Hugging Face's logo

chanwoochan
/
daru_mob_comp_minimum_0715

Safetensors
Gr00tN1d7
Model card Files
xet
 Community
chanwoochan commited on
Commit
7d2a928
·
verified ·
1 Parent(s): ae043fb

Upload folder using huggingface_hub

Browse files
Files changed (34) hide show
  1. checkpoint-50000/config.json +88 -0
  2. checkpoint-50000/embodiment_id.json +57 -0
  3. checkpoint-50000/experiment_cfg/conf.yaml +251 -0
  4. checkpoint-50000/experiment_cfg/config.yaml +249 -0
  5. checkpoint-50000/experiment_cfg/dataset_statistics.json +595 -0
  6. checkpoint-50000/experiment_cfg/final_model_config.json +56 -0
  7. checkpoint-50000/experiment_cfg/final_processor_config.json +0 -0
  8. checkpoint-50000/model-00001-of-00003.safetensors +3 -0
  9. checkpoint-50000/model-00002-of-00003.safetensors +3 -0
  10. checkpoint-50000/model-00003-of-00003.safetensors +3 -0
  11. checkpoint-50000/model.safetensors.index.json +0 -0
  12. checkpoint-50000/optimizer.pt +3 -0
  13. checkpoint-50000/processor_config.json +1171 -0
  14. checkpoint-50000/rng_state.pth +3 -0
  15. checkpoint-50000/scheduler.pt +3 -0
  16. checkpoint-50000/statistics.json +0 -0
  17. checkpoint-50000/trainer_state.json +0 -0
  18. checkpoint-50000/training_args.bin +3 -0
  19. checkpoint-50000/wandb_config.json +1 -0
  20. config.json +88 -0
  21. experiment_cfg/conf.yaml +251 -0
  22. experiment_cfg/config.yaml +249 -0
  23. experiment_cfg/dataset_statistics.json +595 -0
  24. experiment_cfg/final_model_config.json +56 -0
  25. experiment_cfg/final_processor_config.json +0 -0
  26. model-00001-of-00003.safetensors +3 -0
  27. model-00002-of-00003.safetensors +3 -0
  28. model-00003-of-00003.safetensors +3 -0
  29. model.safetensors.index.json +0 -0
  30. processor/embodiment_id.json +57 -0
  31. processor/processor_config.json +1171 -0
  32. processor/statistics.json +0 -0
  33. training_args.bin +3 -0
  34. wandb_config.json +1 -0
checkpoint-50000/config.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_horizon": 40,
3
+ "add_pos_embed": true,
4
+ "apply_sincos_state_encoding": false,
5
+ "architectures": [
6
+ "Gr00tN1d7"
7
+ ],
8
+ "attn_dropout": 0.2,
9
+ "attn_implementation": null,
10
+ "backbone_embedding_dim": 2048,
11
+ "backbone_trainable_params_fp32": true,
12
+ "color_jitter_params": {
13
+ "brightness": 0.3,
14
+ "contrast": 0.4,
15
+ "hue": 0.08,
16
+ "saturation": 0.5
17
+ },
18
+ "crop_fraction": 0.95,
19
+ "diffusion_model_cfg": {
20
+ "attention_head_dim": 48,
21
+ "dropout": 0.2,
22
+ "final_dropout": true,
23
+ "interleave_self_attention": true,
24
+ "norm_type": "ada_norm",
25
+ "num_attention_heads": 32,
26
+ "num_layers": 32,
27
+ "output_dim": 1024,
28
+ "positional_embeddings": null
29
+ },
30
+ "dtype": "float32",
31
+ "exclude_state": false,
32
+ "formalize_language": true,
33
+ "hidden_size": 1024,
34
+ "image_crop_size": [
35
+ 230,
36
+ 230
37
+ ],
38
+ "image_target_size": [
39
+ 256,
40
+ 256
41
+ ],
42
+ "letter_box_transform": false,
43
+ "load_bf16": false,
44
+ "max_action_dim": 132,
45
+ "max_num_embodiments": 32,
46
+ "max_seq_len": 1024,
47
+ "max_state_dim": 132,
48
+ "model_dtype": "bfloat16",
49
+ "model_name": "nvidia/Cosmos-Reason2-2B",
50
+ "model_type": "Gr00tN1d7",
51
+ "noise_beta_alpha": 1.5,
52
+ "noise_beta_beta": 1.0,
53
+ "noise_s": 0.999,
54
+ "num_inference_timesteps": 4,
55
+ "num_timestep_buckets": 1000,
56
+ "random_history_crop": true,
57
+ "random_rotation_angle": 0,
58
+ "reproject_vision": false,
59
+ "rtc_ramp_rate": 6.0,
60
+ "select_layer": 16,
61
+ "shortest_image_edge": 256,
62
+ "state_dropout_prob": 0.2,
63
+ "state_gaussian_noise_std": 0.0,
64
+ "transformers_version": "4.57.3",
65
+ "tune_diffusion_model": true,
66
+ "tune_linear": true,
67
+ "tune_llm": false,
68
+ "tune_projector": true,
69
+ "tune_top_llm_layers": 0,
70
+ "tune_visual": false,
71
+ "tune_vlln": true,
72
+ "use_albumentations": true,
73
+ "use_alternate_vl_dit": true,
74
+ "use_flash_attention": true,
75
+ "use_future_tokens": false,
76
+ "use_mean_std": false,
77
+ "use_percentiles": true,
78
+ "use_vl_self_attention": true,
79
+ "use_vlln": true,
80
+ "vl_self_attention_cfg": {
81
+ "attention_head_dim": 64,
82
+ "dropout": 0.2,
83
+ "final_dropout": true,
84
+ "num_attention_heads": 32,
85
+ "num_layers": 4,
86
+ "positional_embeddings": null
87
+ }
88
+ }
checkpoint-50000/embodiment_id.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "robocasa_panda_omron": 13,
3
+ "oxe_droid": 17,
4
+ "oxe_fractal": 18,
5
+ "oxe_language_table": 19,
6
+ "oxe_bridge": 20,
7
+ "unknown": 22,
8
+ "gr1_unified": 20,
9
+ "agibot": 26,
10
+ "sim_behavior_r1_pro": 23,
11
+ "xdof": 24,
12
+ "xdof_oss_data": 25,
13
+ "unitree_g1_full_body_with_waist_height_nav_cmd": 25,
14
+ "real_r1_pro_sharpa": 27,
15
+ "real_r1_pro_sharpa_add_view": 27,
16
+ "real_r1_pro_sharpa_relative_arm_joint": 26,
17
+ "real_r1_pro_sharpa_delta_eef": 26,
18
+ "real_r1_pro_sharpa_absolute_eef": 26,
19
+ "real_r1_pro_sharpa_meanstd": 26,
20
+ "real_r1_pro_sharpa_relative_eef": 26,
21
+ "real_r1_pro_sharpa_relative_eef_add_view": 26,
22
+ "real_r1_pro_sharpa_relative_eef_relative_hand": 26,
23
+ "real_r1_pro_sharpa_relative_eef_human": 26,
24
+ "real_r1_pro_sharpa_relative_eef_human_add_view": 26,
25
+ "real_r1_pro_sharpa_relative_eef_human_relative_hand": 26,
26
+ "real_r1_pro_sharpa_relative_eef_egodex": 26,
27
+ "real_r1_pro_sharpa_relative_eef_egodex_relative_hand": 26,
28
+ "real_r1_pro_sharpa_relative_eef_egodex_wrist_only": 26,
29
+ "real_r1_pro_sharpa_relative_eef_maxinsights": 26,
30
+ "real_r1_pro_sharpa_relative_eef_maxinsights_relative_hand": 26,
31
+ "real_r1_pro_sharpa_relative_eef_mecka": 26,
32
+ "real_r1_pro_sharpa_relative_eef_mecka_relative_hand": 26,
33
+ "real_g1_relative_eef_absolute_joints": 25,
34
+ "real_g1_relative_eef_absolute_joints_wrist_cam": 25,
35
+ "real_g1_relative_eef_relative_joints": 25,
36
+ "real_r1_pro_sharpa_relative_eef_relative_hand_relative_joint": 26,
37
+ "real_r1_pro_sharpa_relative_joint": 29,
38
+ "oxe_droid_relative_eef_relative_joint": 24,
39
+ "oxe_droid_relative_eef_relative_joint_swapped": 24,
40
+ "oxe_droid_relative_eef_relative_joint_upweight_z": 24,
41
+ "oxe_droid_relative_eef_relative_joint_upweight_z_swapped": 24,
42
+ "oxe_droid_relative_eef_relative_joint_3view": 24,
43
+ "oxe_droid_relative_eef_relative_joint_3view_swapped": 24,
44
+ "oxe_droid_relative_eef": 24,
45
+ "oxe_droid_joint_position_relative": 24,
46
+ "xdof_relative_eef_relative_joint": 27,
47
+ "xdof_relative_eef_relative_joint_subtask": 27,
48
+ "xdof_relative_eef": 27,
49
+ "xdof_relative_joint": 28,
50
+ "simpler_env_google": 0,
51
+ "simpler_env_widowx": 1,
52
+ "libero_sim": 2,
53
+ "droid_sim": 3,
54
+ "new_embodiment": 10,
55
+ "robocasa_gr1_tabletop": 10,
56
+ "unitree_g1_sonic": 11
57
+ }
checkpoint-50000/experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,251 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d7
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Cosmos-Reason2-2B
6
+ backbone_model_type: qwen
7
+ model_revision: null
8
+ tune_top_llm_layers: 0
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: false
12
+ select_layer: 12
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ backbone_trainable_params_fp32: true
17
+ image_crop_size:
18
+ - 230
19
+ - 230
20
+ image_target_size:
21
+ - 256
22
+ - 256
23
+ shortest_image_edge: null
24
+ crop_fraction: null
25
+ random_rotation_angle: null
26
+ color_jitter_params: null
27
+ use_albumentations_transforms: true
28
+ letter_box_transform: false
29
+ extra_augmentation_config: null
30
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_percentiles: true
33
+ use_relative_action: true
34
+ max_state_dim: 132
35
+ max_action_dim: 132
36
+ action_horizon: 40
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ state_history_length: 1
40
+ add_pos_embed: true
41
+ attn_dropout: 0.2
42
+ use_vlln: true
43
+ max_seq_len: 1024
44
+ use_alternate_vl_dit: true
45
+ attend_text_every_n_blocks: 2
46
+ diffusion_model_cfg:
47
+ positional_embeddings: null
48
+ num_layers: 16
49
+ num_attention_heads: 32
50
+ attention_head_dim: 48
51
+ norm_type: ada_norm
52
+ dropout: 0.2
53
+ final_dropout: true
54
+ output_dim: 1024
55
+ interleave_self_attention: true
56
+ num_inference_timesteps: 4
57
+ noise_beta_alpha: 1.5
58
+ noise_beta_beta: 1.0
59
+ noise_s: 0.999
60
+ num_timestep_buckets: 1000
61
+ tune_projector: true
62
+ tune_diffusion_model: true
63
+ tune_vlln: true
64
+ state_dropout_prob: 0.2
65
+ exclude_state: false
66
+ use_mean_std: false
67
+ max_num_embodiments: 32
68
+ data:
69
+ datasets:
70
+ - dataset_paths:
71
+ - chanwoochan/daru_mob_comp_minimum_0715
72
+ embodiment_tag: new_embodiment
73
+ mix_ratio: 1.0
74
+ dataset_type: physical_embodiment
75
+ val_dataset_path: null
76
+ modality_configs:
77
+ new_embodiment:
78
+ video:
79
+ delta_indices:
80
+ - 0
81
+ modality_keys:
82
+ - head_left
83
+ - head_right
84
+ - hand_left
85
+ - hand_right
86
+ sin_cos_embedding_keys: null
87
+ mean_std_embedding_keys: null
88
+ action_configs: null
89
+ state:
90
+ delta_indices:
91
+ - 0
92
+ modality_keys:
