      "step": 820
+    },
+    {
+      "grad_norm": 0.6513978838920593,
+      "learning_rate": 9.788980470147132e-05,
+      "loss": 0.05,
+      "step": 830
+    },
+    {
+      "grad_norm": 0.6479138731956482,
+      "learning_rate": 9.780986315061218e-05,
+      "loss": 0.0518,
+      "step": 840
+    },
+    {
+      "grad_norm": 0.7067680358886719,
+      "learning_rate": 9.772846926598491e-05,
+      "loss": 0.0568,
+      "step": 850
+    },
+    {
+      "grad_norm": 0.6965780258178711,
+      "learning_rate": 9.76456255201146e-05,
+      "loss": 0.052,
+      "step": 860
+    },
+    {
+      "grad_norm": 0.6729057431221008,
+      "learning_rate": 9.756133442956923e-05,
+      "loss": 0.0502,
+      "step": 870
+    },
+    {
+      "grad_norm": 0.5427769422531128,
+      "learning_rate": 9.747559855488313e-05,
+      "loss": 0.0513,
+      "step": 880
+    },
+    {
+      "grad_norm": 0.6974294781684875,
+      "learning_rate": 9.73884205004793e-05,
+      "loss": 0.0526,
+      "step": 890
+    },
+    {
+      "grad_norm": 0.6217942833900452,
+      "learning_rate": 9.729980291459019e-05,
+      "loss": 0.0523,
+      "step": 900
+    },
+    {
+      "grad_norm": 0.7760837078094482,
+      "learning_rate": 9.720974848917735e-05,
+      "loss": 0.0498,
+      "step": 910
+    },
+    {
+      "grad_norm": 0.624162495136261,
+      "learning_rate": 9.711825995984957e-05,
+      "loss": 0.0499,
+      "step": 920
+    },
+    {
+      "grad_norm": 0.6548763513565063,
+      "learning_rate": 9.702534010577991e-05,
+      "loss": 0.0507,
+      "step": 930
+    },
+    {
+      "grad_norm": 0.74265056848526,
+      "learning_rate": 9.693099174962103e-05,
+      "loss": 0.0512,
+      "step": 940
+    },
+    {
+      "grad_norm": 0.5611060857772827,
+      "learning_rate": 9.683521775741977e-05,
+      "loss": 0.0514,
+      "step": 950
+    },
+    {
+      "grad_norm": 0.7143239974975586,
+      "learning_rate": 9.673802103852979e-05,
+      "loss": 0.0493,
+      "step": 960
+    },
+    {
+      "grad_norm": 0.6779119968414307,
+      "learning_rate": 9.663940454552342e-05,
+      "loss": 0.0499,
+      "step": 970
+    },
+    {
+      "grad_norm": 0.6951384544372559,
+      "learning_rate": 9.65393712741018e-05,
+      "loss": 0.0474,
+      "step": 980
+    },
+    {
+      "grad_norm": 0.6736275553703308,
+      "learning_rate": 9.6437924263004e-05,
+      "loss": 0.0485,
+      "step": 990
+    },
+    {
+      "grad_norm": 0.5961176156997681,
+      "learning_rate": 9.63350665939146e-05,
+      "loss": 0.0477,
+      "step": 1000
+    },
+    {
+      "grad_norm": 0.7116590142250061,
+      "learning_rate": 9.623080139137023e-05,
+      "loss": 0.0466,
+      "step": 1010
+    },
+    {
+      "grad_norm": 0.5824304223060608,
+      "learning_rate": 9.612513182266447e-05,
+      "loss": 0.0475,
+      "step": 1020
+    },
+    {
+      "grad_norm": 0.8167940378189087,
+      "learning_rate": 9.601806109775179e-05,
+      "loss": 0.0485,
+      "step": 1030
+    },
+    {
+      "grad_norm": 0.6272838711738586,
+      "learning_rate": 9.590959246914995e-05,
+      "loss": 0.0466,
+      "step": 1040
+    },
+    {
+      "grad_norm": 0.8012779951095581,
+      "learning_rate": 9.579972923184122e-05,
+      "loss": 0.0487,
+      "step": 1050
+    },
+    {
+      "grad_norm": 0.5613420009613037,
+      "learning_rate": 9.568847472317232e-05,
+      "loss": 0.0448,
+      "step": 1060
+    },
+    {
+      "grad_norm": 0.7028784155845642,
+      "learning_rate": 9.557583232275303e-05,
+      "loss": 0.0474,
+      "step": 1070
+    },
+    {
+      "grad_norm": 0.705496609210968,
+      "learning_rate": 9.546180545235344e-05,
+      "loss": 0.0456,
+      "step": 1080
+    },
+    {
+      "grad_norm": 0.6239430904388428,
+      "learning_rate": 9.534639757580013e-05,
+      "loss": 0.0442,
+      "step": 1090
+    },
+    {
+      "grad_norm": 0.6760733723640442,
+      "learning_rate": 9.522961219887092e-05,
+      "loss": 0.046,
+      "step": 1100
+    },
+    {
+      "grad_norm": 0.6449405550956726,
+      "learning_rate": 9.511145286918828e-05,
+      "loss": 0.0475,
+      "step": 1110
+    },
+    {
+      "grad_norm": 0.5605469346046448,
+      "learning_rate": 9.499192317611167e-05,
+      "loss": 0.0438,
+      "step": 1120
+    },
+    {
+      "grad_norm": 0.6307807564735413,
+      "learning_rate": 9.487102675062851e-05,
+      "loss": 0.0468,
+      "step": 1130
+    },
+    {
+      "grad_norm": 0.5671112537384033,
+      "learning_rate": 9.474876726524374e-05,
+      "loss": 0.0426,
+      "step": 1140
+    },
+    {
+      "grad_norm": 0.8449708223342896,
+      "learning_rate": 9.462514843386845e-05,
+      "loss": 0.0434,
+      "step": 1150
+    },
+    {
+      "grad_norm": 0.6328504085540771,
+      "learning_rate": 9.450017401170689e-05,
+      "loss": 0.0464,
+      "step": 1160
+    },
+    {
+      "grad_norm": 0.5694561004638672,
+      "learning_rate": 9.437384779514256e-05,
+      "loss": 0.0462,
+      "step": 1170
+    },
+    {
+      "grad_norm": 0.5980640053749084,
+      "learning_rate": 9.424617362162271e-05,
+      "loss": 0.0465,
+      "step": 1180
+    },
+    {
+      "grad_norm": 0.6236596703529358,
+      "learning_rate": 9.411715536954196e-05,
+      "loss": 0.0434,
+      "step": 1190
+    },
+    {
+      "grad_norm": 0.5503287315368652,
+      "learning_rate": 9.39867969581243e-05,
+      "loss": 0.0438,
+      "step": 1200
+    },
+    {
+      "grad_norm": 0.7289694547653198,
+      "learning_rate": 9.385510234730415e-05,
+      "loss": 0.0452,
+      "step": 1210
+    },
+    {
+      "grad_norm": 0.5549660325050354,
+      "learning_rate": 9.372207553760603e-05,
+      "loss": 0.0435,
+      "step": 1220
+    },
+    {
+      "grad_norm": 0.6124976277351379,
+      "learning_rate": 9.358772057002312e-05,
+      "loss": 0.0423,
+      "step": 1230
+    },
+    {
+      "grad_norm": 0.6705859899520874,
+      "learning_rate": 9.345204152589428e-05,
+      "loss": 0.0425,
+      "step": 1240
+    },
+    {
+      "grad_norm": 0.5378178954124451,
+      "learning_rate": 9.331504252678037e-05,
+      "loss": 0.045,
+      "step": 1250
+    },
+    {
+      "grad_norm": 0.729077935218811,
+      "learning_rate": 9.317672773433876e-05,
+      "loss": 0.0431,
+      "step": 1260
+    },
+    {
+      "grad_norm": 0.5443227887153625,
+      "learning_rate": 9.30371013501972e-05,
+      "loss": 0.0436,
+      "step": 1270
+    },
+    {
+      "grad_norm": 0.6833415031433105,
+      "learning_rate": 9.289616761582587e-05,
+      "loss": 0.0461,
+      "step": 1280
+    },
+    {
+      "grad_norm": 0.6458922028541565,
+      "learning_rate": 9.275393081240882e-05,
+      "loss": 0.0424,
+      "step": 1290
+    },
+    {
+      "grad_norm": 0.726543664932251,
+      "learning_rate": 9.261039526071374e-05,
+      "loss": 0.0447,
+      "step": 1300
+    },
+    {
+      "grad_norm": 0.5732786655426025,
+      "learning_rate": 9.246556532096078e-05,
+      "loss": 0.0431,
+      "step": 1310
+    },
+    {
+      "grad_norm": 0.7566348910331726,
+      "learning_rate": 9.231944539269009e-05,
+      "loss": 0.0436,
+      "step": 1320
+    },
+    {
+      "grad_norm": 0.772743284702301,
+      "learning_rate": 9.217203991462815e-05,
+      "loss": 0.0435,
+      "step": 1330
+    },
+    {
+      "grad_norm": 0.6129595637321472,
+      "learning_rate": 9.202335336455296e-05,
+      "loss": 0.0423,
+      "step": 1340
+    },
+    {
+      "grad_norm": 0.5188651084899902,
+      "learning_rate": 9.187339025915802e-05,
+      "loss": 0.0411,
+      "step": 1350
+    },
+    {
+      "grad_norm": 0.6343339085578918,
+      "learning_rate": 9.17221551539151e-05,
+      "loss": 0.0397,
+      "step": 1360
+    },
+    {
+      "grad_norm": 0.6620864868164062,
+      "learning_rate": 9.156965264293586e-05,
+      "loss": 0.0409,
+      "step": 1370
+    },
+    {
+      "grad_norm": 0.6305266618728638,
+      "learning_rate": 9.141588735883232e-05,
+      "loss": 0.0393,
+      "step": 1380
+    },
+    {
+      "grad_norm": 0.7356075644493103,
+      "learning_rate": 9.126086397257612e-05,
+      "loss": 0.0417,
+      "step": 1390
+    },
+    {
+      "grad_norm": 0.5255911350250244,
+      "learning_rate": 9.110458719335659e-05,
+      "loss": 0.0409,
+      "step": 1400
+    },
+    {
+      "grad_norm": 0.582068920135498,
+      "learning_rate": 9.094706176843777e-05,
+      "loss": 0.0424,
+      "step": 1410
+    },
+    {
+      "grad_norm": 0.48159053921699524,
+      "learning_rate": 9.078829248301417e-05,
+      "loss": 0.039,
+      "step": 1420
+    },
+    {
+      "grad_norm": 0.6647977232933044,
+      "learning_rate": 9.062828416006539e-05,
+      "loss": 0.0424,
+      "step": 1430
+    },
+    {
+      "grad_norm": 0.6569739580154419,
+      "learning_rate": 9.046704166020961e-05,
+      "loss": 0.0405,
+      "step": 1440
+    },
+    {
+      "grad_norm": 0.5740228891372681,
+      "learning_rate": 9.030456988155596e-05,
+      "loss": 0.0403,
+      "step": 1450
+    },
+    {
+      "grad_norm": 0.5312358140945435,
+      "learning_rate": 9.014087375955573e-05,
+      "loss": 0.04,
+      "step": 1460
+    },
+    {
+      "grad_norm": 0.6075960993766785,