93
+ - left_arm
94
+ - right_arm
95
+ - left_hand
96
+ - right_hand
97
+ - head
98
+ - waist
99
+ sin_cos_embedding_keys: null
100
+ mean_std_embedding_keys: null
101
+ action_configs: null
102
+ action:
103
+ delta_indices:
104
+ - 0
105
+ - 1
106
+ - 2
107
+ - 3
108
+ - 4
109
+ - 5
110
+ - 6
111
+ - 7
112
+ - 8
113
+ - 9
114
+ - 10
115
+ - 11
116
+ - 12
117
+ - 13
118
+ - 14
119
+ - 15
120
+ modality_keys:
121
+ - left_arm
122
+ - right_arm
123
+ - left_hand
124
+ - right_hand
125
+ - head
126
+ - waist
127
+ - mob_comp
128
+ - task_goal
129
+ sin_cos_embedding_keys: null
130
+ mean_std_embedding_keys: null
131
+ action_configs:
132
+ - rep: DELTA
133
+ type: NON_EEF
134
+ format: DEFAULT
135
+ state_key: null
136
+ - rep: DELTA
137
+ type: NON_EEF
138
+ format: DEFAULT
139
+ state_key: null
140
+ - rep: DELTA
141
+ type: NON_EEF
142
+ format: DEFAULT
143
+ state_key: null
144
+ - rep: DELTA
145
+ type: NON_EEF
146
+ format: DEFAULT
147
+ state_key: null
148
+ - rep: DELTA
149
+ type: NON_EEF
150
+ format: DEFAULT
151
+ state_key: null
152
+ - rep: DELTA
153
+ type: NON_EEF
154
+ format: DEFAULT
155
+ state_key: null
156
+ - rep: DELTA
157
+ type: NON_EEF
158
+ format: DEFAULT
159
+ state_key: null
160
+ - rep: DELTA
161
+ type: NON_EEF
162
+ format: DEFAULT
163
+ state_key: null
164
+ language:
165
+ delta_indices:
166
+ - 0
167
+ modality_keys:
168
+ - annotation.human.task_description
169
+ sin_cos_embedding_keys: null
170
+ mean_std_embedding_keys: null
171
+ action_configs: null
172
+ download_cache: false
173
+ shard_size: 1024
174
+ episode_sampling_rate: 0.1
175
+ num_shards_per_epoch: 100000
176
+ ds_weights_alpha: null
177
+ override_pretraining_statistics: true
178
+ mode: single_turn
179
+ random_chop: 0.0
180
+ mock_dataset_mode: false
181
+ shuffle: true
182
+ seed: 42
183
+ multiprocessing_context: fork
184
+ allow_padding: false
185
+ subsample_ratio: 1.0
186
+ image_crop_size:
187
+ - 244
188
+ - 244
189
+ image_target_size:
190
+ - 224
191
+ - 224
192
+ training:
193
+ output_dir: output/model/daru_mob_comp_minimum_0715
194
+ experiment_name: null
195
+ max_steps: 50000
196
+ global_batch_size: 24
197
+ per_gpu_batch_size: null
198
+ gradient_accumulation_steps: 1
199
+ learning_rate: 0.0001
200
+ lr_scheduler_type: cosine
201
+ weight_decay: 1.0e-05
202
+ warmup_ratio: 0.05
203
+ warmup_steps: 0
204
+ max_grad_norm: 1.0
205
+ optim: adamw_torch
206
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
207
+ skip_weight_loading: false
208
+ tf32: true
209
+ fp16: false
210
+ bf16: true
211
+ eval_bf16: true
212
+ logging_steps: 10
213
+ save_steps: 50000
214
+ save_total_limit: 3
215
+ save_vl_model: false
216
+ save_only_model: false
217
+ resume_from_checkpoint: false
218
+ upload_checkpoints: false
219
+ upload_every: 1000
220
+ upload_last_n_checkpoints: 5
221
+ max_concurrent_uploads: 2
222
+ eval_strategy: 'no'
223
+ eval_steps: 500
224
+ eval_set_split_ratio: 0.1
225
+ eval_batch_size: 2
226
+ save_best_eval_metric_name: ''
227
+ save_best_eval_metric_greater_is_better: true
228
+ deepspeed_stage: 2
229
+ gradient_checkpointing: false
230
+ transformers_trust_remote_code: true
231
+ transformers_local_files_only: false
232
+ transformers_cache_dir: null
233
+ transformers_access_token: null
234
+ use_ddp: false
235
+ ddp_bucket_cap_mb: 100
236
+ num_gpus: 1
237
+ dataloader_num_workers: 2
238
+ remove_unused_columns: false
239
+ use_wandb: true
240
+ wandb_project: finetune-gr00t-n1d7
241
+ enable_profiling: false
242
+ max_retries: 3
243
+ assert_loss_less_than: null
244
+ add_rl_callback: false
245
+ enable_open_loop_eval: false
246
+ open_loop_eval_traj_ids:
247
+ - 0
248
+ open_loop_eval_steps_per_traj: 100
249
+ open_loop_eval_plot_indices: null
250
+ max_steps: 50000
251
+ save_steps: 50000
checkpoint-50000/experiment_cfg/config.yaml ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d7
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Cosmos-Reason2-2B
6
+ backbone_model_type: qwen
7
+ model_revision: null
8
+ tune_top_llm_layers: 0
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: false
12
+ select_layer: 12
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ backbone_trainable_params_fp32: true
17
+ image_crop_size:
18
+ - 230
19
+ - 230
20
+ image_target_size:
21
+ - 256
22
+ - 256
23
+ shortest_image_edge: null
24
+ crop_fraction: null
25
+ random_rotation_angle: null
26
+ color_jitter_params: null
27
+ use_albumentations_transforms: true
28
+ letter_box_transform: false
29
+ extra_augmentation_config: null
30
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_percentiles: true
33
+ use_relative_action: true
34
+ max_state_dim: 132
35
+ max_action_dim: 132
36
+ action_horizon: 40
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ state_history_length: 1
40
+ add_pos_embed: true
41
+ attn_dropout: 0.2
42
+ use_vlln: true
43
+ max_seq_len: 1024
44
+ use_alternate_vl_dit: true
45
+ attend_text_every_n_blocks: 2
46
+ diffusion_model_cfg:
47
+ positional_embeddings: null
48
+ num_layers: 16
49
+ num_attention_heads: 32
50
+ attention_head_dim: 48
51
+ norm_type: ada_norm
52
+ dropout: 0.2
53
+ final_dropout: true
54
+ output_dim: 1024
55
+ interleave_self_attention: true
56
+ num_inference_timesteps: 4
57
+ noise_beta_alpha: 1.5
58
+ noise_beta_beta: 1.0
59
+ noise_s: 0.999
60
+ num_timestep_buckets: 1000
61
+ tune_projector: true
62
+ tune_diffusion_model: true
63
+ tune_vlln: true
64
+ state_dropout_prob: 0.2
65
+ exclude_state: false
66
+ use_mean_std: false
67
+ max_num_embodiments: 32
68
+ data:
69
+ datasets:
70
+ - dataset_paths:
71
+ - chanwoochan/daru_mob_comp_minimum_0715
72
+ embodiment_tag: new_embodiment
73
+ mix_ratio: 1.0
74
+ dataset_type: physical_embodiment
75
+ val_dataset_path: null
76
+ modality_configs:
77
+ new_embodiment:
78
+ video:
79
+ delta_indices:
80
+ - 0
81
+ modality_keys:
82
+ - head_left
83
+ - head_right
84
+ - hand_left
85
+ - hand_right
86
+ sin_cos_embedding_keys: null
87
+ mean_std_embedding_keys: null
88
+ action_configs: null
89
+ state:
90
+ delta_indices:
91
+ - 0
92
+ modality_keys:
93
+ - left_arm
94
+ - right_arm
95
+ - left_hand
96
+ - right_hand
97
+ - head
98
+ - waist
99
+ sin_cos_embedding_keys: null
100
+ mean_std_embedding_keys: null
101
+ action_configs: null
102
+ action:
103
+ delta_indices:
104
+ - 0
105
+ - 1
106
+ - 2
107
+ - 3
108
+ - 4
109
+ - 5
110
+ - 6
111
+ - 7
112
+ - 8
113
+ - 9
114
+ - 10
115
+ - 11
116
+ - 12
117
+ - 13
118
+ - 14
119
+ - 15
120
+ modality_keys:
121
+ - left_arm
122
+ - right_arm
123
+ - left_hand
124
+ - right_hand
125
+ - head
126
+ - waist
127
+ - mob_comp
128
+ - task_goal
129
+ sin_cos_embedding_keys: null
130
+ mean_std_embedding_keys: null
131
+ action_configs:
132
+ - rep: delta
133
+ type: non_eef
134
+ format: default
135
+ state_key: null
136
+ - rep: delta
137
+ type: non_eef
138
+ format: default
139
+ state_key: null
140
+ - rep: delta
141
+ type: non_eef
142
+ format: default
143
+ state_key: null
144
+ - rep: delta
145
+ type: non_eef
146
+ format: default
147
+ state_key: null
148
+ - rep: delta
149
+ type: non_eef
150
+ format: default
151
+ state_key: null
152
+ - rep: delta
153
+ type: non_eef
154
+ format: default
155
+ state_key: null
156
+ - rep: delta
157
+ type: non_eef
158
+ format: default
159
+ state_key: null
160
+ - rep: delta
161
+ type: non_eef
162
+ format: default
163
+ state_key: null
164
+ language:
165
+ delta_indices:
166
+ - 0
167
+ modality_keys:
168
+ - annotation.human.task_description
169
+ sin_cos_embedding_keys: null
170
+ mean_std_embedding_keys: null
171
+ action_configs: null
172
+ download_cache: false
173
+ shard_size: 1024
174
+ episode_sampling_rate: 0.1
175
+ num_shards_per_epoch: 100000
176
+ ds_weights_alpha: null
177
+ override_pretraining_statistics: true
178
+ mode: single_turn
179
+ random_chop: 0.0
180
+ mock_dataset_mode: false
181
+ shuffle: true
182
+ seed: 42
183
+ multiprocessing_context: fork
184
+ allow_padding: false
185
+ subsample_ratio: 1.0
186
+ image_crop_size:
187
+ - 244
188
+ - 244
189
+ image_target_size:
190
+ - 224
191
+ - 224
192
+ training:
193
+ output_dir: output/model/daru_mob_comp_minimum_0715
194
+ experiment_name: null
195
+ max_steps: 50000
196
+ global_batch_size: 24
197
+ per_gpu_batch_size: null
198
+ gradient_accumulation_steps: 1
199
+ learning_rate: 0.0001
200
+ lr_scheduler_type: cosine
201
+ weight_decay: 1.0e-05
202
+ warmup_ratio: 0.05
203
+ warmup_steps: 0
204
+ max_grad_norm: 1.0
205
+ optim: adamw_torch
206
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
207
+ skip_weight_loading: false
208
+ tf32: true
209
+ fp16: false
210
+ bf16: true
211
+ eval_bf16: true
212
+ logging_steps: 10
213
+ save_steps: 50000
214
+ save_total_limit: 3
215
+ save_vl_model: false
216
+ save_only_model: false
217
+ resume_from_checkpoint: false
218
+ upload_checkpoints: false
219
+ upload_every: 1000
220
+ upload_last_n_checkpoints: 5
221
+ max_concurrent_uploads: 2
222
+ eval_strategy: 'no'
223
+ eval_steps: 500
224
+ eval_set_split_ratio: 0.1
225
+ eval_batch_size: 2
226
+ save_best_eval_metric_name: ''
227
+ save_best_eval_metric_greater_is_better: true
228
+ deepspeed_stage: 2
229
+ gradient_checkpointing: false
230
+ transformers_trust_remote_code: true
231
+ transformers_local_files_only: false
232
+ transformers_cache_dir: null
233
+ transformers_access_token: null
234
+ use_ddp: false
235
+ ddp_bucket_cap_mb: 100
236
+ num_gpus: 1
237
+ dataloader_num_workers: 2
238
+ remove_unused_columns: false
239
+ use_wandb: true
240
+ wandb_project: finetune-gr00t-n1d7
241
+ enable_profiling: false
242
+ max_retries: 3
243
+ assert_loss_less_than: null
244
+ add_rl_callback: false
245
+ enable_open_loop_eval: false
246
+ open_loop_eval_traj_ids:
247
+ - 0
248
+ open_loop_eval_steps_per_traj: 100
249
+ open_loop_eval_plot_indices: null
checkpoint-50000/experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,595 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "left_arm": {
5
+ "min": [
6
+ -0.9032078981399536,
7
+ -0.17104226350784302,
8
+ -1.1798425912857056,
9
+ -2.3226840496063232,
10
+ -1.4744541645050049,
11
+ -0.5880014300346375,
12
+ -0.7271041870117188
13
+ ],
14
+ "max": [
15
+ 0.3963642716407776,
16
+ 0.7513642311096191,
17
+ 0.5494296550750732,
18
+ -0.6892305016517639,
19
+ 0.7433357238769531,
20
+ 0.6255260109901428,
21
+ 1.0030406713485718
22
+ ],
23
+ "mean": [
24
+ -0.2869558334350586,
25
+ 0.17532365024089813,
26
+ -0.47614753246307373,
27
+ -1.7166075706481934,
28
+ -0.18068324029445648,
29
+ 0.5105005502700806,
30
+ 0.2532295286655426
31
+ ],
32
+ "std": [
33
+ 0.19354814291000366,
34
+ 0.1314506232738495,
35
+ 0.2711745500564575,
36
+ 0.3270381689071655,
37
+ 0.39534708857536316,
38
+ 0.21731241047382355,
39
+ 0.25014322996139526
40
+ ],
41
+ "q01": [
42
+ -0.7380753207206726,
43
+ -0.12591328606009483,
44
+ -0.8992686891555786,
45
+ -2.238983335494995,
46
+ -1.4360568523406982,
47
+ -0.29329036206007003,
48
+ -0.5905199331045151
49
+ ],
50
+ "q99": [
51
+ 0.2684316337108612,
52
+ 0.590570544600487,
53
+ 0.4492477476596832,
54
+ -0.7885397672653198,
55
+ 0.5502028489112855,
56
+ 0.6190682649612427,
57
+ 0.7808114385604861
58
+ ]
59
+ },
60
+ "right_arm": {
61
+ "min": [
62
+ -1.0232865810394287,
63
+ -0.34435346722602844,
64
+ -2.2130775451660156,
65
+ -2.2280874252319336,
66
+ -0.3998549282550812,
67
+ -0.7140141725540161,
68
+ -0.07504915446043015
69
+ ],
70
+ "max": [
71
+ 0.36023595929145813,
72
+ 0.2612757980823517,
73
+ 1.5369369983673096,
74
+ -0.5520476698875427,
75
+ 1.4470524787902832,
76
+ -0.1054178848862648,
77
+ 1.0857692956924438
78
+ ],
79
+ "mean": [
80
+ -0.26461294293403625,
81
+ -0.10266675800085068,
82
+ 0.6031351089477539,
83
+ -1.7845441102981567,
84
+ 0.26721641421318054,
85
+ -0.5852886438369751,
86
+ 0.5154456496238708
87
+ ],
88
+ "std": [
89
+ 0.22050070762634277,
90
+ 0.09380312263965607,
91
+ 0.2636542320251465,
92
+ 0.27053317427635193,
93
+ 0.3317246437072754,
94
+ 0.0631183311343193,
95
+ 0.21471618115901947
96
+ ],
97
+ "q01": [
98
+ -0.7713866585493088,
99
+ -0.304285945892334,
100
+ 0.017029177658259866,
101
+ -2.16717529296875,
102
+ -0.15329226270318033,
103
+ -0.6524040699005127,
104
+ 0.19320794939994812
105
+ ],
106
+ "q99": [
107
+ 0.2689552307128906,
108
+ 0.17824872985482232,
109
+ 1.2808971405029297,
110
+ -0.9394356924295425,
111
+ 1.4196507930755615,
112
+ -0.25132742524147034,
113
+ 1.0289466321468355
114
+ ]
115
+ },
116
+ "left_hand": {
117
+ "min": [
118
+ -0.8945000171661377,
119
+ -0.1889999955892563,
120
+ -0.007000000216066837,
121
+ -0.12600000202655792,
122
+ -0.9514999985694885,
123
+ -0.9599999785423279
124
+ ],
125
+ "max": [
126
+ 0.6884999871253967,
127
+ 0.9120000004768372,
128
+ 0.9595000147819519,
129
+ 0.9120000004768372,
130
+ 0.9934999942779541,
131
+ 0.9934999942779541
132
+ ],
133
+ "mean": [
134
+ 0.1297818273305893,
135
+ 0.11738239973783493,
136
+ 0.11905454844236374,
137
+ 0.10788065940141678,
138
+ 0.09944233298301697,
139
+ 0.6733852624893188
140
+ ],
141
+ "std": [
142
+ 0.21148353815078735,
143
+ 0.21655233204364777,
144
+ 0.20444288849830627,
145
+ 0.19637329876422882,
146
+ 0.20284469425678253,
147
+ 0.2246493250131607
148
+ ],
149
+ "q01": [
150
+ -0.0010000000474974513,
151
+ -0.0005000000237487257,
152
+ 0.0,
153
+ -0.0010000000474974513,
154
+ -0.003000000026077032,
155
+ 0.07050000131130219
156
+ ],
157
+ "q99": [
158
+ 0.6449999809265137,
159
+ 0.6345000267028809,
160
+ 0.6065000295639038,
161
+ 0.5899999737739563,
162
+ 0.8355000019073486,
163
+ 0.9700000286102295
164
+ ]
165
+ },
166
+ "right_hand": {
167
+ "min": [
168
+ -0.9714999794960022,
169
+ -0.1264999955892563,
170
+ -0.6384999752044678,
171
+ -0.6334999799728394,
172
+ -0.014499999582767487,
173
+ 0.421999990940094
174
+ ],
175
+ "max": [
176
+ 0.9679999947547913,
177
+ 0.8519999980926514,
178
+ 0.8759999871253967,
179
+ 0.8190000057220459,
180
+ 0.8234999775886536,
181
+ 0.987500011920929
182
+ ],
183
+ "mean": [
184
+ 0.0794672966003418,
185
+ 0.08586394041776657,
186
+ 0.0845729261636734,
187
+ 0.07792633771896362,
188
+ 0.0825234204530716,
189
+ 0.6479356288909912
190
+ ],
191
+ "std": [
192
+ 0.18681903183460236,
193
+ 0.19215397536754608,
194
+ 0.18897227942943573,
195
+ 0.17260189354419708,
196
+ 0.18959352374076843,
197
+ 0.17591802775859833
198
+ ],
199
+ "q01": [
200
+ -0.0044999998062849045,
201
+ -0.0005000000237487257,
202
+ -0.004999999888241291,
203
+ -0.0010000000474974513,
204
+ -0.004000000189989805,
205
+ 0.5199999809265137
206
+ ],
207
+ "q99": [
208
+ 0.8370000123977661,
209
+ 0.8305000066757202,
210
+ 0.6029999852180481,
211
+ 0.5400000214576721,
212
+ 0.7455000281333923,
213
+ 0.9729999899864197
214
+ ]
215
+ },
216
+ "head": {
217
+ "min": [
218
+ -0.6981316804885864,
219
+ -0.12217304855585098
220
+ ],
221
+ "max": [
222
+ 0.8552113175392151,
223
+ 0.13962633907794952
224
+ ],
225
+ "mean": [
226
+ 0.06253354251384735,
227
+ -0.0034417936112731695
228
+ ],
229
+ "std": [
230
+ 0.30650585889816284,
231
+ 0.05744078382849693
232
+ ],
233
+ "q01": [
234
+ -0.6457718014717102,
235
+ -0.12217304855585098
236
+ ],
237
+ "q99": [
238
+ 0.7853981852531433,
239
+ 0.12217304855585098
240
+ ]
241
+ },
242
+ "waist": {
243
+ "min": [
244
+ -0.7402533292770386
245
+ ],
246
+ "max": [
247
+ 0.541886031627655
248
+ ],
249
+ "mean": [
250
+ -0.04419296979904175
251
+ ],
252
+ "std": [
253
+ 0.23224681615829468
254
+ ],
255
+ "q01": [
256
+ -0.6261920928955078
257
+ ],
258
+ "q99": [
259
+ 0.4083695709705353
260
+ ]
261
+ }
262
+ },
263
+ "action": {
264
+ "left_arm": {
265
+ "min": [
266
+ -0.8499535322189331,
267
+ -0.25394493341445923,
268
+ -1.2947511672973633,
269
+ -2.356194019317627,
270
+ -1.5707954168319702,
271
+ -0.6283181309700012,
272
+ -0.8141002058982849
273
+ ],
274
+ "max": [
275
+ 0.4256761372089386,
276
+ 0.7752004861831665,
277
+ 0.6364154815673828,
278
+ -0.5632686018943787,
279
+ 0.7934372425079346,
280
+ 0.6283181309700012,
281
+ 1.047196865081787
282
+ ],
283
+ "mean": [
284
+ -0.2504061758518219,
285
+ 0.16202592849731445,
286
+ -0.4816209673881531,
287
+ -1.684709906578064,
288
+ -0.184758722782135,
289
+ 0.5335910320281982,
290
+ 0.25200003385543823
291
+ ],
292
+ "std": [
293
+ 0.18740738928318024,
294
+ 0.13428013026714325,
295
+ 0.279079794883728,
296
+ 0.3418758809566498,
297
+ 0.4204392731189728,
298
+ 0.23198546469211578,
299
+ 0.254506379365921
300
+ ],
301
+ "q01": [
302
+ -0.7154676306247711,
303
+ -0.16645458683371545,
304
+ -0.9364646309614182,
305
+ -2.213869159221649,
306
+ -1.5707954168319702,
307
+ -0.3343559408187866,
308
+ -0.6008876466751099
309
+ ],
310
+ "q99": [
311
+ 0.2805453842878344,
312
+ 0.5947800517082216,
313
+ 0.4674780580401421,
314
+ -0.7327411317825316,
315
+ 0.5889149415493011,
316
+ 0.6283181309700012,
317
+ 0.8024686759710313
318
+ ]
319
+ },
320
+ "right_arm": {
321
+ "min": [
322
+ -1.042810082435608,
323
+ -0.3474646210670471,
324
+ -0.10725196450948715,
325
+ -2.3385281562805176,
326
+ -0.34228163957595825,
327
+ -0.6283181309700012,
328
+ -0.09903465956449509
329
+ ],
330
+ "max": [
331
+ 0.4206028878688812,
332
+ 0.3490656316280365,
333
+ 1.653267502784729,
334
+ -0.4729674756526947,
335
+ 1.5707954168319702,
336
+ -0.04136136919260025,
337
+ 1.047196865081787
338
+ ],
339
+ "mean": [
340
+ -0.2439965158700943,
341
+ -0.1011262983083725,
342
+ 0.6109874248504639,
343
+ -1.7618523836135864,
344
+ 0.2710442841053009,
345
+ -0.6060145497322083,
346
+ 0.5119675397872925
347
+ ],
348
+ "std": [
349
+ 0.22085259854793549,
350
+ 0.10317016392946243,
351
+ 0.27652469277381897,
352
+ 0.2890540063381195,
353
+ 0.3692925274372101,
354
+ 0.07110434770584106,
355
+ 0.23511721193790436
356
+ ],
357
+ "q01": [
358
+ -0.7713390624523163,
359
+ -0.3190961357951164,
360
+ 0.01663753580302,
361
+ -2.2107937860488893,
362
+ -0.21337952315807343,
363
+ -0.6283181309700012,
364
+ 0.1675674919784069
365
+ ],
366
+ "q99": [
367
+ 0.2990204724669457,
368
+ 0.23830342411994943,
369
+ 1.362928637266159,
370
+ -0.900239096879959,
371
+ 1.5707954168319702,
372
+ -0.22352868109941482,
373
+ 1.047196865081787
374
+ ]
375
+ },
376
+ "left_hand": {
377
+ "min": [
378
+ 0.0,
379
+ 0.0,
380
+ 0.0,
381
+ 0.0,
382
+ 0.0,
383
+ 0.5
384
+ ],
385
+ "max": [
386
+ 1.0,
387
+ 1.0,
388
+ 1.0,
389