+      "learning_rate": 8.997595826685243e-05,
+      "loss": 0.0447,
+      "step": 1470
+    },
+    {
+      "grad_norm": 0.4863235056400299,
+      "learning_rate": 8.980982841313074e-05,
+      "loss": 0.04,
+      "step": 1480
+    },
+    {
+      "grad_norm": 0.5700269937515259,
+      "learning_rate": 8.964248924496435e-05,
+      "loss": 0.0381,
+      "step": 1490
+    },
+    {
+      "grad_norm": 0.6507738828659058,
+      "learning_rate": 8.947394584566258e-05,
+      "loss": 0.0389,
+      "step": 1500
+    },
+    {
+      "grad_norm": 0.5595966577529907,
+      "learning_rate": 8.930420333511606e-05,
+      "loss": 0.0391,
+      "step": 1510
+    },
+    {
+      "grad_norm": 0.6669788360595703,
+      "learning_rate": 8.913326686964117e-05,
+      "loss": 0.0389,
+      "step": 1520
+    },
+    {
+      "grad_norm": 0.6669943332672119,
+      "learning_rate": 8.89611416418234e-05,
+      "loss": 0.0383,
+      "step": 1530
+    },
+    {
+      "grad_norm": 0.6302319765090942,
+      "learning_rate": 8.878783288035957e-05,
+      "loss": 0.0387,
+      "step": 1540
+    },
+    {
+      "grad_norm": 0.5663589239120483,
+      "learning_rate": 8.86133458498991e-05,
+      "loss": 0.0409,
+      "step": 1550
+    },
+    {
+      "grad_norm": 0.6851902008056641,
+      "learning_rate": 8.843768585088393e-05,
+      "loss": 0.039,
+      "step": 1560
+    },
+    {
+      "grad_norm": 0.4790905714035034,
+      "learning_rate": 8.82608582193877e-05,
+      "loss": 0.0397,
+      "step": 1570
+    },
+    {
+      "grad_norm": 0.5178065896034241,
+      "learning_rate": 8.80828683269535e-05,
+      "loss": 0.0378,
+      "step": 1580
+    },
+    {
+      "grad_norm": 0.5918937921524048,
+      "learning_rate": 8.790372158043074e-05,
+      "loss": 0.0372,
+      "step": 1590
+    },
+    {
+      "grad_norm": 0.5967026948928833,
+      "learning_rate": 8.772342342181095e-05,
+      "loss": 0.0382,
+      "step": 1600
+    },
+    {
+      "grad_norm": 0.6606245636940002,
+      "learning_rate": 8.75419793280624e-05,
+      "loss": 0.0375,
+      "step": 1610
+    },
+    {
+      "grad_norm": 0.6581755876541138,
+      "learning_rate": 8.735939481096378e-05,
+      "loss": 0.0376,
+      "step": 1620
+    },
+    {
+      "grad_norm": 0.6129235029220581,
+      "learning_rate": 8.717567541693673e-05,
+      "loss": 0.0394,
+      "step": 1630
+    },
+    {
+      "grad_norm": 0.6601637601852417,
+      "learning_rate": 8.699082672687734e-05,
+      "loss": 0.0389,
+      "step": 1640
+    },
+    {
+      "grad_norm": 0.5138643980026245,
+      "learning_rate": 8.680485435598673e-05,
+      "loss": 0.0423,
+      "step": 1650
+    },
+    {
+      "grad_norm": 0.48632222414016724,
+      "learning_rate": 8.661776395360029e-05,
+      "loss": 0.0391,
+      "step": 1660
+    },
+    {
+      "grad_norm": 0.7064632773399353,
+      "learning_rate": 8.642956120301626e-05,
+      "loss": 0.0372,
+      "step": 1670
+    },
+    {
+      "grad_norm": 0.599163293838501,
+      "learning_rate": 8.624025182132292e-05,
+      "loss": 0.0387,
+      "step": 1680
+    },
+    {
+      "grad_norm": 0.49687978625297546,
+      "learning_rate": 8.604984155922506e-05,
+      "loss": 0.0384,
+      "step": 1690
+    },
+    {
+      "grad_norm": 0.5044621229171753,
+      "learning_rate": 8.585833620086918e-05,
+      "loss": 0.037,
+      "step": 1700
+    },
+    {
+      "grad_norm": 0.48257848620414734,
+      "learning_rate": 8.566574156366784e-05,
+      "loss": 0.0355,
+      "step": 1710
+    },
+    {
+      "grad_norm": 0.7500448822975159,
+      "learning_rate": 8.547206349812298e-05,
+      "loss": 0.0396,
+      "step": 1720
+    },
+    {
+      "grad_norm": 0.5865467190742493,
+      "learning_rate": 8.527730788764805e-05,
+      "loss": 0.0392,
+      "step": 1730
+    },
+    {
+      "grad_norm": 0.5636748671531677,
+      "learning_rate": 8.508148064838948e-05,
+      "loss": 0.0362,
+      "step": 1740
+    },
+    {
+      "grad_norm": 0.6705951690673828,
+      "learning_rate": 8.488458772904684e-05,
+      "loss": 0.0366,
+      "step": 1750
+    },
+    {
+      "grad_norm": 0.5554336309432983,
+      "learning_rate": 8.468663511069217e-05,
+      "loss": 0.0363,
+      "step": 1760
+    },
+    {
+      "grad_norm": 0.45949798822402954,
+      "learning_rate": 8.448762880658825e-05,
+      "loss": 0.0383,
+      "step": 1770
+    },
+    {
+      "grad_norm": 0.5537176132202148,
+      "learning_rate": 8.428757486200603e-05,
+      "loss": 0.0367,
+      "step": 1780
+    },
+    {
+      "grad_norm": 0.4168303608894348,
+      "learning_rate": 8.40864793540409e-05,
+      "loss": 0.0396,
+      "step": 1790
+    },
+    {
+      "grad_norm": 0.6133352518081665,
+      "learning_rate": 8.388434839142813e-05,
+      "loss": 0.0378,
+      "step": 1800
+    },
+    {
+      "grad_norm": 0.484385222196579,
+      "learning_rate": 8.368118811435726e-05,
+      "loss": 0.035,
+      "step": 1810
+    },
+    {
+      "grad_norm": 0.4967355728149414,
+      "learning_rate": 8.347700469428564e-05,
+      "loss": 0.0351,
+      "step": 1820
+    },
+    {
+      "grad_norm": 0.5309926867485046,
+      "learning_rate": 8.327180433375091e-05,
+      "loss": 0.0364,
+      "step": 1830
+    },
+    {
+      "grad_norm": 0.5246663093566895,
+      "learning_rate": 8.306559326618259e-05,
+      "loss": 0.0358,
+      "step": 1840
+    },
+    {
+      "grad_norm": 0.5756788849830627,
+      "learning_rate": 8.285837775571276e-05,
+      "loss": 0.0357,
+      "step": 1850
+    },
+    {
+      "grad_norm": 0.5469930768013,
+      "learning_rate": 8.265016409698573e-05,
+      "loss": 0.0365,
+      "step": 1860
+    },
+    {
+      "grad_norm": 0.5380656123161316,
+      "learning_rate": 8.244095861496686e-05,
+      "loss": 0.036,
+      "step": 1870
+    },
+    {
+      "grad_norm": 0.45056721568107605,
+      "learning_rate": 8.223076766475035e-05,
+      "loss": 0.0358,
+      "step": 1880
+    },
+    {
+      "grad_norm": 0.7063160538673401,
+      "learning_rate": 8.201959763136633e-05,
+      "loss": 0.0344,
+      "step": 1890
+    },
+    {
+      "grad_norm": 0.5377547144889832,
+      "learning_rate": 8.180745492958674e-05,
+      "loss": 0.0364,
+      "step": 1900
+    },
+    {
+      "grad_norm": 0.4956304132938385,
+      "learning_rate": 8.159434600373061e-05,
+      "loss": 0.0335,
+      "step": 1910
+    },
+    {
+      "grad_norm": 0.650954008102417,
+      "learning_rate": 8.138027732746818e-05,
+      "loss": 0.0357,
+      "step": 1920
+    },
+    {
+      "grad_norm": 0.43918663263320923,
+      "learning_rate": 8.116525540362434e-05,
+      "loss": 0.0341,
+      "step": 1930
+    },
+    {
+      "grad_norm": 0.4600161612033844,
+      "learning_rate": 8.094928676398101e-05,
+      "loss": 0.0348,
+      "step": 1940
+    },
+    {
+      "grad_norm": 0.5328519344329834,
+      "learning_rate": 8.073237796907882e-05,
+      "loss": 0.034,
+      "step": 1950
+    },
+    {
+      "grad_norm": 0.6386666893959045,
+      "learning_rate": 8.051453560801772e-05,
+      "loss": 0.0384,
+      "step": 1960
+    },
+    {
+      "grad_norm": 0.5290015935897827,
+      "learning_rate": 8.029576629825687e-05,
+      "loss": 0.034,
+      "step": 1970
+    },
+    {
+      "grad_norm": 0.511027455329895,
+      "learning_rate": 8.007607668541362e-05,
+      "loss": 0.0344,
+      "step": 1980
+    },
+    {
+      "grad_norm": 0.5814989805221558,
+      "learning_rate": 7.985547344306161e-05,
+      "loss": 0.0379,
+      "step": 1990
+    },
+    {
+      "grad_norm": 0.48753857612609863,
+      "learning_rate": 7.963396327252812e-05,
+      "loss": 0.0367,
+      "step": 2000
+    },
+    {
+      "grad_norm": 0.511787474155426,
+      "learning_rate": 7.941155290269038e-05,
+      "loss": 0.0353,
+      "step": 2010
+    },
+    {
+      "grad_norm": 0.5462181568145752,
+      "learning_rate": 7.918824908977123e-05,
+      "loss": 0.0335,
+      "step": 2020
+    },
+    {
+      "grad_norm": 0.42617759108543396,
+      "learning_rate": 7.896405861713394e-05,
+      "loss": 0.034,
+      "step": 2030
+    },
+    {
+      "grad_norm": 0.4321683645248413,
+      "learning_rate": 7.873898829507606e-05,
+      "loss": 0.0337,
+      "step": 2040
+    },
+    {
+      "grad_norm": 0.48458221554756165,
+      "learning_rate": 7.851304496062254e-05,
+      "loss": 0.0336,
+      "step": 2050
+    },
+    {
+      "grad_norm": 0.4516434967517853,
+      "learning_rate": 7.828623547731818e-05,
+      "loss": 0.0334,
+      "step": 2060
+    },
+    {
+      "grad_norm": 0.5588269233703613,
+      "learning_rate": 7.80585667350189e-05,
+      "loss": 0.0343,
+      "step": 2070
+    },
+    {
+      "grad_norm": 0.5814851522445679,
+      "learning_rate": 7.783004564968263e-05,
+      "loss": 0.0343,
+      "step": 2080
+    },
+    {
+      "grad_norm": 0.548805832862854,
+      "learning_rate": 7.760067916315921e-05,
+      "loss": 0.0333,
+      "step": 2090
+    },
+    {
+      "grad_norm": 0.47741568088531494,
+      "learning_rate": 7.737047424297941e-05,
+      "loss": 0.0335,
+      "step": 2100
+    },
+    {
+      "grad_norm": 0.5174888968467712,
+      "learning_rate": 7.713943788214337e-05,