+ 1.0,
390
+ 1.0,
391
+ 1.0
392
+ ],
393
+ "mean": [
394
+ 0.22313059866428375,
395
+ 0.22313059866428375,
396
+ 0.22313059866428375,
397
+ 0.22313059866428375,
398
+ 0.22313059866428375,
399
+ 0.6710783839225769
400
+ ],
401
+ "std": [
402
+ 0.41156288981437683,
403
+ 0.41156288981437683,
404
+ 0.41156288981437683,
405
+ 0.41156288981437683,
406
+ 0.41156288981437683,
407
+ 0.22488965094089508
408
+ ],
409
+ "q01": [
410
+ 0.0,
411
+ 0.0,
412
+ 0.0,
413
+ 0.0,
414
+ 0.0,
415
+ 0.5
416
+ ],
417
+ "q99": [
418
+ 1.0,
419
+ 1.0,
420
+ 1.0,
421
+ 1.0,
422
+ 1.0,
423
+ 0.9999073147773743
424
+ ]
425
+ },
426
+ "right_hand": {
427
+ "min": [
428
+ 0.0,
429
+ 0.0,
430
+ 0.0,
431
+ 0.0,
432
+ 0.0,
433
+ 0.5
434
+ ],
435
+ "max": [
436
+ 1.0,
437
+ 1.0,
438
+ 1.0,
439
+ 1.0,
440
+ 1.0,
441
+ 1.0
442
+ ],
443
+ "mean": [
444
+ 0.16685084998607635,
445
+ 0.16685084998607635,
446
+ 0.16685084998607635,
447
+ 0.16685084998607635,
448
+ 0.16685084998607635,
449
+ 0.6144992709159851
450
+ ],
451
+ "std": [
452
+ 0.36956050992012024,
453
+ 0.36956050992012024,
454
+ 0.36956050992012024,
455
+ 0.36956050992012024,
456
+ 0.36956050992012024,
457
+ 0.20083360373973846
458
+ ],
459
+ "q01": [
460
+ 0.0,
461
+ 0.0,
462
+ 0.0,
463
+ 0.0,
464
+ 0.0,
465
+ 0.5
466
+ ],
467
+ "q99": [
468
+ 1.0,
469
+ 1.0,
470
+ 1.0,
471
+ 1.0,
472
+ 1.0,
473
+ 0.999963641166687
474
+ ]
475
+ },
476
+ "head": {
477
+ "min": [
478
+ -0.7581518888473511,
479
+ -0.11258255690336227
480
+ ],
481
+ "max": [
482
+ 0.8933978080749512,
483
+ 0.1990974098443985
484
+ ],
485
+ "mean": [
486
+ 0.0644804909825325,
487
+ 0.02685731090605259
488
+ ],
489
+ "std": [
490
+ 0.3190660774707794,
491
+ 0.062273312360048294
492
+ ],
493
+ "q01": [
494
+ -0.685428461432457,
495
+ -0.10292920261621476
496
+ ],
497
+ "q99": [
498
+ 0.807757226228714,
499
+ 0.15959951758384708
500
+ ]
501
+ },
502
+ "waist": {
503
+ "min": [
504
+ -0.7449859976768494
505
+ ],
506
+ "max": [
507
+ 0.5428807735443115
508
+ ],
509
+ "mean": [
510
+ -0.04446377605199814
511
+ ],
512
+ "std": [
513
+ 0.2328842282295227
514
+ ],
515
+ "q01": [
516
+ -0.6253079611063004
517
+ ],
518
+ "q99": [
519
+ 0.41605783462524415
520
+ ]
521
+ },
522
+ "mob_comp": {
523
+ "min": [
524
+ 0.0,
525
+ 0.0,
526
+ 0.0,
527
+ 0.0,
528
+ 0.0,
529
+ 0.0
530
+ ],
531
+ "max": [
532
+ 1.0,
533
+ 0.0,
534
+ 1.0,
535
+ 0.0,
536
+ 1.0,
537
+ 1.0
538
+ ],
539
+ "mean": [
540
+ 0.03673145920038223,
541
+ 0.0,
542
+ 0.1621939092874527,
543
+ 0.0,
544
+ 0.03323742002248764,
545
+ 0.12208585441112518
546
+ ],
547
+ "std": [
548
+ 0.18808671832084656,
549
+ 0.0,
550
+ 0.36869359016418457,
551
+ 0.0,
552
+ 0.17924492061138153,
553
+ 0.32738250494003296
554
+ ],
555
+ "q01": [
556
+ 0.0,
557
+ 0.0,
558
+ 0.0,
559
+ 0.0,
560
+ 0.0,
561
+ 0.0
562
+ ],
563
+ "q99": [
564
+ 1.0,
565
+ 0.0,
566
+ 1.0,
567
+ 0.0,
568
+ 1.0,
569
+ 1.0
570
+ ]
571
+ },
572
+ "task_goal": {
573
+ "min": [
574
+ 0.0
575
+ ],
576
+ "max": [
577
+ 1.0
578
+ ],
579
+ "mean": [
580
+ 0.07475482672452927
581
+ ],
582
+ "std": [
583
+ 0.2630324363708496
584
+ ],
585
+ "q01": [
586
+ 0.0
587
+ ],
588
+ "q99": [
589
+ 1.0
590
+ ]
591
+ }
592
+ },
593
+ "relative_action": {}
594
+ }
595
+ }
checkpoint-50000/experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "Gr00tN1d7",
3
+ "model_dtype": "bfloat16",
4
+ "model_name": "nvidia/Cosmos-Reason2-2B",
5
+ "backbone_model_type": "qwen",
6
+ "model_revision": null,
7
+ "tune_top_llm_layers": 0,
8
+ "backbone_embedding_dim": 2048,
9
+ "tune_llm": false,
10
+ "tune_visual": false,
11
+ "select_layer": 16,
12
+ "reproject_vision": false,
13
+ "use_flash_attention": true,
14
+ "load_bf16": false,
15
+ "backbone_trainable_params_fp32": true,
16
+ "letter_box_transform": false,
17
+ "extra_augmentation_config": null,
18
+ "apply_sincos_state_encoding": false,
19
+ "use_percentiles": true,
20
+ "use_relative_action": false,
21
+ "max_state_dim": 132,
22
+ "max_action_dim": 132,
23
+ "action_horizon": 40,
24
+ "hidden_size": 1024,
25
+ "input_embedding_dim": 1536,
26
+ "state_history_length": 1,
27
+ "add_pos_embed": true,
28
+ "attn_dropout": 0.2,
29
+ "use_vlln": true,
30
+ "max_seq_len": 1024,
31
+ "use_alternate_vl_dit": true,
32
+ "attend_text_every_n_blocks": 2,
33
+ "diffusion_model_cfg": {
34
+ "attention_head_dim": 48,
35
+ "dropout": 0.2,
36
+ "final_dropout": true,
37
+ "interleave_self_attention": true,
38
+ "norm_type": "ada_norm",
39
+ "num_attention_heads": 32,
40
+ "num_layers": 32,
41
+ "output_dim": 1024,
42
+ "positional_embeddings": null
43
+ },
44
+ "num_inference_timesteps": 4,
45
+ "noise_beta_alpha": 1.5,
46
+ "noise_beta_beta": 1.0,
47
+ "noise_s": 0.999,
48
+ "num_timestep_buckets": 1000,
49
+ "tune_projector": true,
50
+ "tune_diffusion_model": true,
51
+ "tune_vlln": true,
52
+ "state_dropout_prob": 0.2,
53
+ "exclude_state": false,
54
+ "use_mean_std": false,
55
+ "max_num_embodiments": 32
56
+ }
checkpoint-50000/experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-50000/model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbcaea5ee88f1e0f1465043920a2647c67e7de17d24adfd1c477742a6168edec
3
+ size 4986649584
checkpoint-50000/model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca72f26be25f0bf67d1e8c7d8cccdb1a1efb5e9fcc623262f900dd72ea924b52
3
+ size 4970792616
checkpoint-50000/model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ebf7872ce1490869f9660b4c00750db30840739bfb020bb7ad0718e297a4992
3
+ size 2618758696
checkpoint-50000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-50000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b7cadf4383a082db1b6cd58acbd6879b9d54d3d63884c09284db0cf846efd3f
3
+ size 12964594710
checkpoint-50000/processor_config.json ADDED
@@ -0,0 +1,1171 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "processor_class": "Gr00tN1d7Processor",
3
+ "processor_kwargs": {
4
+ "modality_configs": {
5
+ "real_g1_relative_eef_relative_joints": {
6
+ "video": {
7
+ "delta_indices": [
8
+ -20,
9
+ 0
10
+ ],
11
+ "modality_keys": [
12
+ "ego_view"
13
+ ],
14
+ "sin_cos_embedding_keys": null,
15
+ "mean_std_embedding_keys": null,
16
+ "action_configs": null
17
+ },
18
+ "state": {
19
+ "delta_indices": [
20
+ 0
21
+ ],
22
+ "modality_keys": [
23
+ "left_wrist_eef_9d",
24
+ "right_wrist_eef_9d",
25
+ "left_hand",
26
+ "right_hand",
27
+ "left_arm",
28
+ "right_arm",
29
+ "waist"
30
+ ],
31
+ "sin_cos_embedding_keys": null,
32
+ "mean_std_embedding_keys": null,
33
+ "action_configs": null
34
+ },
35
+ "action": {
36
+ "delta_indices": [
37
+ 0,
38
+ 1,
39
+ 2,
40
+ 3,
41
+ 4,
42
+ 5,
43
+ 6,
44
+ 7,
45
+ 8,
46
+ 9,
47
+ 10,
48
+ 11,
49
+ 12,
50
+ 13,
51
+ 14,
52
+ 15,
53
+ 16,
54
+ 17,
55
+ 18,
56
+ 19,
57
+ 20,
58
+ 21,
59
+ 22,
60
+ 23,
61
+ 24,
62
+ 25,
63
+ 26,
64
+ 27,
65
+ 28,
66
+ 29,
67
+ 30,
68
+ 31,
69
+ 32,
70
+ 33,
71
+ 34,
72
+ 35,
73
+ 36,
74
+ 37,
75
+ 38,
76
+ 39
77
+ ],
78
+ "modality_keys": [
79
+ "left_wrist_eef_9d",
80
+ "right_wrist_eef_9d",
81
+ "left_hand",
82
+ "right_hand",
83
+ "left_arm",
84
+ "right_arm",
85
+ "waist",
86
+ "base_height_command",
87
+ "navigate_command"
88
+ ],
89
+ "sin_cos_embedding_keys": null,
90
+ "mean_std_embedding_keys": null,
91
+ "action_configs": [
92
+ {
93
+ "rep": "RELATIVE",
94
+ "type": "EEF",
95
+ "format": "XYZ_ROT6D",
96
+ "state_key": "left_wrist_eef_9d"
97
+ },
98
+ {
99
+ "rep": "RELATIVE",
100
+ "type": "EEF",
101
+ "format": "XYZ_ROT6D",
102
+ "state_key": "right_wrist_eef_9d"
103
+ },
104
+ {
105
+ "rep": "ABSOLUTE",
106
+ "type": "NON_EEF",
107
+ "format": "DEFAULT",
108
+ "state_key": "left_hand"
109
+ },
110
+ {
111
+ "rep": "ABSOLUTE",
112
+ "type": "NON_EEF",
113
+ "format": "DEFAULT",
114
+ "state_key": "right_hand"
115
+ },
116
+ {
117
+ "rep": "RELATIVE",
118
+ "type": "NON_EEF",
119
+ "format": "DEFAULT",
120
+ "state_key": "left_arm"
121
+ },
122
+ {
123
+ "rep": "RELATIVE",
124
+ "type": "NON_EEF",
125
+ "format": "DEFAULT",
126
+ "state_key": "right_arm"
127
+ },
128
+ {
129
+ "rep": "ABSOLUTE",
130
+ "type": "NON_EEF",
131
+ "format": "DEFAULT",
132
+ "state_key": "waist"
133
+ },
134
+ {
135
+ "rep": "ABSOLUTE",
136
+ "type": "NON_EEF",
137
+ "format": "DEFAULT",
138
+ "state_key": "base_height_command"
139
+ },
140
+ {
141
+ "rep": "ABSOLUTE",
142
+ "type": "NON_EEF",
143
+ "format": "DEFAULT",
144
+ "state_key": "navigate_command"
145
+ }
146
+ ]
147
+ },
148
+ "language": {
149
+ "delta_indices": [
150
+ 0
151
+ ],
152
+ "modality_keys": [
153
+ "annotation.human.task_description"
154
+ ],
155
+ "sin_cos_embedding_keys": null,
156
+ "mean_std_embedding_keys": null,
157
+ "action_configs": null
158
+ }
159
+ },
160
+ "real_r1_pro_sharpa_relative_eef_mecka": {
161
+ "video": {
162
+ "delta_indices": [
163
+ -30,
164
+ 0
165
+ ],
166
+ "modality_keys": [
167
+ "ego_view_cropratio_res320x240_freq30"
168
+ ],
169
+ "sin_cos_embedding_keys": null,
170
+ "mean_std_embedding_keys": null,
171
+ "action_configs": null
172
+ },
173
+ "state": {
174
+ "delta_indices": [
175
+ 0
176
+ ],
177
+ "modality_keys": [
178
+ "left_wrist_eef",
179
+ "right_wrist_eef",
180
+ "left_hand_joints",
181
+ "right_hand_joints"
182
+ ],
183