+      "loss": 0.0333,
+      "step": 2110
+    },
+    {
+      "grad_norm": 0.5185810923576355,
+      "learning_rate": 7.690757709890812e-05,
+      "loss": 0.0336,
+      "step": 2120
+    },
+    {
+      "grad_norm": 0.5128524303436279,
+      "learning_rate": 7.66748989365744e-05,
+      "loss": 0.0334,
+      "step": 2130
+    },
+    {
+      "grad_norm": 0.39958009123802185,
+      "learning_rate": 7.644141046327271e-05,
+      "loss": 0.0312,
+      "step": 2140
+    },
+    {
+      "grad_norm": 0.49459007382392883,
+      "learning_rate": 7.620711877174866e-05,
+      "loss": 0.033,
+      "step": 2150
+    },
+    {
+      "grad_norm": 0.5272864699363708,
+      "learning_rate": 7.597203097914732e-05,
+      "loss": 0.0342,
+      "step": 2160
+    },
+    {
+      "grad_norm": 0.5177196860313416,
+      "learning_rate": 7.573615422679726e-05,
+      "loss": 0.0352,
+      "step": 2170
+    },
+    {
+      "grad_norm": 0.4378889203071594,
+      "learning_rate": 7.549949567999345e-05,
+      "loss": 0.0313,
+      "step": 2180
+    },
+    {
+      "grad_norm": 0.47708237171173096,
+      "learning_rate": 7.526206252777968e-05,
+      "loss": 0.0346,
+      "step": 2190
+    },
+    {
+      "grad_norm": 0.6027482151985168,
+      "learning_rate": 7.50238619827301e-05,
+      "loss": 0.0337,
+      "step": 2200
+    },
+    {
+      "grad_norm": 0.4342920780181885,
+      "learning_rate": 7.478490128073022e-05,
+      "loss": 0.0335,
+      "step": 2210
+    },
+    {
+      "grad_norm": 0.531221866607666,
+      "learning_rate": 7.454518768075704e-05,
+      "loss": 0.0334,
+      "step": 2220
+    },
+    {
+      "grad_norm": 0.42581605911254883,
+      "learning_rate": 7.430472846465856e-05,
+      "loss": 0.0322,
+      "step": 2230
+    },
+    {
+      "grad_norm": 0.42176100611686707,
+      "learning_rate": 7.406353093693253e-05,
+      "loss": 0.0325,
+      "step": 2240
+    },
+    {
+      "grad_norm": 0.5013089179992676,
+      "learning_rate": 7.382160242450469e-05,
+      "loss": 0.033,
+      "step": 2250
+    },
+    {
+      "grad_norm": 0.4330039918422699,
+      "learning_rate": 7.357895027650598e-05,
+      "loss": 0.0312,
+      "step": 2260
+    },
+    {
+      "grad_norm": 0.4245281219482422,
+      "learning_rate": 7.333558186404958e-05,
+      "loss": 0.0316,
+      "step": 2270
+    },
+    {
+      "grad_norm": 0.5211013555526733,
+      "learning_rate": 7.309150458000668e-05,
+      "loss": 0.0325,
+      "step": 2280
+    },
+    {
+      "grad_norm": 0.6483867168426514,
+      "learning_rate": 7.284672583878219e-05,
+      "loss": 0.0326,
+      "step": 2290
+    },
+    {
+      "grad_norm": 0.4503920376300812,
+      "learning_rate": 7.260125307608929e-05,
+      "loss": 0.0337,
+      "step": 2300
+    },
+    {
+      "grad_norm": 0.40689945220947266,
+      "learning_rate": 7.235509374872373e-05,
+      "loss": 0.033,
+      "step": 2310
+    },
+    {
+      "grad_norm": 0.5707908272743225,
+      "learning_rate": 7.210825533433719e-05,
+      "loss": 0.0322,
+      "step": 2320
+    },
+    {
+      "grad_norm": 0.4584868252277374,
+      "learning_rate": 7.186074533121013e-05,
+      "loss": 0.0328,
+      "step": 2330
+    },
+    {
+      "grad_norm": 0.4298429489135742,
+      "learning_rate": 7.161257125802413e-05,
+      "loss": 0.0319,
+      "step": 2340
+    },
+    {
+      "grad_norm": 0.5181540846824646,
+      "learning_rate": 7.136374065363334e-05,
+      "loss": 0.0337,
+      "step": 2350
+    },
+    {
+      "grad_norm": 0.5237898826599121,
+      "learning_rate": 7.11142610768356e-05,
+      "loss": 0.0322,
+      "step": 2360
+    },
+    {
+      "grad_norm": 0.478541761636734,
+      "learning_rate": 7.086414010614276e-05,
+      "loss": 0.0314,
+      "step": 2370
+    },
+    {
+      "grad_norm": 0.48820534348487854,
+      "learning_rate": 7.061338533955043e-05,
+      "loss": 0.0337,
+      "step": 2380
+    },
+    {
+      "grad_norm": 0.48146507143974304,
+      "learning_rate": 7.036200439430725e-05,
+      "loss": 0.0334,
+      "step": 2390
+    },
+    {
+      "grad_norm": 0.5339111685752869,
+      "learning_rate": 7.01100049066835e-05,
+      "loss": 0.0318,
+      "step": 2400
+    },
+    {
+      "grad_norm": 0.48041751980781555,
+      "learning_rate": 6.985739453173903e-05,
+      "loss": 0.0331,
+      "step": 2410
+    },
+    {
+      "grad_norm": 0.5311410427093506,
+      "learning_rate": 6.960418094309085e-05,
+      "loss": 0.032,
+      "step": 2420
+    },
+    {
+      "grad_norm": 0.3880246877670288,
+      "learning_rate": 6.93503718326799e-05,
+      "loss": 0.0315,
+      "step": 2430
+    },
+    {
+      "grad_norm": 0.44451695680618286,
+      "learning_rate": 6.909597491053751e-05,
+      "loss": 0.0302,
+      "step": 2440
+    },
+    {
+      "grad_norm": 0.5587266683578491,
+      "learning_rate": 6.884099790455113e-05,
+      "loss": 0.0306,
+      "step": 2450
+    },
+    {
+      "grad_norm": 0.5494639873504639,
+      "learning_rate": 6.858544856022952e-05,
+      "loss": 0.0327,
+      "step": 2460
+    },
+    {
+      "grad_norm": 0.42823129892349243,
+      "learning_rate": 6.83293346404676e-05,
+      "loss": 0.0327,
+      "step": 2470
+    },
+    {
+      "grad_norm": 0.4961199462413788,
+      "learning_rate": 6.80726639253105e-05,
+      "loss": 0.0343,
+      "step": 2480
+    },
+    {
+      "grad_norm": 0.5614944696426392,
+      "learning_rate": 6.781544421171732e-05,
+      "loss": 0.0327,
+      "step": 2490
+    },
+    {
+      "grad_norm": 0.45186933875083923,
+      "learning_rate": 6.755768331332424e-05,
+      "loss": 0.0335,
+      "step": 2500
+    },
+    {
+      "grad_norm": 0.4534369111061096,
+      "learning_rate": 6.729938906020713e-05,
+      "loss": 0.0327,
+      "step": 2510
+    },
+    {
+      "grad_norm": 0.3477000892162323,
+      "learning_rate": 6.704056929864376e-05,
+      "loss": 0.0306,
+      "step": 2520
+    },
+    {
+      "grad_norm": 0.43540525436401367,
+      "learning_rate": 6.67812318908754e-05,
+      "loss": 0.0298,
+      "step": 2530
+    },
+    {
+      "grad_norm": 0.3721942901611328,
+      "learning_rate": 6.6521384714868e-05,
+      "loss": 0.029,
+      "step": 2540
+    },
+    {
+      "grad_norm": 0.5104391574859619,
+      "learning_rate": 6.626103566407295e-05,
+      "loss": 0.0303,
+      "step": 2550
+    },
+    {
+      "grad_norm": 0.38681259751319885,
+      "learning_rate": 6.600019264718713e-05,
+      "loss": 0.0301,
+      "step": 2560
+    },
+    {
+      "grad_norm": 0.4433761239051819,
+      "learning_rate": 6.573886358791285e-05,
+      "loss": 0.0319,
+      "step": 2570
+    },
+    {
+      "grad_norm": 0.413533478975296,
+      "learning_rate": 6.547705642471703e-05,
+      "loss": 0.0309,
+      "step": 2580
+    },
+    {
+      "grad_norm": 0.3686458170413971,
+      "learning_rate": 6.521477911059008e-05,
+      "loss": 0.0282,
+      "step": 2590
+    },
+    {
+      "grad_norm": 0.38820889592170715,
+      "learning_rate": 6.495203961280434e-05,
+      "loss": 0.0301,
+      "step": 2600
+    },
+    {
+      "grad_norm": 0.4620518982410431,
+      "learning_rate": 6.468884591267204e-05,
+      "loss": 0.0302,
+      "step": 2610
+    },
+    {
+      "grad_norm": 0.5960747599601746,
+      "learning_rate": 6.44252060053028e-05,
+      "loss": 0.03,
+      "step": 2620
+    },
+    {
+      "grad_norm": 0.6030673980712891,
+      "learning_rate": 6.416112789936086e-05,
+      "loss": 0.0312,
+      "step": 2630
+    },
+    {
+      "grad_norm": 0.5022603273391724,
+      "learning_rate": 6.389661961682173e-05,
+      "loss": 0.0302,
+      "step": 2640
+    },
+    {
+      "grad_norm": 0.4202682375907898,
+      "learning_rate": 6.363168919272846e-05,
+      "loss": 0.0295,
+      "step": 2650
+    },
+    {
+      "grad_norm": 0.40000995993614197,
+      "learning_rate": 6.336634467494768e-05,
+      "loss": 0.0286,
+      "step": 2660
+    },
+    {
+      "grad_norm": 0.49214911460876465,
+      "learning_rate": 6.310059412392505e-05,
+      "loss": 0.0308,
+      "step": 2670
+    },
+    {
+      "grad_norm": 0.4698426127433777,
+      "learning_rate": 6.283444561244042e-05,
+      "loss": 0.0291,
+      "step": 2680
+    },
+    {
+      "grad_norm": 0.48712682723999023,
+      "learning_rate": 6.256790722536251e-05,
+      "loss": 0.0288,
+      "step": 2690
+    },
+    {
+      "grad_norm": 0.46532660722732544,
+      "learning_rate": 6.230098705940354e-05,
+      "loss": 0.0292,
+      "step": 2700
+    },
+    {
+      "grad_norm": 0.422234445810318,
+      "learning_rate": 6.203369322287306e-05,
+      "loss": 0.0295,
+      "step": 2710
+    },
+    {
+      "grad_norm": 0.3877371549606323,
+      "learning_rate": 6.17660338354317e-05,
+      "loss": 0.0281,
+      "step": 2720
+    },
+    {
+      "grad_norm": 0.4343086779117584,
+      "learning_rate": 6.149801702784456e-05,
+      "loss": 0.0279,
+      "step": 2730
+    },
+    {
+      "grad_norm": 0.4151124954223633,
+      "learning_rate": 6.122965094173424e-05,
+      "loss": 0.0288,
+      "step": 2740
+    },
+    {
+      "grad_norm": 0.5097172260284424,
+      "learning_rate": 6.0960943729333374e-05,