+ "sin_cos_embedding_keys": null,
184
+ "mean_std_embedding_keys": null,
185
+ "action_configs": null
186
+ },
187
+ "action": {
188
+ "delta_indices": [
189
+ 0,
190
+ 1,
191
+ 2,
192
+ 3,
193
+ 4,
194
+ 5,
195
+ 6,
196
+ 7,
197
+ 8,
198
+ 9,
199
+ 10,
200
+ 11,
201
+ 12,
202
+ 13,
203
+ 14,
204
+ 15,
205
+ 16,
206
+ 17,
207
+ 18,
208
+ 19,
209
+ 20,
210
+ 21,
211
+ 22,
212
+ 23,
213
+ 24,
214
+ 25,
215
+ 26,
216
+ 27,
217
+ 28,
218
+ 29,
219
+ 30,
220
+ 31,
221
+ 32,
222
+ 33,
223
+ 34,
224
+ 35,
225
+ 36,
226
+ 37,
227
+ 38,
228
+ 39
229
+ ],
230
+ "modality_keys": [
231
+ "left_wrist_eef",
232
+ "right_wrist_eef",
233
+ "left_hand_joints",
234
+ "right_hand_joints"
235
+ ],
236
+ "sin_cos_embedding_keys": null,
237
+ "mean_std_embedding_keys": null,
238
+ "action_configs": [
239
+ {
240
+ "rep": "RELATIVE",
241
+ "type": "EEF",
242
+ "format": "XYZ_ROT6D",
243
+ "state_key": "left_wrist_eef"
244
+ },
245
+ {
246
+ "rep": "RELATIVE",
247
+ "type": "EEF",
248
+ "format": "XYZ_ROT6D",
249
+ "state_key": "right_wrist_eef"
250
+ },
251
+ {
252
+ "rep": "ABSOLUTE",
253
+ "type": "NON_EEF",
254
+ "format": "DEFAULT",
255
+ "state_key": "left_hand_joints"
256
+ },
257
+ {
258
+ "rep": "ABSOLUTE",
259
+ "type": "NON_EEF",
260
+ "format": "DEFAULT",
261
+ "state_key": "right_hand_joints"
262
+ }
263
+ ]
264
+ },
265
+ "language": {
266
+ "delta_indices": [
267
+ 0
268
+ ],
269
+ "modality_keys": [
270
+ "annotation.human.coarse_action"
271
+ ],
272
+ "sin_cos_embedding_keys": null,
273
+ "mean_std_embedding_keys": null,
274
+ "action_configs": null
275
+ }
276
+ },
277
+ "real_r1_pro_sharpa_relative_eef_human": {
278
+ "video": {
279
+ "delta_indices": [
280
+ -20,
281
+ 0
282
+ ],
283
+ "modality_keys": [
284
+ "ego_view_res320x240_freq20",
285
+ "left_wrist_view_res320x240_freq20",
286
+ "right_wrist_view_res320x240_freq20"
287
+ ],
288
+ "sin_cos_embedding_keys": null,
289
+ "mean_std_embedding_keys": null,
290
+ "action_configs": null
291
+ },
292
+ "state": {
293
+ "delta_indices": [
294
+ 0
295
+ ],
296
+ "modality_keys": [
297
+ "left_wrist_eef",
298
+ "right_wrist_eef",
299
+ "left_hand_joints",
300
+ "right_hand_joints"
301
+ ],
302
+ "sin_cos_embedding_keys": null,
303
+ "mean_std_embedding_keys": null,
304
+ "action_configs": null
305
+ },
306
+ "action": {
307
+ "delta_indices": [
308
+ 0,
309
+ 1,
310
+ 2,
311
+ 3,
312
+ 4,
313
+ 5,
314
+ 6,
315
+ 7,
316
+ 8,
317
+ 9,
318
+ 10,
319
+ 11,
320
+ 12,
321
+ 13,
322
+ 14,
323
+ 15,
324
+ 16,
325
+ 17,
326
+ 18,
327
+ 19,
328
+ 20,
329
+ 21,
330
+ 22,
331
+ 23,
332
+ 24,
333
+ 25,
334
+ 26,
335
+ 27,
336
+ 28,
337
+ 29,
338
+ 30,
339
+ 31,
340
+ 32,
341
+ 33,
342
+ 34,
343
+ 35,
344
+ 36,
345
+ 37,
346
+ 38,
347
+ 39
348
+ ],
349
+ "modality_keys": [
350
+ "left_wrist_eef",
351
+ "right_wrist_eef",
352
+ "left_hand_joints",
353
+ "right_hand_joints"
354
+ ],
355
+ "sin_cos_embedding_keys": null,
356
+ "mean_std_embedding_keys": null,
357
+ "action_configs": [
358
+ {
359
+ "rep": "RELATIVE",
360
+ "type": "EEF",
361
+ "format": "XYZ_ROT6D",
362
+ "state_key": "left_wrist_eef"
363
+ },
364
+ {
365
+ "rep": "RELATIVE",
366
+ "type": "EEF",
367
+ "format": "XYZ_ROT6D",
368
+ "state_key": "right_wrist_eef"
369
+ },
370
+ {
371
+ "rep": "ABSOLUTE",
372
+ "type": "NON_EEF",
373
+ "format": "DEFAULT",
374
+ "state_key": "left_hand_joints"
375
+ },
376
+ {
377
+ "rep": "ABSOLUTE",
378
+ "type": "NON_EEF",
379
+ "format": "DEFAULT",
380
+ "state_key": "right_hand_joints"
381
+ }
382
+ ]
383
+ },
384
+ "language": {
385
+ "delta_indices": [
386
+ 0
387
+ ],
388
+ "modality_keys": [
389
+ "annotation.human.coarse_action"
390
+ ],
391
+ "sin_cos_embedding_keys": null,
392
+ "mean_std_embedding_keys": null,
393
+ "action_configs": null
394
+ }
395
+ },
396
+ "real_r1_pro_sharpa_relative_eef": {
397
+ "video": {
398
+ "delta_indices": [
399
+ -20,
400
+ 0
401
+ ],
402
+ "modality_keys": [
403
+ "ego_view_res320x240_freq20",
404
+ "left_wrist_view_res320x240_freq20",
405
+ "right_wrist_view_res320x240_freq20"
406
+ ],
407
+ "sin_cos_embedding_keys": null,
408
+ "mean_std_embedding_keys": null,
409
+ "action_configs": null
410
+ },
411
+ "state": {
412
+ "delta_indices": [
413
+ 0
414
+ ],
415
+ "modality_keys": [
416
+ "left_wrist_eef",
417
+ "right_wrist_eef",
418
+ "left_hand_joints",
419
+ "right_hand_joints"
420
+ ],
421
+ "sin_cos_embedding_keys": null,
422
+ "mean_std_embedding_keys": null,
423
+ "action_configs": null
424
+ },
425
+ "action": {
426
+ "delta_indices": [
427
+ 0,
428
+ 1,
429
+ 2,
430
+ 3,
431
+ 4,
432
+ 5,
433
+ 6,
434
+ 7,
435
+ 8,
436
+ 9,
437
+ 10,
438
+ 11,
439
+ 12,
440
+ 13,
441
+ 14,
442
+ 15,
443
+ 16,
444
+ 17,
445
+ 18,
446
+ 19,
447
+ 20,
448
+ 21,
449
+ 22,
450
+ 23,
451
+ 24,
452
+ 25,
453
+ 26,
454
+ 27,
455
+ 28,
456
+ 29,
457
+ 30,
458
+ 31,
459
+ 32,
460
+ 33,
461
+ 34,
462
+ 35,
463
+ 36,
464
+ 37,
465
+ 38,
466
+ 39
467
+ ],
468
+ "modality_keys": [
469
+ "left_wrist_eef",
470
+ "right_wrist_eef",
471
+ "left_hand_joints",
472
+ "right_hand_joints"
473
+ ],
474
+ "sin_cos_embedding_keys": null,
475
+ "mean_std_embedding_keys": null,
476
+ "action_configs": [
477
+ {
478
+ "rep": "RELATIVE",
479
+ "type": "EEF",
480
+ "format": "XYZ_ROT6D",
481
+ "state_key": "left_wrist_eef"
482
+ },
483
+ {
484
+ "rep": "RELATIVE",
485
+ "type": "EEF",
486
+ "format": "XYZ_ROT6D",
487
+ "state_key": "right_wrist_eef"
488
+ },
489
+ {
490
+ "rep": "ABSOLUTE",
491
+ "type": "NON_EEF",
492
+ "format": "DEFAULT",
493
+ "state_key": "left_hand_joints"
494
+ },
495
+ {
496
+ "rep": "ABSOLUTE",
497
+ "type": "NON_EEF",
498
+ "format": "DEFAULT",
499
+ "state_key": "right_hand_joints"
500
+ }
501
+ ]
502
+ },
503
+ "language": {
504
+ "delta_indices": [
505
+ 0
506
+ ],
507
+ "modality_keys": [
508
+ "annotation.human.coarse_action"
509
+ ],
510
+ "sin_cos_embedding_keys": null,
511
+ "mean_std_embedding_keys": null,
512
+ "action_configs": null
513
+ }
514
+ },
515
+ "xdof_relative_eef_relative_joint": {
516
+ "video": {
517
+ "delta_indices": [
518
+ -30,
519
+ 0
520
+ ],
521
+ "modality_keys": [
522
+ "top_camera-images-rgb_320_240",
523
+ "left_camera-images-rgb_320_240",
524
+ "right_camera-images-rgb_320_240"
525
+ ],
526
+ "sin_cos_embedding_keys": null,
527
+ "mean_std_embedding_keys": null,
528
+ "action_configs": null
529
+ },
530
+ "state": {
531
+ "delta_indices": [
532
+ 0
533
+ ],
534
+ "modality_keys": [
535
+ "left_wrist_eef",
536
+ "right_wrist_eef",
537
+ "left_gripper_pos",
538
+ "right_gripper_pos",
539
+ "left_joint_pos",
540
+ "right_joint_pos"
541
+ ],
542
+ "sin_cos_embedding_keys": null,
543
+ "mean_std_embedding_keys": null,
544
+ "action_configs": null
545
+ },
546
+ "action": {
547
+ "delta_indices": [
548
+ 0,
549
+ 1,
550
+ 2,
551
+ 3,
552
+ 4,
553
+ 5,
554
+ 6,
555
+ 7,
556
+ 8,
557
+ 9,
558
+ 10,
559
+ 11,
560
+ 12,
561
+ 13,
562
+ 14,
563
+ 15,
564
+ 16,
565
+ 17,
566
+ 18,
567
+ 19,
568
+ 20,
569
+ 21,
570
+ 22,
571
+ 23,
572
+ 24,
573
+ 25,
574
+ 26,
575
+ 27,
576
+ 28,
577
+ 29,
578
+ 30,
579
+ 31,
580
+ 32,
581
+ 33,
582
+ 34,
583
+ 35,
584
+ 36,
585
+ 37,
586
+ 38,
587
+ 39
588
+ ],
589
+ "modality_keys": [
590
+ "left_wrist_eef",
591
+ "right_wrist_eef",
592
+ "left_gripper_pos",
593
+ "right_gripper_pos",
594
+ "left_joint_pos",
595
+ "right_joint_pos"
596
+ ],
597
+ "sin_cos_embedding_keys": null,
598
+ "mean_std_embedding_keys": null,
599
+ "action_configs": [
600
+ {
601
+ "rep": "RELATIVE",
602
+ "type": "EEF",
603
+ "format": "XYZ_ROT6D",
604
+ "state_key": "left_wrist_eef"
605
+ },
606
+ {
607
+ "rep": "RELATIVE",
608
+ "type": "EEF",
609
+ "format": "XYZ_ROT6D",
610
+ "state_key": "right_wrist_eef"
611
+ },
612
+ {
613
+ "rep": "ABSOLUTE",
614
+ "type": "NON_EEF",
615
+ "format": "DEFAULT",
616
+ "state_key": "left_gripper_pos"
617
+ },
618
+ {
619
+ "rep": "ABSOLUTE",
620
+ "type": "NON_EEF",
621
+ "format": "DEFAULT",
622
+ "state_key": "right_gripper_pos"
623
+ },
624
+ {
625
+ "rep": "RELATIVE",
626
+ "type": "NON_EEF",
627
+ "format": "DEFAULT",
628
+ "state_key": "left_joint_pos"
629
+ },
630
+ {
631
+ "rep": "RELATIVE",
632
+ "type": "NON_EEF",
633
+ "format": "DEFAULT",
634
+ "state_key": "right_joint_pos"
635
+ }
636
+ ]
637
+ },
638
+ "language": {
639
+ "delta_indices": [
640
+ 0
641
+ ],
642
+ "modality_keys": [
643
+ "annotation.task"
644
+ ],
645
+ "sin_cos_embedding_keys": null,
646
+ "mean_std_embedding_keys": null,
647
+ "action_configs": null
648
+ }
649
+ },
650
+ "real_r1_pro_sharpa_relative_eef_maxinsights": {
651
+ "video": {
652
+ "delta_indices": [
653
+ -30,
654
+ 0
655
+ ],
656
+ "modality_keys": [
657
+ "ego_view_cropratio_res320x240_freq30"
658
+ ],
659
+ "sin_cos_embedding_keys": null,
660
+ "mean_std_embedding_keys": null,
661
+ "action_configs": null
662
+ },
663
+ "state": {
664
+ "delta_indices": [
665
+ 0
666
+ ],
667
+ "modality_keys": [
668
+ "left_wrist_eef",
669
+ "right_wrist_eef",
670
+ "left_hand_joints",
671
+ "right_hand_joints"
672
+ ],
673
+ "sin_cos_embedding_keys": null,
674
+ "mean_std_embedding_keys": null,
675
+ "action_configs": null
676
+ },
677
+ "action": {
678
+ "delta_indices": [
679
+ 0,
680
+ 1,
681
+ 2,
682
+ 3,
683
+ 4,