+      "loss": 0.0306,
+      "step": 2750
+    },
+    {
+      "grad_norm": 0.5084396600723267,
+      "learning_rate": 6.069190355323717e-05,
+      "loss": 0.0284,
+      "step": 2760
+    },
+    {
+      "grad_norm": 0.46339768171310425,
+      "learning_rate": 6.042253858615532e-05,
+      "loss": 0.0284,
+      "step": 2770
+    },
+    {
+      "grad_norm": 0.4735918343067169,
+      "learning_rate": 6.015285701066382e-05,
+      "loss": 0.0304,
+      "step": 2780
+    },
+    {
+      "grad_norm": 0.4199438691139221,
+      "learning_rate": 5.988286701895631e-05,
+      "loss": 0.0321,
+      "step": 2790
+    },
+    {
+      "grad_norm": 0.5070331692695618,
+      "learning_rate": 5.961257681259535e-05,
+      "loss": 0.0313,
+      "step": 2800
+    },
+    {
+      "grad_norm": 0.419735848903656,
+      "learning_rate": 5.934199460226317e-05,
+      "loss": 0.0306,
+      "step": 2810
+    },
+    {
+      "grad_norm": 0.4207720160484314,
+      "learning_rate": 5.9071128607512285e-05,
+      "loss": 0.0295,
+      "step": 2820
+    },
+    {
+      "grad_norm": 0.3231958746910095,
+      "learning_rate": 5.8799987056515804e-05,
+      "loss": 0.0288,
+      "step": 2830
+    },
+    {
+      "grad_norm": 0.31990593671798706,
+      "learning_rate": 5.8528578185817514e-05,
+      "loss": 0.0296,
+      "step": 2840
+    },
+    {
+      "grad_norm": 0.41715797781944275,
+      "learning_rate": 5.825691024008162e-05,
+      "loss": 0.0287,
+      "step": 2850
+    },
+    {
+      "grad_norm": 0.3970031142234802,
+      "learning_rate": 5.798499147184233e-05,
+      "loss": 0.0286,
+      "step": 2860
+    },
+    {
+      "grad_norm": 0.3385803699493408,
+      "learning_rate": 5.771283014125317e-05,
+      "loss": 0.0309,
+      "step": 2870
+    },
+    {
+      "grad_norm": 0.42199286818504333,
+      "learning_rate": 5.7440434515836064e-05,
+      "loss": 0.0283,
+      "step": 2880
+    },
+    {
+      "grad_norm": 0.4184113144874573,
+      "learning_rate": 5.7167812870230094e-05,
+      "loss": 0.0281,
+      "step": 2890
+    },
+    {
+      "grad_norm": 0.4537411630153656,
+      "learning_rate": 5.689497348594035e-05,
+      "loss": 0.0282,
+      "step": 2900
+    },
+    {
+      "grad_norm": 0.3758944869041443,
+      "learning_rate": 5.662192465108613e-05,
+      "loss": 0.0291,
+      "step": 2910
+    },
+    {
+      "grad_norm": 0.43658238649368286,
+      "learning_rate": 5.634867466014932e-05,
+      "loss": 0.0276,
+      "step": 2920
+    },
+    {
+      "grad_norm": 0.435166597366333,
+      "learning_rate": 5.607523181372234e-05,
+      "loss": 0.029,
+      "step": 2930
+    },
+    {
+      "grad_norm": 0.39249086380004883,
+      "learning_rate": 5.5801604418256117e-05,
+      "loss": 0.0281,
+      "step": 2940
+    },
+    {
+      "grad_norm": 0.38982075452804565,
+      "learning_rate": 5.552780078580756e-05,
+      "loss": 0.0267,
+      "step": 2950
+    },
+    {
+      "grad_norm": 0.397555410861969,
+      "learning_rate": 5.525382923378728e-05,
+      "loss": 0.028,
+      "step": 2960
+    },
+    {
+      "grad_norm": 0.4088326692581177,
+      "learning_rate": 5.49796980847068e-05,
+      "loss": 0.0299,
+      "step": 2970
+    },
+    {
+      "grad_norm": 0.45356932282447815,
+      "learning_rate": 5.470541566592573e-05,
+      "loss": 0.0296,
+      "step": 2980
+    },
+    {
+      "grad_norm": 0.3560492694377899,
+      "learning_rate": 5.443099030939887e-05,
+      "loss": 0.0278,
+      "step": 2990
+    },
+    {
+      "grad_norm": 0.35022470355033875,
+      "learning_rate": 5.415643035142309e-05,
+      "loss": 0.0275,
+      "step": 3000
+    },
+    {
+      "grad_norm": 0.40406009554862976,
+      "learning_rate": 5.3881744132384104e-05,
+      "loss": 0.0286,
+      "step": 3010
+    },
+    {
+      "grad_norm": 0.3887314796447754,
+      "learning_rate": 5.360693999650303e-05,
+      "loss": 0.0278,
+      "step": 3020
+    },
+    {
+      "grad_norm": 0.44561782479286194,
+      "learning_rate": 5.3332026291583016e-05,
+      "loss": 0.0274,
+      "step": 3030
+    },
+    {
+      "grad_norm": 0.4164481461048126,
+      "learning_rate": 5.305701136875566e-05,
+      "loss": 0.0292,
+      "step": 3040
+    },
+    {
+      "grad_norm": 0.3928636610507965,
+      "learning_rate": 5.278190358222721e-05,
+      "loss": 0.0281,
+      "step": 3050
+    },
+    {
+      "grad_norm": 0.4332394003868103,
+      "learning_rate": 5.25067112890249e-05,
+      "loss": 0.0286,
+      "step": 3060
+    },
+    {
+      "grad_norm": 0.33109888434410095,
+      "learning_rate": 5.2231442848743064e-05,
+      "loss": 0.0284,
+      "step": 3070
+    },
+    {
+      "grad_norm": 0.44803181290626526,
+      "learning_rate": 5.1956106623289145e-05,
+      "loss": 0.0261,
+      "step": 3080
+    },
+    {
+      "grad_norm": 0.39142924547195435,
+      "learning_rate": 5.168071097662972e-05,
+      "loss": 0.0261,
+      "step": 3090
+    },
+    {
+      "grad_norm": 0.532752513885498,
+      "learning_rate": 5.1405264274536445e-05,
+      "loss": 0.0278,
+      "step": 3100
+    },
+    {
+      "grad_norm": 0.4373730421066284,
+      "learning_rate": 5.112977488433188e-05,
+      "loss": 0.0281,
+      "step": 3110
+    },
+    {
+      "grad_norm": 0.395621120929718,
+      "learning_rate": 5.085425117463533e-05,
+      "loss": 0.027,
+      "step": 3120
+    },
+    {
+      "grad_norm": 0.43923935294151306,
+      "learning_rate": 5.057870151510864e-05,
+      "loss": 0.0264,
+      "step": 3130
+    },
+    {
+      "grad_norm": 0.37473028898239136,
+      "learning_rate": 5.030313427620197e-05,
+      "loss": 0.0261,
+      "step": 3140
+    },
+    {
+      "grad_norm": 0.44424837827682495,
+      "learning_rate": 5.0027557828899426e-05,
+      "loss": 0.0269,
+      "step": 3150
+    },
+    {
+      "grad_norm": 0.40588265657424927,
+      "learning_rate": 4.975198054446492e-05,
+      "loss": 0.0262,
+      "step": 3160
+    },
+    {
+      "grad_norm": 0.4514497220516205,
+      "learning_rate": 4.947641079418773e-05,
+      "loss": 0.0267,
+      "step": 3170
+    },
+    {
+      "grad_norm": 0.5240787267684937,
+      "learning_rate": 4.920085694912828e-05,
+      "loss": 0.0287,
+      "step": 3180
+    },
+    {
+      "grad_norm": 0.37988799810409546,
+      "learning_rate": 4.892532737986387e-05,
+      "loss": 0.0265,
+      "step": 3190
+    },
+    {
+      "grad_norm": 0.376968115568161,
+      "learning_rate": 4.864983045623434e-05,
+      "loss": 0.0292,
+      "step": 3200
+    },
+    {
+      "grad_norm": 0.3950600326061249,
+      "learning_rate": 4.837437454708784e-05,
+      "loss": 0.0279,
+      "step": 3210
+    },
+    {
+      "grad_norm": 0.37644240260124207,
+      "learning_rate": 4.809896802002662e-05,
+      "loss": 0.0259,
+      "step": 3220
+    },
+    {
+      "grad_norm": 0.39114123582839966,
+      "learning_rate": 4.7823619241152854e-05,
+      "loss": 0.0256,
+      "step": 3230
+    },
+    {
+      "grad_norm": 0.2755090594291687,
+      "learning_rate": 4.754833657481445e-05,
+      "loss": 0.0248,
+      "step": 3240
+    },
+    {
+      "grad_norm": 0.3851339519023895,
+      "learning_rate": 4.7273128383351015e-05,
+      "loss": 0.0274,
+      "step": 3250
+    },
+    {
+      "grad_norm": 0.4715338349342346,
+      "learning_rate": 4.699800302683981e-05,
+      "loss": 0.0273,
+      "step": 3260
+    },
+    {
+      "grad_norm": 0.30057212710380554,
+      "learning_rate": 4.6722968862841806e-05,
+      "loss": 0.0251,
+      "step": 3270
+    },
+    {
+      "grad_norm": 0.3959311544895172,
+      "learning_rate": 4.6448034246147754e-05,
+      "loss": 0.0267,
+      "step": 3280
+    },
+    {
+      "grad_norm": 0.35625332593917847,
+      "learning_rate": 4.6173207528524476e-05,
+      "loss": 0.0255,
+      "step": 3290
+    },
+    {
+      "grad_norm": 0.35969534516334534,
+      "learning_rate": 4.58984970584611e-05,
+      "loss": 0.0263,
+      "step": 3300
+    },
+    {
+      "grad_norm": 0.5300602316856384,
+      "learning_rate": 4.562391118091544e-05,
+      "loss": 0.0279,
+      "step": 3310
+    },
+    {
+      "grad_norm": 0.5815778374671936,
+      "learning_rate": 4.534945823706056e-05,
+      "loss": 0.0291,
+      "step": 3320
+    },
+    {
+      "grad_norm": 0.41548240184783936,
+      "learning_rate": 4.507514656403137e-05,
+      "loss": 0.0273,
+      "step": 3330
+    },
+    {
+      "grad_norm": 0.38885802030563354,
+      "learning_rate": 4.480098449467132e-05,
+      "loss": 0.0265,
+      "step": 3340
+    },
+    {
+      "grad_norm": 0.3587275445461273,
+      "learning_rate": 4.452698035727929e-05,
+      "loss": 0.0255,
+      "step": 3350
+    },
+    {
+      "grad_norm": 0.42804276943206787,
+      "learning_rate": 4.425314247535668e-05,
+      "loss": 0.0272,
+      "step": 3360
+    },
+    {
+      "grad_norm": 0.37650251388549805,
+      "learning_rate": 4.3979479167354477e-05,
+      "loss": 0.0263,
+      "step": 3370
+    },
+    {
+      "grad_norm": 0.3508603274822235,
+      "learning_rate": 4.370599874642055e-05,
+      "loss": 0.0288,
+      "step": 3380
+    },
+    {
+      "grad_norm": 0.3932643532752991,
+      "learning_rate": 4.3432709520147205e-05,
+      "loss": 0.0271,
+      "step": 3390
+    },
+    {
+      "grad_norm": 0.3967253863811493,
+      "learning_rate": 4.315961979031875e-05,
+      "loss": 0.0262,
+      "step": 3400
+    },
+    {
+      "grad_norm": 0.34985652565956116,
+      "learning_rate": 4.2886737852659325e-05,
+      "loss": 0.0266,
+      "step": 3410
+    },
+    {
+      "grad_norm": 0.4587637484073639,
+      "learning_rate": 4.261407199658093e-05,
+      "loss": 0.0261,
+      "step": 3420
+    },
+    {
+      "grad_norm": 0.40246912837028503,
+      "learning_rate": 4.234163050493158e-05,
+      "loss": 0.0248,
+      "step": 3430
+    },
+    {
+      "grad_norm": 0.4552555978298187,
+      "learning_rate": 4.2069421653743706e-05,
+      "loss": 0.0264,
+      "step": 3440
+    },
+    {
+      "grad_norm": 0.31220653653144836,
+      "learning_rate": 4.179745371198276e-05,
+      "loss": 0.0253,
+      "step": 3450
+    },
+    {
+      "grad_norm": 0.38440898060798645,
+      "learning_rate": 4.1525734941296026e-05,
+      "loss": 0.0263,
+      "step": 3460
+    },
+    {
+      "grad_norm": 0.3631040155887604,
+      "learning_rate": 4.125427359576162e-05,
+      "loss": 0.0254,
+      "step": 3470
+    },
+    {
+      "grad_norm": 0.41102805733680725,
+      "learning_rate": 4.0983077921637815e-05,
+      "loss": 0.0253,
+      "step": 3480
+    },
+    {
+      "grad_norm": 0.38240450620651245,
+      "learning_rate": 4.07121561571125e-05,
+      "loss": 0.0255,
+      "step": 3490
+    },
+    {
+      "grad_norm": 0.3399236798286438,
+      "learning_rate": 4.044151653205292e-05,
+      "loss": 0.0238,
+      "step": 3500
+    },
+    {
+      "grad_norm": 0.37793099880218506,
+      "learning_rate": 4.0171167267755696e-05,
+      "loss": 0.0248,
+      "step": 3510
+    },
+    {
+      "grad_norm": 0.47253522276878357,
+      "learning_rate": 3.9901116576697083e-05,
+      "loss": 0.0243,
+      "step": 3520
+    },
+    {
+      "grad_norm": 0.42441272735595703,
+      "learning_rate": 3.963137266228349e-05,
+      "loss": 0.0239,
+      "step": 3530
+    },
+    {
+      "grad_norm": 0.378666490316391,
+      "learning_rate": 3.93619437186023e-05,
+      "loss": 0.0265,
+      "step": 3540
+    },
+    {
+      "grad_norm": 0.41540664434432983,
+      "learning_rate": 3.9092837930172884e-05,
+      "loss": 0.0253,
+      "step": 3550
+    },
+    {
+      "grad_norm": 0.3746972680091858,
+      "learning_rate": 3.8824063471698105e-05,
+      "loss": 0.0251,
+      "step": 3560
+    },
+    {
+      "grad_norm": 0.3388042151927948,
+      "learning_rate": 3.855562850781589e-05,
+      "loss": 0.0249,
+      "step": 3570
+    },
+    {
+      "grad_norm": 0.32412514090538025,
+      "learning_rate": 3.828754119285123e-05,
+      "loss": 0.0247,
+      "step": 3580
+    },
+    {
+      "grad_norm": 0.32984834909439087,
+      "learning_rate": 3.801980967056851e-05,
+      "loss": 0.0257,
+      "step": 3590
+    },
+    {
+      "grad_norm": 0.3018534481525421,
+      "learning_rate": 3.77524420739241e-05,
+      "loss": 0.0254,
+      "step": 3600
+    },
+    {
+      "grad_norm": 0.34701016545295715,
+      "learning_rate": 3.748544652481927e-05,
+      "loss": 0.024,
+      "step": 3610
+    },
+    {
+      "grad_norm": 0.4032657742500305,
+      "learning_rate": 3.721883113385353e-05,
+      "loss": 0.0268,
+      "step": 3620
+    },
+    {
+      "grad_norm": 0.3056441843509674,
+      "learning_rate": 3.695260400007819e-05,
+      "loss": 0.023,
+      "step": 3630
+    },
+    {
+      "grad_norm": 0.34800854325294495,
+      "learning_rate": 3.6686773210750385e-05,
+      "loss": 0.0253,
+      "step": 3640
+    },
+    {
+      "grad_norm": 0.3697214424610138,
+      "learning_rate": 3.642134684108737e-05,
+      "loss": 0.0246,
+      "step": 3650
+    },
+    {
+      "grad_norm": 0.31739336252212524,
+      "learning_rate": 3.615633295402123e-05,
+      "loss": 0.0254,
+      "step": 3660
+    },
+    {
+      "grad_norm": 0.3576298952102661,
+      "learning_rate": 3.5891739599953945e-05,
+      "loss": 0.0231,
+      "step": 3670
+    },
+    {
+      "grad_norm": 0.4195830225944519,
+      "learning_rate": 3.5627574816512846e-05,
+      "loss": 0.0271,
+      "step": 3680
+    },
+    {
+      "grad_norm": 0.38255468010902405,
+      "learning_rate": 3.536384662830648e-05,
+      "loss": 0.0263,
+      "step": 3690
+    },
+    {
+      "grad_norm": 0.31736278533935547,
+      "learning_rate": 3.5100563046680764e-05,
+      "loss": 0.0243,
+      "step": 3700
+    },
+    {
+      "grad_norm": 0.38382217288017273,
+      "learning_rate": 3.483773206947572e-05,
+      "loss": 0.0233,
+      "step": 3710
+    },
+    {
+      "grad_norm": 0.37553671002388,
+      "learning_rate": 3.457536168078247e-05,
+      "loss": 0.0233,
+      "step": 3720
+    },
+    {
+      "grad_norm": 0.45702260732650757,
+      "learning_rate": 3.431345985070067e-05,
+      "loss": 0.0248,
+      "step": 3730
+    },
+    {
+      "grad_norm": 0.30532407760620117,
+      "learning_rate": 3.40520345350965e-05,
+      "loss": 0.0232,
+      "step": 3740
+    },
+    {
+      "grad_norm": 0.3560055196285248,
+      "learning_rate": 3.379109367536089e-05,
+      "loss": 0.0237,
+      "step": 3750
+    },
+    {
+      "grad_norm": 0.3362725079059601,
+      "learning_rate": 3.3530645198168295e-05,
+      "loss": 0.0243,
+      "step": 3760
+    },
+    {
+      "grad_norm": 0.32288989424705505,
+      "learning_rate": 3.327069701523595e-05,
+      "loss": 0.0232,
+      "step": 3770
+    },
+    {
+      "grad_norm": 0.35202544927597046,
+      "learning_rate": 3.301125702308353e-05,
+      "loss": 0.0248,
+      "step": 3780
+    },
+    {
+      "grad_norm": 0.3243369460105896,
+      "learning_rate": 3.275233310279321e-05,
+      "loss": 0.023,
+      "step": 3790
+    },
+    {
+      "grad_norm": 0.3195122480392456,
+      "learning_rate": 3.249393311977037e-05,
+      "loss": 0.024,
+      "step": 3800
+    },
+    {
+      "grad_norm": 0.26877087354660034,
+      "learning_rate": 3.223606492350451e-05,
+      "loss": 0.0232,
+      "step": 3810
+    },
+    {
+      "grad_norm": 0.36210641264915466,
+      "learning_rate": 3.197873634733096e-05,
+      "loss": 0.0241,
+      "step": 3820
+    },
+    {
+      "grad_norm": 0.3320061266422272,
+      "learning_rate": 3.172195520819285e-05,
+      "loss": 0.0247,
+      "step": 3830
+    },
+    {
+      "grad_norm": 0.44787928462028503,
+      "learning_rate": 3.146572930640362e-05,
+      "loss": 0.0233,
+      "step": 3840
+    },
+    {
+      "grad_norm": 0.3908366858959198,
+      "learning_rate": 3.121006642541014e-05,
+      "loss": 0.0246,
+      "step": 3850
+    },
+    {
+      "grad_norm": 0.3663296103477478,
+      "learning_rate": 3.095497433155626e-05,
+      "loss": 0.0236,
+      "step": 3860
+    },
+    {
+      "grad_norm": 0.36651811003685,
+      "learning_rate": 3.070046077384682e-05,
+      "loss": 0.0234,
+      "step": 3870
+    },
+    {
+      "grad_norm": 0.2849855422973633,
+      "learning_rate": 3.0446533483712304e-05,
+      "loss": 0.0231,
+      "step": 3880
+    },
+    {
+      "grad_norm": 0.30322885513305664,
+      "learning_rate": 3.0193200174774038e-05,
+      "loss": 0.0258,
+      "step": 3890
+    },
+    {
+      "grad_norm": 0.34874120354652405,
+      "learning_rate": 2.994046854260974e-05,
+      "loss": 0.0217,
+      "step": 3900
+    },
+    {
+      "grad_norm": 0.3394639790058136,
+      "learning_rate": 2.9688346264519866e-05,
+      "loss": 0.0218,
+      "step": 3910
+    },
+    {
+      "grad_norm": 0.2971331775188446,
+      "learning_rate": 2.943684099929436e-05,
+      "loss": 0.0235,
+      "step": 3920
+    },
+    {
+      "grad_norm": 0.3201124370098114,
+      "learning_rate": 2.918596038697995e-05,
+      "loss": 0.0233,
+      "step": 3930
+    },
+    {
+      "grad_norm": 0.3573104739189148,
+      "learning_rate": 2.8935712048648112e-05,
+      "loss": 0.0222,
+      "step": 3940
+    },
+    {
+      "grad_norm": 0.3573231101036072,
+      "learning_rate": 2.8686103586163626e-05,
+      "loss": 0.0232,
+      "step": 3950
+    },
+    {
+      "grad_norm": 0.3325694799423218,
+      "learning_rate": 2.843714258195346e-05,
+      "loss": 0.0229,
+      "step": 3960
+    },
+    {
+      "grad_norm": 0.349811315536499,
+      "learning_rate": 2.8188836598776662e-05,
+      "loss": 0.0236,
+      "step": 3970
+    },
+    {
+      "grad_norm": 0.3447701334953308,
+      "learning_rate": 2.7941193179494484e-05,
+      "loss": 0.0236,
+      "step": 3980
+    },
+    {
+      "grad_norm": 0.3687022030353546,
+      "learning_rate": 2.7694219846841262e-05,
+      "loss": 0.0249,
+      "step": 3990
+    },
+    {
+      "grad_norm": 0.3964892327785492,
+      "learning_rate": 2.7447924103195976e-05,
+      "loss": 0.0222,
+      "step": 4000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 6000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9223372036854775807,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 128,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-4000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c25b10b9ea2b48c7d392e8e72ee1e37b2d2fadcf822e76df74171586e2f74941
+size 7633