684
+ 5,
685
+ 6,
686
+ 7,
687
+ 8,
688
+ 9,
689
+ 10,
690
+ 11,
691
+ 12,
692
+ 13,
693
+ 14,
694
+ 15,
695
+ 16,
696
+ 17,
697
+ 18,
698
+ 19,
699
+ 20,
700
+ 21,
701
+ 22,
702
+ 23,
703
+ 24,
704
+ 25,
705
+ 26,
706
+ 27,
707
+ 28,
708
+ 29,
709
+ 30,
710
+ 31,
711
+ 32,
712
+ 33,
713
+ 34,
714
+ 35,
715
+ 36,
716
+ 37,
717
+ 38,
718
+ 39
719
+ ],
720
+ "modality_keys": [
721
+ "left_wrist_eef",
722
+ "right_wrist_eef",
723
+ "left_hand_joints",
724
+ "right_hand_joints"
725
+ ],
726
+ "sin_cos_embedding_keys": null,
727
+ "mean_std_embedding_keys": null,
728
+ "action_configs": [
729
+ {
730
+ "rep": "RELATIVE",
731
+ "type": "EEF",
732
+ "format": "XYZ_ROT6D",
733
+ "state_key": "left_wrist_eef"
734
+ },
735
+ {
736
+ "rep": "RELATIVE",
737
+ "type": "EEF",
738
+ "format": "XYZ_ROT6D",
739
+ "state_key": "right_wrist_eef"
740
+ },
741
+ {
742
+ "rep": "ABSOLUTE",
743
+ "type": "NON_EEF",
744
+ "format": "DEFAULT",
745
+ "state_key": "left_hand_joints"
746
+ },
747
+ {
748
+ "rep": "ABSOLUTE",
749
+ "type": "NON_EEF",
750
+ "format": "DEFAULT",
751
+ "state_key": "right_hand_joints"
752
+ }
753
+ ]
754
+ },
755
+ "language": {
756
+ "delta_indices": [
757
+ 0
758
+ ],
759
+ "modality_keys": [
760
+ "annotation.human.coarse_action"
761
+ ],
762
+ "sin_cos_embedding_keys": null,
763
+ "mean_std_embedding_keys": null,
764
+ "action_configs": null
765
+ }
766
+ },
767
+ "xdof_relative_eef_relative_joint_subtask": {
768
+ "video": {
769
+ "delta_indices": [
770
+ -30,
771
+ 0
772
+ ],
773
+ "modality_keys": [
774
+ "top_camera-images-rgb_320_240",
775
+ "left_camera-images-rgb_320_240",
776
+ "right_camera-images-rgb_320_240"
777
+ ],
778
+ "sin_cos_embedding_keys": null,
779
+ "mean_std_embedding_keys": null,
780
+ "action_configs": null
781
+ },
782
+ "state": {
783
+ "delta_indices": [
784
+ 0
785
+ ],
786
+ "modality_keys": [
787
+ "left_wrist_eef",
788
+ "right_wrist_eef",
789
+ "left_gripper_pos",
790
+ "right_gripper_pos",
791
+ "left_joint_pos",
792
+ "right_joint_pos"
793
+ ],
794
+ "sin_cos_embedding_keys": null,
795
+ "mean_std_embedding_keys": null,
796
+ "action_configs": null
797
+ },
798
+ "action": {
799
+ "delta_indices": [
800
+ 0,
801
+ 1,
802
+ 2,
803
+ 3,
804
+ 4,
805
+ 5,
806
+ 6,
807
+ 7,
808
+ 8,
809
+ 9,
810
+ 10,
811
+ 11,
812
+ 12,
813
+ 13,
814
+ 14,
815
+ 15,
816
+ 16,
817
+ 17,
818
+ 18,
819
+ 19,
820
+ 20,
821
+ 21,
822
+ 22,
823
+ 23,
824
+ 24,
825
+ 25,
826
+ 26,
827
+ 27,
828
+ 28,
829
+ 29,
830
+ 30,
831
+ 31,
832
+ 32,
833
+ 33,
834
+ 34,
835
+ 35,
836
+ 36,
837
+ 37,
838
+ 38,
839
+ 39
840
+ ],
841
+ "modality_keys": [
842
+ "left_wrist_eef",
843
+ "right_wrist_eef",
844
+ "left_gripper_pos",
845
+ "right_gripper_pos",
846
+ "left_joint_pos",
847
+ "right_joint_pos"
848
+ ],
849
+ "sin_cos_embedding_keys": null,
850
+ "mean_std_embedding_keys": null,
851
+ "action_configs": [
852
+ {
853
+ "rep": "RELATIVE",
854
+ "type": "EEF",
855
+ "format": "XYZ_ROT6D",
856
+ "state_key": "left_wrist_eef"
857
+ },
858
+ {
859
+ "rep": "RELATIVE",
860
+ "type": "EEF",
861
+ "format": "XYZ_ROT6D",
862
+ "state_key": "right_wrist_eef"
863
+ },
864
+ {
865
+ "rep": "ABSOLUTE",
866
+ "type": "NON_EEF",
867
+ "format": "DEFAULT",
868
+ "state_key": "left_gripper_pos"
869
+ },
870
+ {
871
+ "rep": "ABSOLUTE",
872
+ "type": "NON_EEF",
873
+ "format": "DEFAULT",
874
+ "state_key": "right_gripper_pos"
875
+ },
876
+ {
877
+ "rep": "RELATIVE",
878
+ "type": "NON_EEF",
879
+ "format": "DEFAULT",
880
+ "state_key": "left_joint_pos"
881
+ },
882
+ {
883
+ "rep": "RELATIVE",
884
+ "type": "NON_EEF",
885
+ "format": "DEFAULT",
886
+ "state_key": "right_joint_pos"
887
+ }
888
+ ]
889
+ },
890
+ "language": {
891
+ "delta_indices": [
892
+ 0
893
+ ],
894
+ "modality_keys": [
895
+ "annotation.sub_task"
896
+ ],
897
+ "sin_cos_embedding_keys": null,
898
+ "mean_std_embedding_keys": null,
899
+ "action_configs": null
900
+ }
901
+ },
902
+ "oxe_droid_relative_eef_relative_joint": {
903
+ "video": {
904
+ "delta_indices": [
905
+ -15,
906
+ 0
907
+ ],
908
+ "modality_keys": [
909
+ "exterior_image_1_left",
910
+ "wrist_image_left"
911
+ ],
912
+ "sin_cos_embedding_keys": null,
913
+ "mean_std_embedding_keys": null,
914
+ "action_configs": null
915
+ },
916
+ "state": {
917
+ "delta_indices": [
918
+ 0
919
+ ],
920
+ "modality_keys": [
921
+ "eef_9d",
922
+ "gripper_position",
923
+ "joint_position"
924
+ ],
925
+ "sin_cos_embedding_keys": null,
926
+ "mean_std_embedding_keys": null,
927
+ "action_configs": null
928
+ },
929
+ "action": {
930
+ "delta_indices": [
931
+ 0,
932
+ 1,
933
+ 2,
934
+ 3,
935
+ 4,
936
+ 5,
937
+ 6,
938
+ 7,
939
+ 8,
940
+ 9,
941
+ 10,
942
+ 11,
943
+ 12,
944
+ 13,
945
+ 14,
946
+ 15,
947
+ 16,
948
+ 17,
949
+ 18,
950
+ 19,
951
+ 20,
952
+ 21,
953
+ 22,
954
+ 23,
955
+ 24,
956
+ 25,
957
+ 26,
958
+ 27,
959
+ 28,
960
+ 29,
961
+ 30,
962
+ 31,
963
+ 32,
964
+ 33,
965
+ 34,
966
+ 35,
967
+ 36,
968
+ 37,
969
+ 38,
970
+ 39
971
+ ],
972
+ "modality_keys": [
973
+ "eef_9d",
974
+ "gripper_position",
975
+ "joint_position"
976
+ ],
977
+ "sin_cos_embedding_keys": null,
978
+ "mean_std_embedding_keys": null,
979
+ "action_configs": [
980
+ {
981
+ "rep": "RELATIVE",
982
+ "type": "EEF",
983
+ "format": "XYZ_ROT6D",
984
+ "state_key": "eef_9d"
985
+ },
986
+ {
987
+ "rep": "ABSOLUTE",
988
+ "type": "NON_EEF",
989
+ "format": "DEFAULT",
990
+ "state_key": "gripper_position"
991
+ },
992
+ {
993
+ "rep": "RELATIVE",
994
+ "type": "NON_EEF",
995
+ "format": "DEFAULT",
996
+ "state_key": "joint_position"
997
+ }
998
+ ]
999
+ },
1000
+ "language": {
1001
+ "delta_indices": [
1002
+ 0
1003
+ ],
1004
+ "modality_keys": [
1005
+ "annotation.language.language_instruction"
1006
+ ],
1007
+ "sin_cos_embedding_keys": null,
1008
+ "mean_std_embedding_keys": null,
1009
+ "action_configs": null
1010
+ }
1011
+ },
1012
+ "new_embodiment": {
1013
+ "video": {
1014
+ "delta_indices": [
1015
+ 0
1016
+ ],
1017
+ "modality_keys": [
1018
+ "head_left",
1019
+ "head_right",
1020
+ "hand_left",
1021
+ "hand_right"
1022
+ ],
1023
+ "sin_cos_embedding_keys": null,
1024
+ "mean_std_embedding_keys": null,
1025
+ "action_configs": null
1026
+ },
1027
+ "state": {
1028
+ "delta_indices": [
1029
+ 0
1030
+ ],
1031
+ "modality_keys": [
1032
+ "left_arm",
1033
+ "right_arm",
1034
+ "left_hand",
1035
+ "right_hand",
1036
+ "head",
1037
+ "waist"
1038
+ ],
1039
+ "sin_cos_embedding_keys": null,
1040
+ "mean_std_embedding_keys": null,
1041
+ "action_configs": null
1042
+ },
1043
+ "action": {
1044
+ "delta_indices": [
1045
+ 0,
1046
+ 1,
1047
+ 2,
1048
+ 3,
1049
+ 4,
1050
+ 5,
1051
+ 6,
1052
+ 7,
1053
+ 8,
1054
+ 9,
1055
+ 10,
1056
+ 11,
1057
+ 12,
1058
+ 13,
1059
+ 14,
1060
+ 15
1061
+ ],
1062
+ "modality_keys": [
1063
+ "left_arm",
1064
+ "right_arm",
1065
+ "left_hand",
1066
+ "right_hand",
1067
+ "head",
1068
+ "waist",
1069
+ "mob_comp",
1070
+ "task_goal"
1071
+ ],
1072
+ "sin_cos_embedding_keys": null,
1073
+ "mean_std_embedding_keys": null,
1074
+ "action_configs": [
1075
+ {
1076
+ "rep": "DELTA",
1077
+ "type": "NON_EEF",
1078
+ "format": "DEFAULT",
1079
+ "state_key": null
1080
+ },
1081
+ {
1082
+ "rep": "DELTA",
1083
+ "type": "NON_EEF",
1084
+ "format": "DEFAULT",
1085
+ "state_key": null
1086
+ },
1087
+ {
1088
+ "rep": "DELTA",
1089
+ "type": "NON_EEF",
1090
+ "format": "DEFAULT",
1091
+ "state_key": null
1092
+ },
1093
+ {
1094
+ "rep": "DELTA",
1095
+ "type": "NON_EEF",
1096
+ "format": "DEFAULT",
1097
+ "state_key": null
1098
+ },
1099
+ {
1100
+ "rep": "DELTA",
1101
+ "type": "NON_EEF",
1102
+ "format": "DEFAULT",
1103
+ "state_key": null
1104
+ },
1105
+ {
1106
+ "rep": "DELTA",
1107
+ "type": "NON_EEF",
1108
+ "format": "DEFAULT",
1109
+ "state_key": null
1110
+ },
1111
+ {
1112
+ "rep": "DELTA",
1113
+ "type": "NON_EEF",
1114
+ "format": "DEFAULT",
1115
+ "state_key": null
1116
+ },
1117
+ {
1118
+ "rep": "DELTA",
1119
+ "type": "NON_EEF",
1120
+ "format": "DEFAULT",
1121
+ "state_key": null
1122
+ }
1123
+ ]
1124
+ },
1125
+ "language": {
1126
+ "delta_indices": [
1127
+ 0
1128
+ ],
1129
+ "modality_keys": [
1130
+ "annotation.human.task_description"
1131
+ ],
1132
+ "sin_cos_embedding_keys": null,
1133
+ "mean_std_embedding_keys": null,
1134
+ "action_configs": null
1135
+ }
1136
+ }
1137
+ },
1138
+ "image_crop_size": [
1139
+ 230,
1140
+ 230
1141
+ ],
1142
+ "image_target_size": [
1143
+ 256,
1144
+ 256
1145
+ ],
1146
+ "use_albumentations": true,
1147
+ "random_rotation_angle": 0,
1148
+ "color_jitter_params": {
1149
+ "brightness": 0.3,
1150
+ "contrast": 0.4,
1151
+ "saturation": 0.5,
1152
+ "hue": 0.08
1153
+ },
1154
+ "shortest_image_edge": 256,
1155
+ "crop_fraction": 0.95,
1156
+ "letter_box_transform": false,
1157
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1158
+ "model_type": "qwen",
1159
+ "formalize_language": true,
1160
+ "max_state_dim": 132,
1161
+ "max_action_dim": 132,
1162
+ "max_action_horizon": 40,
1163
+ "use_percentiles": true,
1164
+ "use_mean_std": false,
1165
+ "clip_outliers": true,
1166
+ "apply_sincos_state_encoding": false,
1167
+ "use_relative_action": true,
1168
+ "exclude_state": false,
1169
+ "state_dropout_prob": 0.2