checkpoint-4000/wandb_config.json

@@ -0,0 +1 @@
+{"project": "finetune-gr00t-n1d6", "run_id": "g1_finetune"}

checkpoint-4000/zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if ZERO_STAGE not in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info("Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info("Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

checkpoint-6000/config.json

@@ -0,0 +1,70 @@
+{
+  "action_horizon": 50,
+  "add_pos_embed": true,
+  "apply_sincos_state_encoding": true,
+  "architectures": [
+    "Gr00tN1d6"
+  ],
+  "attn_dropout": 0.2,
+  "attn_implementation": null,
+  "backbone_embedding_dim": 2048,
+  "backbone_model_type": "eagle",
+  "backbone_trainable_params_fp32": true,
+  "collator_overwrite_image_inputs": false,
+  "color_jitter_params": {
+    "brightness": 0.1,
+    "contrast": 0.1,
+    "hue": 0.1,
+    "saturation": 0.1
+  },
+  "crop_fraction": 0.95,
+  "diffusion_model_cfg": {
+    "attention_head_dim": 48,
+    "dropout": 0.2,
+    "final_dropout": true,
+    "interleave_self_attention": true,
+    "norm_type": "ada_norm",
+    "num_attention_heads": 32,
+    "num_layers": 32,
+    "output_dim": 1024,
+    "positional_embeddings": null
+  },
+  "eagle_collator": true,
+  "formalize_language": true,
+  "gemma_collator": false,
+  "hidden_size": 1024,
+  "image_crop_size": null,
+  "image_target_size": null,
+  "input_embedding_dim": 1536,
+  "load_bf16": true,
+  "max_action_dim": 128,
+  "max_num_embodiments": 32,
+  "max_seq_len": 1024,
+  "max_state_dim": 128,
+  "model_dtype": "bfloat16",
+  "model_name": "nvidia/Eagle-Block2A-2B-v2",
+  "model_type": "Gr00tN1d6",
+  "noise_beta_alpha": 1.5,
+  "noise_beta_beta": 1.0,
+  "noise_s": 0.999,
+  "num_inference_timesteps": 4,
+  "num_timestep_buckets": 1000,
+  "random_rotation_angle": null,
+  "reproject_vision": false,
+  "select_layer": 16,
+  "shortest_image_edge": 256,
+  "state_dropout_prob": 0.0,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "tune_diffusion_model": true,
+  "tune_llm": false,
+  "tune_projector": true,
+  "tune_top_llm_layers": 4,
+  "tune_visual": false,
+  "tune_vlln": true,
+  "use_albumentations_transforms": true,
+  "use_alternate_vl_dit": true,
+  "use_flash_attention": true,
+  "use_relative_action": true,
+  "use_vlln": true
+}

checkpoint-6000/embodiment_id.json

@@ -0,0 +1,11 @@
+{
+  "robocasa_panda_omron": 13,
+  "gr1": 20,
+  "behavior_r1_pro": 24,
+  "unitree_g1": 8,
+  "oxe_google": 0,
+  "oxe_widowx": 1,
+  "libero_panda": 2,
+  "oxe_droid": 16,
+  "new_embodiment": 10
+}

checkpoint-6000/experiment_cfg/conf.yaml

@@ -0,0 +1,252 @@
+load_config_path: null
+model:
+  model_type: Gr00tN1d6
+  model_dtype: bfloat16
+  model_name: nvidia/Eagle-Block2A-2B-v2
+  backbone_model_type: eagle
+  model_revision: null
+  tune_top_llm_layers: 4
+  backbone_embedding_dim: 2048
+  tune_llm: false
+  tune_visual: false
+  select_layer: 16
+  reproject_vision: false
+  use_flash_attention: true
+  load_bf16: false
+  collator_overwrite_image_inputs: false
+  eagle_collator: true
+  backbone_trainable_params_fp32: true
+  image_crop_size: null
+  image_target_size: null
+  shortest_image_edge: 256
+  crop_fraction: 0.95
+  random_rotation_angle: null
+  color_jitter_params:
+    brightness: 0.3
+    contrast: 0.4
+    saturation: 0.5
+    hue: 0.08
+  use_albumentations_transforms: true
+  formalize_language: true
+  apply_sincos_state_encoding: false
+  use_relative_action: true
+  max_state_dim: 29
+  max_action_dim: 29
+  action_horizon: 16
+  hidden_size: 1024
+  input_embedding_dim: 1536
+  add_pos_embed: true
+  attn_dropout: 0.2
+  use_vlln: true
+  max_seq_len: 1024
+  use_alternate_vl_dit: true
+  attend_text_every_n_blocks: 2
+  diffusion_model_cfg:
+    positional_embeddings: null
+    num_layers: 32
+    num_attention_heads: 32
+    attention_head_dim: 48
+    norm_type: ada_norm
+    dropout: 0.2
+    final_dropout: true
+    output_dim: 1024
+    interleave_self_attention: true
+  num_inference_timesteps: 4
+  noise_beta_alpha: 1.5
+  noise_beta_beta: 1.0
+  noise_s: 0.999
+  num_timestep_buckets: 1000
+  tune_projector: true
+  tune_diffusion_model: true
+  tune_vlln: true
+  state_dropout_prob: 0.0
+  state_additive_noise_scale: 0.0
+  max_num_embodiments: 32
+data:
+  datasets:
+  - dataset_paths:
+    - /data/PnPCubeLine
+    embodiment_tag: unitree_g1
+    mix_ratio: 1.0
+    dataset_type: physical_embodiment
+    val_dataset_path: null
+  modality_configs:
+    unitree_g1:
+      video:
+        delta_indices:
+        - 0
+        modality_keys:
+        - ego_view
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+      state:
+        delta_indices:
+        - 0
+        modality_keys:
+        - left_leg
+        - right_leg
+        - waist
+        - left_arm
+        - right_arm
+        - left_hand
+        - right_hand
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+      action:
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        - 24
+        - 25
+        - 26
+        - 27
+        - 28
+        - 29
+        modality_keys:
+        - left_arm
+        - right_arm
+        - left_hand
+        - right_hand
+        - waist
+        - base_height_command
+        - navigate_command
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs:
+        - rep: RELATIVE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: RELATIVE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+        - rep: ABSOLUTE
+          type: NON_EEF
+          format: DEFAULT
+          state_key: null
+      language:
+        delta_indices:
+        - 0
+        modality_keys:
+        - annotation.human.task_description
+        sin_cos_embedding_keys: null
+        mean_std_embedding_keys: null
+        action_configs: null
+  download_cache: false
+  shard_size: 1024
+  episode_sampling_rate: 0.1
+  num_shards_per_epoch: 100000
+  override_pretraining_statistics: false
+  mode: single_turn
+  random_chop: 0.0
+  mock_dataset_mode: false
+  shuffle: true
+  seed: 42
+  multiprocessing_context: fork
+  allow_padding: false
+  subsample_ratio: 1.0
+  image_crop_size:
+  - 244
+  - 244
+  image_target_size:
+  - 224
+  - 224
+  video_backend: torchcodec
+training:
+  output_dir: /checkpoints/g1_finetune
+  experiment_name: null
+  max_steps: 6000
+  global_batch_size: 512
+  batch_size: null
+  gradient_accumulation_steps: 1
+  learning_rate: 0.0001
+  lr_scheduler_type: cosine
+  weight_decay: 1.0e-05
+  warmup_ratio: 0.05
+  warmup_steps: 0
+  max_grad_norm: 1.0
+  optim: adamw_torch
+  start_from_checkpoint: nvidia/GR00T-N1.6-3B
+  tf32: true
+  fp16: false
+  bf16: true
+  eval_bf16: true
+  logging_steps: 10
+  save_steps: 2000
+  save_total_limit: 2
+  save_vl_model: false
+  upload_checkpoints: false
+  upload_every: 1000
+  upload_last_n_checkpoints: 5
+  max_concurrent_uploads: 2
+  eval_strategy: 'no'
+  eval_steps: 500
+  eval_set_split_ratio: 0.1
+  eval_batch_size: 2
+  save_best_eval_metric_name: ''
+  save_best_eval_metric_greater_is_better: true
+  deepspeed_stage: 2
+  gradient_checkpointing: false
+  transformers_trust_remote_code: true
+  transformers_local_files_only: false
+  transformers_cache_dir: null
+  transformers_access_token: null
+  use_ddp: false
+  ddp_bucket_cap_mb: 100
+  num_gpus: 4
+  dataloader_num_workers: 6
+  remove_unused_columns: false
+  use_wandb: false
+  wandb_project: finetune-gr00t-n1d6
+  enable_profiling: false
+  max_retries: 3
+  assert_loss_less_than: null
+  add_rl_callback: false
+  enable_open_loop_eval: false
+  open_loop_eval_traj_ids:
+  - 0
+  open_loop_eval_steps_per_traj: 100
+  open_loop_eval_plot_indices: null
+max_steps: 6000
+save_steps: 2000

checkpoint-6000/experiment_cfg/config.yaml

@@ -0,0 +1,291 @@
+!!python/object:gr00t.configs.base_config.Config
+data: !!python/object:gr00t.configs.data.data_config.DataConfig
+  allow_padding: false
+  datasets:
+  - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
+    dataset_paths:
+    - /data/PnPCubeLine
+    dataset_type: physical_embodiment
+    embodiment_tag: unitree_g1
+    mix_ratio: 1.0
+    val_dataset_path: null
+  download_cache: false
+  episode_sampling_rate: 0.1
+  image_crop_size:
+  - 244
+  - 244
+  image_target_size:
+  - 224
+  - 224
+  mock_dataset_mode: false
+  modality_configs:
+    unitree_g1:
+      action: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs:
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
+          - default
+          rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
+          - relative
+          state_key: null
+          type: &id003 !!python/object/apply:gr00t.data.types.ActionType
+          - non_eef
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id002
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: &id004 !!python/object/apply:gr00t.data.types.ActionRepresentation
+          - absolute
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id004
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id004
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id004
+          state_key: null
+          type: *id003
+        - !!python/object:gr00t.data.types.ActionConfig
+          format: *id001
+          rep: *id004
+          state_key: null
+          type: *id003
+        delta_indices:
+        - 0
+        - 1
+        - 2
+        - 3
+        - 4
+        - 5
+        - 6
+        - 7
+        - 8
+        - 9
+        - 10
+        - 11
+        - 12
+        - 13
+        - 14
+        - 15
+        - 16
+        - 17
+        - 18
+        - 19
+        - 20
+        - 21
+        - 22
+        - 23
+        - 24
+        - 25
+        - 26
+        - 27
+        - 28
+        - 29
+        mean_std_embedding_keys: null
+        modality_keys:
+        - left_arm
+        - right_arm
+        - left_hand
+        - right_hand
+        - waist
+        - base_height_command
+        - navigate_command
+        sin_cos_embedding_keys: null
+      language: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - 0
+        mean_std_embedding_keys: null
+        modality_keys:
+        - annotation.human.task_description
+        sin_cos_embedding_keys: null
+      state: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - 0
+        mean_std_embedding_keys: null
+        modality_keys:
+        - left_leg
+        - right_leg
+        - waist
+        - left_arm
+        - right_arm
+        - left_hand
+        - right_hand
+        sin_cos_embedding_keys: null
+      video: !!python/object:gr00t.data.types.ModalityConfig
+        action_configs: null
+        delta_indices:
+        - 0
+        mean_std_embedding_keys: null
+        modality_keys:
+        - ego_view
+        sin_cos_embedding_keys: null
+  mode: single_turn
+  multiprocessing_context: fork
+  num_shards_per_epoch: 100000
+  override_pretraining_statistics: false
+  random_chop: 0.0
+  seed: 42
+  shard_size: 1024
+  shuffle: true
+  subsample_ratio: 1.0
+  video_backend: torchcodec
+load_config_path: null
+model: !!python/object:gr00t.configs.model.gr00t_n1d6.Gr00tN1d6Config
+  _attn_implementation_autoset: false
+  _attn_implementation_internal: null
+  _commit_hash: null
+  _name_or_path: ''
+  add_cross_attention: false
+  architectures: null
+  backbone_model_type: eagle
+  backbone_trainable_params_fp32: true
+  bad_words_ids: null
+  begin_suppress_tokens: null
+  bos_token_id: null
+  chunk_size_feed_forward: 0
+  color_jitter_params:
+    brightness: 0.3
+    contrast: 0.4
+    hue: 0.08
+    saturation: 0.5
+  cross_attention_hidden_size: null
+  decoder_start_token_id: null
+  diffusion_model_cfg:
+    attention_head_dim: 48
+    dropout: 0.2
+    final_dropout: true
+    interleave_self_attention: true
+    norm_type: ada_norm
+    num_attention_heads: 32
+    num_layers: 32
+    output_dim: 1024
+    positional_embeddings: null
+  diversity_penalty: 0.0
+  do_sample: false
+  eagle_collator: true
+  early_stopping: false
+  encoder_no_repeat_ngram_size: 0
+  eos_token_id: null
+  exponential_decay_length_penalty: null
+  finetuning_task: null
+  forced_bos_token_id: null
+  forced_eos_token_id: null
+  id2label:
+    0: LABEL_0
+    1: LABEL_1
+  is_decoder: false
+  is_encoder_decoder: false
+  label2id:
+    LABEL_0: 0
+    LABEL_1: 1
+  length_penalty: 1.0
+  load_bf16: false
+  max_length: 20
+  min_length: 0
+  model_name: nvidia/Eagle-Block2A-2B-v2
+  no_repeat_ngram_size: 0
+  num_beam_groups: 1
+  num_beams: 1
+  num_return_sequences: 1
+  output_attentions: false
+  output_hidden_states: false
+  output_scores: false
+  pad_token_id: null
+  prefix: null
+  problem_type: null
+  pruned_heads: {}
+  random_rotation_angle: null
+  remove_invalid_values: false
+  repetition_penalty: 1.0
+  reproject_vision: false
+  return_dict: true
+  return_dict_in_generate: false
+  sep_token_id: null
+  state_dropout_prob: 0.0
+  suppress_tokens: null
+  task_specific_params: null
+  temperature: 1.0
+  tf_legacy_loss: false
+  tie_encoder_decoder: false
+  tie_word_embeddings: true
+  tokenizer_class: null
+  top_k: 50
+  top_p: 1.0
+  torch_dtype: null
+  torchscript: false
+  transformers_version: null
+  tune_diffusion_model: true
+  tune_llm: false
+  tune_projector: true
+  tune_visual: false
+  typical_p: 1.0
+  use_bfloat16: false
+  use_relative_action: true
+training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
+  add_rl_callback: false
+  assert_loss_less_than: null
+  batch_size: null
+  bf16: true
+  dataloader_num_workers: 6
+  ddp_bucket_cap_mb: 100
+  deepspeed_stage: 2
+  enable_open_loop_eval: false
+  enable_profiling: false
+  eval_batch_size: 2
+  eval_bf16: true
+  eval_set_split_ratio: 0.1
+  eval_steps: 500
+  eval_strategy: 'no'
+  experiment_name: null
+  fp16: false
+  global_batch_size: 512
+  gradient_accumulation_steps: 1
+  gradient_checkpointing: false
+  learning_rate: 0.0001
+  logging_steps: 10
+  lr_scheduler_type: cosine
+  max_concurrent_uploads: 2
+  max_grad_norm: 1.0
+  max_retries: 3
+  max_steps: 6000
+  num_gpus: 4
+  open_loop_eval_plot_indices: null
+  open_loop_eval_steps_per_traj: 100
+  open_loop_eval_traj_ids:
+  - 0
+  optim: adamw_torch
+  output_dir: /checkpoints/g1_finetune
+  remove_unused_columns: false
+  save_best_eval_metric_greater_is_better: true
+  save_best_eval_metric_name: ''
+  save_steps: 2000
+  save_total_limit: 2
+  save_vl_model: false
+  start_from_checkpoint: nvidia/GR00T-N1.6-3B
+  tf32: true
+  transformers_access_token: null
+  transformers_cache_dir: null
+  transformers_local_files_only: false
+  transformers_trust_remote_code: true
+  upload_checkpoints: false
+  upload_every: 1000
+  upload_last_n_checkpoints: 5
+  use_ddp: false
+  use_wandb: false
+  wandb_project: finetune-gr00t-n1d6
+  warmup_ratio: 0.05
+  warmup_steps: 0
+  weight_decay: 1.0e-05