1170
+ }
1171
+ }
checkpoint-50000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:713ad76d9ffea3b4599bf7f33dd60940baeca3848f584e4053cb6d4be5f2b3aa
3
+ size 14645
checkpoint-50000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e2d10041ab6fabfcdf847f500f1a0149161e21c91e14cb0d5946a9abf608723
3
+ size 1465
checkpoint-50000/statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-50000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-50000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c06c6312892d4f8856e3cf36eda5c6087283cb9454423c024dd03521616ec9db
3
+ size 5841
checkpoint-50000/wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d7", "run_id": "daru_mob_comp_minimum_0715"}
config.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_horizon": 40,
3
+ "add_pos_embed": true,
4
+ "apply_sincos_state_encoding": false,
5
+ "architectures": [
6
+ "Gr00tN1d7"
7
+ ],
8
+ "attn_dropout": 0.2,
9
+ "attn_implementation": null,
10
+ "backbone_embedding_dim": 2048,
11
+ "backbone_trainable_params_fp32": true,
12
+ "color_jitter_params": {
13
+ "brightness": 0.3,
14
+ "contrast": 0.4,
15
+ "hue": 0.08,
16
+ "saturation": 0.5
17
+ },
18
+ "crop_fraction": 0.95,
19
+ "diffusion_model_cfg": {
20
+ "attention_head_dim": 48,
21
+ "dropout": 0.2,
22
+ "final_dropout": true,
23
+ "interleave_self_attention": true,
24
+ "norm_type": "ada_norm",
25
+ "num_attention_heads": 32,
26
+ "num_layers": 32,
27
+ "output_dim": 1024,
28
+ "positional_embeddings": null
29
+ },
30
+ "dtype": "float32",
31
+ "exclude_state": false,
32
+ "formalize_language": true,
33
+ "hidden_size": 1024,
34
+ "image_crop_size": [
35
+ 230,
36
+ 230
37
+ ],
38
+ "image_target_size": [
39
+ 256,
40
+ 256
41
+ ],
42
+ "letter_box_transform": false,
43
+ "load_bf16": false,
44
+ "max_action_dim": 132,
45
+ "max_num_embodiments": 32,
46
+ "max_seq_len": 1024,
47
+ "max_state_dim": 132,
48
+ "model_dtype": "bfloat16",
49
+ "model_name": "nvidia/Cosmos-Reason2-2B",
50
+ "model_type": "Gr00tN1d7",
51
+ "noise_beta_alpha": 1.5,
52
+ "noise_beta_beta": 1.0,
53
+ "noise_s": 0.999,
54
+ "num_inference_timesteps": 4,
55
+ "num_timestep_buckets": 1000,
56
+ "random_history_crop": true,
57
+ "random_rotation_angle": 0,
58
+ "reproject_vision": false,
59
+ "rtc_ramp_rate": 6.0,
60
+ "select_layer": 16,
61
+ "shortest_image_edge": 256,
62
+ "state_dropout_prob": 0.2,
63
+ "state_gaussian_noise_std": 0.0,
64
+ "transformers_version": "4.57.3",
65
+ "tune_diffusion_model": true,
66
+ "tune_linear": true,
67
+ "tune_llm": false,
68
+ "tune_projector": true,
69
+ "tune_top_llm_layers": 0,
70
+ "tune_visual": false,
71
+ "tune_vlln": true,
72
+ "use_albumentations": true,
73
+ "use_alternate_vl_dit": true,
74
+ "use_flash_attention": true,
75
+ "use_future_tokens": false,
76
+ "use_mean_std": false,
77
+ "use_percentiles": true,
78
+ "use_vl_self_attention": true,
79
+ "use_vlln": true,
80
+ "vl_self_attention_cfg": {
81
+ "attention_head_dim": 64,
82
+ "dropout": 0.2,
83
+ "final_dropout": true,
84
+ "num_attention_heads": 32,
85
+ "num_layers": 4,
86
+ "positional_embeddings": null
87
+ }
88
+ }
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,251 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d7
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Cosmos-Reason2-2B
6
+ backbone_model_type: qwen
7
+ model_revision: null
8
+ tune_top_llm_layers: 0
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: false
12
+ select_layer: 12
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ backbone_trainable_params_fp32: true
17
+ image_crop_size:
18
+ - 230
19
+ - 230
20
+ image_target_size:
21
+ - 256
22
+ - 256
23
+ shortest_image_edge: null
24
+ crop_fraction: null
25
+ random_rotation_angle: null
26
+ color_jitter_params: null
27
+ use_albumentations_transforms: true
28
+ letter_box_transform: false
29
+ extra_augmentation_config: null
30
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_percentiles: true
33
+ use_relative_action: true
34
+ max_state_dim: 132
35
+ max_action_dim: 132
36
+ action_horizon: 40
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ state_history_length: 1
40
+ add_pos_embed: true
41
+ attn_dropout: 0.2
42
+ use_vlln: true
43
+ max_seq_len: 1024
44
+ use_alternate_vl_dit: true
45
+ attend_text_every_n_blocks: 2
46
+ diffusion_model_cfg:
47
+ positional_embeddings: null
48
+ num_layers: 16
49
+ num_attention_heads: 32
50
+ attention_head_dim: 48
51
+ norm_type: ada_norm
52
+ dropout: 0.2
53
+ final_dropout: true
54
+ output_dim: 1024
55
+ interleave_self_attention: true
56
+ num_inference_timesteps: 4
57
+ noise_beta_alpha: 1.5
58
+ noise_beta_beta: 1.0
59
+ noise_s: 0.999
60
+ num_timestep_buckets: 1000
61
+ tune_projector: true
62
+ tune_diffusion_model: true
63
+ tune_vlln: true
64
+ state_dropout_prob: 0.2
65
+ exclude_state: false
66
+ use_mean_std: false
67
+ max_num_embodiments: 32
68
+ data:
69
+ datasets:
70
+ - dataset_paths:
71
+ - chanwoochan/daru_mob_comp_minimum_0715
72
+ embodiment_tag: new_embodiment
73
+ mix_ratio: 1.0
74
+ dataset_type: physical_embodiment
75
+ val_dataset_path: null
76
+ modality_configs:
77
+ new_embodiment:
78
+ video:
79
+ delta_indices:
80
+ - 0
81
+ modality_keys:
82
+ - head_left
83
+ - head_right
84
+ - hand_left
85
+ - hand_right
86
+ sin_cos_embedding_keys: null
87
+ mean_std_embedding_keys: null
88
+ action_configs: null
89
+ state:
90
+ delta_indices:
91
+ - 0
92
+ modality_keys:
93
+ - left_arm
94
+ - right_arm
95
+ - left_hand
96
+ - right_hand
97
+ - head
98
+ - waist
99
+ sin_cos_embedding_keys: null
100
+ mean_std_embedding_keys: null
101
+ action_configs: null
102
+ action:
103
+ delta_indices:
104
+ - 0
105
+ - 1
106
+ - 2
107
+ - 3
108
+ - 4
109
+ - 5
110
+ - 6
111
+ - 7
112
+ - 8
113
+ - 9
114
+ - 10
115
+ - 11
116
+ - 12
117
+ - 13
118
+ - 14
119
+ - 15
120
+ modality_keys:
121
+ - left_arm
122
+ - right_arm
123
+ - left_hand
124
+ - right_hand
125
+ - head
126
+ - waist
127
+ - mob_comp
128
+ - task_goal
129
+ sin_cos_embedding_keys: null
130
+ mean_std_embedding_keys: null
131
+ action_configs:
132
+ - rep: DELTA
133
+ type: NON_EEF
134
+ format: DEFAULT
135
+ state_key: null
136
+ - rep: DELTA
137
+ type: NON_EEF
138
+ format: DEFAULT
139
+ state_key: null
140
+ - rep: DELTA
141
+ type: NON_EEF
142
+ format: DEFAULT
143
+ state_key: null
144
+ - rep: DELTA
145
+ type: NON_EEF
146
+ format: DEFAULT
147
+ state_key: null
148
+ - rep: DELTA
149
+ type: NON_EEF
150
+ format: DEFAULT
151
+ state_key: null
152
+ - rep: DELTA
153
+ type: NON_EEF
154
+ format: DEFAULT
155
+ state_key: null
156
+ - rep: DELTA
157
+ type: NON_EEF
158
+ format: DEFAULT
159
+ state_key: null
160
+ - rep: DELTA
161
+ type: NON_EEF
162
+ format: DEFAULT
163
+ state_key: null
164
+ language:
165
+ delta_indices:
166
+ - 0
167
+ modality_keys:
168
+ - annotation.human.task_description
169
+ sin_cos_embedding_keys: null
170
+ mean_std_embedding_keys: null
171
+ action_configs: null
172
+ download_cache: false
173
+ shard_size: 1024
174
+ episode_sampling_rate: 0.1
175
+ num_shards_per_epoch: 100000
176
+ ds_weights_alpha: null
177
+ override_pretraining_statistics: true
178
+ mode: single_turn
179
+ random_chop: 0.0
180
+ mock_dataset_mode: false
181
+ shuffle: true
182
+ seed: 42
183
+ multiprocessing_context: fork
184
+ allow_padding: false
185
+ subsample_ratio: 1.0
186
+ image_crop_size:
187
+ - 244
188
+ - 244
189
+ image_target_size:
190
+ - 224
191
+ - 224
192
+ training:
193
+ output_dir: output/model/daru_mob_comp_minimum_0715
194
+ experiment_name: null
195
+ max_steps: 50000
196
+ global_batch_size: 24
197
+ per_gpu_batch_size: null
198
+ gradient_accumulation_steps: 1
199
+ learning_rate: 0.0001
200
+ lr_scheduler_type: cosine
201
+ weight_decay: 1.0e-05
202
+ warmup_ratio: 0.05
203
+ warmup_steps: 0
204
+ max_grad_norm: 1.0
205
+ optim: adamw_torch
206
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
207
+ skip_weight_loading: false
208
+ tf32: true
209
+ fp16: false
210
+ bf16: true
211
+ eval_bf16: true
212
+ logging_steps: 10
213
+ save_steps: 50000
214
+ save_total_limit: 3
215
+ save_vl_model: false
216
+ save_only_model: false
217
+ resume_from_checkpoint: false
218
+ upload_checkpoints: false
219
+ upload_every: 1000
220
+ upload_last_n_checkpoints: 5
221
+ max_concurrent_uploads: 2
222
+ eval_strategy: 'no'
223
+ eval_steps: 500
224
+ eval_set_split_ratio: 0.1
225
+ eval_batch_size: 2
226
+ save_best_eval_metric_name: ''
227
+ save_best_eval_metric_greater_is_better: true
228
+ deepspeed_stage: 2
229
+ gradient_checkpointing: false
230
+ transformers_trust_remote_code: true
231
+ transformers_local_files_only: false
232
+ transformers_cache_dir: null
233
+ transformers_access_token: null
234
+ use_ddp: false
235
+ ddp_bucket_cap_mb: 100
236
+ num_gpus: 1
237
+ dataloader_num_workers: 2
238
+ remove_unused_columns: false
239
+ use_wandb: true
240
+ wandb_project: finetune-gr00t-n1d7
241
+ enable_profiling: false
242
+ max_retries: 3
243
+ assert_loss_less_than: null
244
+ add_rl_callback: false
245
+ enable_open_loop_eval: false
246
+ open_loop_eval_traj_ids:
247
+ - 0
248
+ open_loop_eval_steps_per_traj: 100