checkpoint-6000/experiment_cfg/final_model_config.json

@@ -0,0 +1,53 @@
+{
+  "model_type": "Gr00tN1d6",
+  "model_dtype": "bfloat16",
+  "model_name": "nvidia/Eagle-Block2A-2B-v2",
+  "backbone_model_type": "eagle",
+  "model_revision": null,
+  "tune_top_llm_layers": 4,
+  "backbone_embedding_dim": 2048,
+  "tune_llm": false,
+  "tune_visual": false,
+  "select_layer": 16,
+  "reproject_vision": false,
+  "use_flash_attention": true,
+  "load_bf16": true,
+  "collator_overwrite_image_inputs": false,
+  "eagle_collator": true,
+  "backbone_trainable_params_fp32": true,
+  "apply_sincos_state_encoding": true,
+  "use_relative_action": true,
+  "max_state_dim": 128,
+  "max_action_dim": 128,
+  "action_horizon": 50,
+  "hidden_size": 1024,
+  "input_embedding_dim": 1536,
+  "add_pos_embed": true,
+  "attn_dropout": 0.2,
+  "use_vlln": true,
+  "max_seq_len": 1024,
+  "use_alternate_vl_dit": true,
+  "attend_text_every_n_blocks": 2,
+  "diffusion_model_cfg": {
+    "attention_head_dim": 48,
+    "dropout": 0.2,
+    "final_dropout": true,
+    "interleave_self_attention": true,
+    "norm_type": "ada_norm",
+    "num_attention_heads": 32,
+    "num_layers": 32,
+    "output_dim": 1024,
+    "positional_embeddings": null
+  },
+  "num_inference_timesteps": 4,
+  "noise_beta_alpha": 1.5,
+  "noise_beta_beta": 1.0,
+  "noise_s": 0.999,
+  "num_timestep_buckets": 1000,
+  "tune_projector": true,
+  "tune_diffusion_model": true,
+  "tune_vlln": true,
+  "state_dropout_prob": 0.0,
+  "state_additive_noise_scale": 0.0,
+  "max_num_embodiments": 32
+}

checkpoint-6000/global_step6000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d716d54756cabdab94f305ee9fe2548dca14c8d55ec6225e672cd7f2b104ce85
+size 4859923697

checkpoint-6000/global_step6000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96da5e13c16537540cecd04633ba9c42405d2038a81544cabf2e81a07253f7a0
+size 4859923761

checkpoint-6000/global_step6000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dab25e68e17f63192bf2f0550da355bab7a5a8a23b8643d7498135100cbc76c6
+size 4859923569

checkpoint-6000/global_step6000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59b204d834aebe4e09576e19c9f92706e25b33d0277d20566867e6d86fa8ea56
+size 4859917873

checkpoint-6000/global_step6000/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb7845d00c4c234892e97bf9f3a806e54e3e889297adbf77a5cb1b8e1e755bf8
+size 9907202435

checkpoint-6000/latest

@@ -0,0 +1 @@
+global_step6000

checkpoint-6000/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fecd3662763ea4a4067d707921778c923bb3c854600c032405302b58375c9534
+size 4991091456

checkpoint-6000/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1ad92be30818a153597b184bc1bc18dad71b35e8bc06dcb299c18d9ef33f159
+size 1582283096

checkpoint-6000/processor_config.json

@@ -0,0 +1,508 @@
+{
+  "processor_class": "Gr00tN1d6Processor",
+  "processor_kwargs": {
+    "modality_configs": {
+      "behavior_r1_pro": {
+        "video": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "observation.images.rgb.head_256_256",
+            "observation.images.rgb.left_wrist_256_256",
+            "observation.images.rgb.right_wrist_256_256"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "state": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "robot_pos",
+            "robot_ori_cos",
+            "robot_ori_sin",
+            "robot_2d_ori",
+            "robot_2d_ori_cos",
+            "robot_2d_ori_sin",
+            "robot_lin_vel",
+            "robot_ang_vel",
+            "arm_left_qpos",
+            "arm_left_qpos_sin",
+            "arm_left_qpos_cos",
+            "eef_left_pos",
+            "eef_left_quat",
+            "gripper_left_qpos",
+            "arm_right_qpos",
+            "arm_right_qpos_sin",
+            "arm_right_qpos_cos",
+            "eef_right_pos",
+            "eef_right_quat",
+            "gripper_right_qpos",
+            "trunk_qpos"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "action": {
+          "delta_indices": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+            12,
+            13,
+            14,
+            15,
+            16,
+            17,
+            18,
+            19,
+            20,
+            21,
+            22,
+            23,
+            24,
+            25,
+            26,
+            27,
+            28,
+            29,
+            30,
+            31
+          ],
+          "modality_keys": [
+            "base",
+            "torso",
+            "left_arm",
+            "left_gripper",
+            "right_arm",
+            "right_gripper"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": [
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": "trunk_qpos"
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": "arm_left_qpos"
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": "arm_right_qpos"
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            }
+          ]
+        },
+        "language": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "annotation.human.coarse_action"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        }
+      },
+      "gr1": {
+        "video": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "ego_view_bg_crop_pad_res256_freq20"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "state": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand",
+            "waist"
+          ],
+          "sin_cos_embedding_keys": [
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand",
+            "waist"
+          ],
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "action": {
+          "delta_indices": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+            12,
+            13,
+            14,
+            15
+          ],
+          "modality_keys": [
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand",
+            "waist"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": [
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            }
+          ]
+        },
+        "language": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "task"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        }
+      },
+      "robocasa_panda_omron": {
+        "video": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "res256_image_side_0",
+            "res256_image_side_1",
+            "res256_image_wrist_0"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "state": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "end_effector_position_relative",
+            "end_effector_rotation_relative",
+            "gripper_qpos",
+            "base_position",
+            "base_rotation"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "action": {
+          "delta_indices": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+            12,
+            13,
+            14,
+            15
+          ],
+          "modality_keys": [
+            "end_effector_position",
+            "end_effector_rotation",
+            "gripper_close",
+            "base_motion",
+            "control_mode"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": [
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            }
+          ]
+        },
+        "language": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "annotation.human.action.task_description"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        }
+      },
+      "unitree_g1": {
+        "video": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "ego_view"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "state": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "left_leg",
+            "right_leg",
+            "waist",
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        },
+        "action": {
+          "delta_indices": [
+            0,
+            1,
+            2,
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+            12,
+            13,
+            14,
+            15,
+            16,
+            17,
+            18,
+            19,
+            20,
+            21,
+            22,
+            23,
+            24,
+            25,
+            26,
+            27,
+            28,
+            29
+          ],
+          "modality_keys": [
+            "left_arm",
+            "right_arm",
+            "left_hand",
+            "right_hand",
+            "waist",
+            "base_height_command",
+            "navigate_command"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": [
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "RELATIVE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            },
+            {
+              "rep": "ABSOLUTE",
+              "type": "NON_EEF",
+              "format": "DEFAULT",
+              "state_key": null
+            }
+          ]
+        },
+        "language": {
+          "delta_indices": [
+            0
+          ],
+          "modality_keys": [
+            "annotation.human.task_description"
+          ],
+          "sin_cos_embedding_keys": null,
+          "mean_std_embedding_keys": null,
+          "action_configs": null
+        }
+      }
+    },
+    "image_crop_size": null,
+    "image_target_size": null,
+    "use_albumentations": true,
+    "random_rotation_angle": null,
+    "color_jitter_params": {
+      "brightness": 0.3,
+      "contrast": 0.4,
+      "saturation": 0.5,
+      "hue": 0.08
+    },
+    "shortest_image_edge": 256,
+    "crop_fraction": 0.95,
+    "model_name": "nvidia/Eagle-Block2A-2B-v2",
+    "model_type": "eagle",
+    "formalize_language": true,
+    "max_state_dim": 128,
+    "max_action_dim": 128,
+    "max_action_horizon": 50,
+    "use_percentiles": false,
+    "clip_outliers": true,
+    "apply_sincos_state_encoding": true,
+    "use_relative_action": true
+  }
+}

checkpoint-6000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9519287980bbfb4ff0f1c0329de2253b6dd2be2b39ad90c99159d18f01d97bfa
+size 15429

checkpoint-6000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83f65f4f53a4142b631ecec686fa131a767bf6d5033059c906cb4f037c82a328
+size 15429

checkpoint-6000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dcd7a12f12b34f5b5c8ad691f5721ecbb5f4591a725af4e652b91c4540b5493
+size 15429

checkpoint-6000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:117be16b69be6f924996dc4bca36bcb32e2a58c09d436f6d80629cea2f152550
+size 15429

checkpoint-6000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c960d31406b345d9ea0d680e904d6145dbe0fc3b960fbf26ae378848e83ec25
+size 1465