249
+ open_loop_eval_plot_indices: null
250
+ max_steps: 50000
251
+ save_steps: 50000
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,249 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d7
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Cosmos-Reason2-2B
6
+ backbone_model_type: qwen
7
+ model_revision: null
8
+ tune_top_llm_layers: 0
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: false
12
+ select_layer: 12
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ backbone_trainable_params_fp32: true
17
+ image_crop_size:
18
+ - 230
19
+ - 230
20
+ image_target_size:
21
+ - 256
22
+ - 256
23
+ shortest_image_edge: null
24
+ crop_fraction: null
25
+ random_rotation_angle: null
26
+ color_jitter_params: null
27
+ use_albumentations_transforms: true
28
+ letter_box_transform: false
29
+ extra_augmentation_config: null
30
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_percentiles: true
33
+ use_relative_action: true
34
+ max_state_dim: 132
35
+ max_action_dim: 132
36
+ action_horizon: 40
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ state_history_length: 1
40
+ add_pos_embed: true
41
+ attn_dropout: 0.2
42
+ use_vlln: true
43
+ max_seq_len: 1024
44
+ use_alternate_vl_dit: true
45
+ attend_text_every_n_blocks: 2
46
+ diffusion_model_cfg:
47
+ positional_embeddings: null
48
+ num_layers: 16
49
+ num_attention_heads: 32
50
+ attention_head_dim: 48
51
+ norm_type: ada_norm
52
+ dropout: 0.2
53
+ final_dropout: true
54
+ output_dim: 1024
55
+ interleave_self_attention: true
56
+ num_inference_timesteps: 4
57
+ noise_beta_alpha: 1.5
58
+ noise_beta_beta: 1.0
59
+ noise_s: 0.999
60
+ num_timestep_buckets: 1000
61
+ tune_projector: true
62
+ tune_diffusion_model: true
63
+ tune_vlln: true
64
+ state_dropout_prob: 0.2
65
+ exclude_state: false
66
+ use_mean_std: false
67
+ max_num_embodiments: 32
68
+ data:
69
+ datasets:
70
+ - dataset_paths:
71
+ - chanwoochan/daru_mob_comp_minimum_0715
72
+ embodiment_tag: new_embodiment
73
+ mix_ratio: 1.0
74
+ dataset_type: physical_embodiment
75
+ val_dataset_path: null
76
+ modality_configs:
77
+ new_embodiment:
78
+ video:
79
+ delta_indices:
80
+ - 0
81
+ modality_keys:
82
+ - head_left
83
+ - head_right
84
+ - hand_left
85
+ - hand_right
86
+ sin_cos_embedding_keys: null
87
+ mean_std_embedding_keys: null
88
+ action_configs: null
89
+ state:
90
+ delta_indices:
91
+ - 0
92
+ modality_keys:
93
+ - left_arm
94
+ - right_arm
95
+ - left_hand
96
+ - right_hand
97
+ - head
98
+ - waist
99
+ sin_cos_embedding_keys: null
100
+ mean_std_embedding_keys: null
101
+ action_configs: null
102
+ action:
103
+ delta_indices:
104
+ - 0
105
+ - 1
106
+ - 2
107
+ - 3
108
+ - 4
109
+ - 5
110
+ - 6
111
+ - 7
112
+ - 8
113
+ - 9
114
+ - 10
115
+ - 11
116
+ - 12
117
+ - 13
118
+ - 14
119
+ - 15
120
+ modality_keys:
121
+ - left_arm
122
+ - right_arm
123
+ - left_hand
124
+ - right_hand
125
+ - head
126
+ - waist
127
+ - mob_comp
128
+ - task_goal
129
+ sin_cos_embedding_keys: null
130
+ mean_std_embedding_keys: null
131
+ action_configs:
132
+ - rep: delta
133
+ type: non_eef
134
+ format: default
135
+ state_key: null
136
+ - rep: delta
137
+ type: non_eef
138
+ format: default
139
+ state_key: null
140
+ - rep: delta
141
+ type: non_eef
142
+ format: default
143
+ state_key: null
144
+ - rep: delta
145
+ type: non_eef
146
+ format: default
147
+ state_key: null
148
+ - rep: delta
149
+ type: non_eef
150
+ format: default
151
+ state_key: null
152
+ - rep: delta
153
+ type: non_eef
154
+ format: default
155
+ state_key: null
156
+ - rep: delta
157
+ type: non_eef
158
+ format: default
159
+ state_key: null
160
+ - rep: delta
161
+ type: non_eef
162
+ format: default
163
+ state_key: null
164
+ language:
165
+ delta_indices:
166
+ - 0
167
+ modality_keys:
168
+ - annotation.human.task_description
169
+ sin_cos_embedding_keys: null
170
+ mean_std_embedding_keys: null
171
+ action_configs: null
172
+ download_cache: false
173
+ shard_size: 1024
174
+ episode_sampling_rate: 0.1
175
+ num_shards_per_epoch: 100000
176
+ ds_weights_alpha: null
177
+ override_pretraining_statistics: true
178
+ mode: single_turn
179
+ random_chop: 0.0
180
+ mock_dataset_mode: false
181
+ shuffle: true
182
+ seed: 42
183
+ multiprocessing_context: fork
184
+ allow_padding: false
185
+ subsample_ratio: 1.0
186
+ image_crop_size:
187
+ - 244
188
+ - 244
189
+ image_target_size:
190
+ - 224
191
+ - 224
192
+ training:
193
+ output_dir: output/model/daru_mob_comp_minimum_0715
194
+ experiment_name: null
195
+ max_steps: 50000
196
+ global_batch_size: 24
197
+ per_gpu_batch_size: null
198
+ gradient_accumulation_steps: 1
199
+ learning_rate: 0.0001
200
+ lr_scheduler_type: cosine
201
+ weight_decay: 1.0e-05
202
+ warmup_ratio: 0.05
203
+ warmup_steps: 0
204
+ max_grad_norm: 1.0
205
+ optim: adamw_torch
206
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
207
+ skip_weight_loading: false
208
+ tf32: true
209
+ fp16: false
210
+ bf16: true
211
+ eval_bf16: true
212
+ logging_steps: 10
213
+ save_steps: 50000
214
+ save_total_limit: 3
215
+ save_vl_model: false
216
+ save_only_model: false
217
+ resume_from_checkpoint: false
218
+ upload_checkpoints: false
219
+ upload_every: 1000
220
+ upload_last_n_checkpoints: 5
221
+ max_concurrent_uploads: 2
222
+ eval_strategy: 'no'
223
+ eval_steps: 500
224
+ eval_set_split_ratio: 0.1
225
+ eval_batch_size: 2
226
+ save_best_eval_metric_name: ''
227
+ save_best_eval_metric_greater_is_better: true
228
+ deepspeed_stage: 2
229
+ gradient_checkpointing: false
230
+ transformers_trust_remote_code: true
231
+ transformers_local_files_only: false
232
+ transformers_cache_dir: null
233
+ transformers_access_token: null
234
+ use_ddp: false
235
+ ddp_bucket_cap_mb: 100
236
+ num_gpus: 1
237
+ dataloader_num_workers: 2
238
+ remove_unused_columns: false
239
+ use_wandb: true
240
+ wandb_project: finetune-gr00t-n1d7
241
+ enable_profiling: false
242
+ max_retries: 3
243
+ assert_loss_less_than: null
244
+ add_rl_callback: false
245
+ enable_open_loop_eval: false
246
+ open_loop_eval_traj_ids:
247
+ - 0
248
+ open_loop_eval_steps_per_traj: 100
249
+ open_loop_eval_plot_indices: null
experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,595 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "left_arm": {
5
+ "min": [
6
+ -0.9032078981399536,
7
+ -0.17104226350784302,
8
+ -1.1798425912857056,
9
+ -2.3226840496063232,
10
+ -1.4744541645050049,
11
+ -0.5880014300346375,
12
+ -0.7271041870117188
13
+ ],
14
+ "max": [
15
+ 0.3963642716407776,
16
+ 0.7513642311096191,
17
+ 0.5494296550750732,
18
+ -0.6892305016517639,
19
+ 0.7433357238769531,
20
+ 0.6255260109901428,
21
+ 1.0030406713485718
22
+ ],
23
+ "mean": [
24
+ -0.2869558334350586,
25
+ 0.17532365024089813,
26
+ -0.47614753246307373,
27
+ -1.7166075706481934,
28
+ -0.18068324029445648,
29
+ 0.5105005502700806,
30
+ 0.2532295286655426
31
+ ],
32
+ "std": [
33
+ 0.19354814291000366,
34
+ 0.1314506232738495,
35
+ 0.2711745500564575,
36
+ 0.3270381689071655,
37
+ 0.39534708857536316,
38
+ 0.21731241047382355,
39
+ 0.25014322996139526
40
+ ],
41
+ "q01": [
42
+ -0.7380753207206726,
43
+ -0.12591328606009483,
44
+ -0.8992686891555786,
45
+ -2.238983335494995,
46
+ -1.4360568523406982,
47
+ -0.29329036206007003,
48
+ -0.5905199331045151
49
+ ],
50
+ "q99": [
51
+ 0.2684316337108612,
52
+ 0.590570544600487,
53
+ 0.4492477476596832,
54
+ -0.7885397672653198,
55
+ 0.5502028489112855,
56
+ 0.6190682649612427,
57
+ 0.7808114385604861
58
+ ]
59
+ },
60
+ "right_arm": {
61
+ "min": [
62
+ -1.0232865810394287,
63
+ -0.34435346722602844,
64
+ -2.2130775451660156,
65
+ -2.2280874252319336,
66
+ -0.3998549282550812,
67
+ -0.7140141725540161,
68
+ -0.07504915446043015
69
+ ],
70
+ "max": [
71
+ 0.36023595929145813,
72
+ 0.2612757980823517,
73
+ 1.5369369983673096,
74
+ -0.5520476698875427,
75
+ 1.4470524787902832,
76
+ -0.1054178848862648,
77
+ 1.0857692956924438
78
+ ],
79
+ "mean": [
80
+ -0.26461294293403625,
81
+ -0.10266675800085068,
82
+ 0.6031351089477539,
83
+ -1.7845441102981567,
84
+ 0.26721641421318054,
85
+ -0.5852886438369751,
86
+ 0.5154456496238708
87
+ ],
88
+ "std": [
89
+ 0.22050070762634277,
90
+ 0.09380312263965607,
91
+ 0.2636542320251465,
92
+ 0.27053317427635193,
93
+ 0.3317246437072754,
94
+ 0.0631183311343193,
95
+ 0.21471618115901947
96
+ ],
97
+ "q01": [
98
+ -0.7713866585493088,
99
+ -0.304285945892334,
100
+ 0.017029177658259866,
101
+ -2.16717529296875,
102
+ -0.15329226270318033,
103
+ -0.6524040699005127,
104
+ 0.19320794939994812
105
+ ],
106
+ "q99": [
107
+ 0.2689552307128906,
108
+ 0.17824872985482232,
109
+ 1.2808971405029297,
110
+ -0.9394356924295425,
111
+ 1.4196507930755615,
112
+ -0.25132742524147034,
113
+ 1.0289466321468355
114
+ ]
115
+ },
116
+ "left_hand": {
117
+ "min": [
118
+ -0.8945000171661377,
119
+ -0.1889999955892563,
120
+ -0.007000000216066837,
121
+ -0.12600000202655792,
122
+ -0.9514999985694885,
123
+ -0.9599999785423279
124
+ ],
125
+ "max": [
126
+ 0.6884999871253967,
127
+ 0.9120000004768372,
128
+ 0.9595000147819519,
129
+ 0.9120000004768372,
130
+ 0.9934999942779541,
131
+ 0.9934999942779541
132
+ ],
133
+ "mean": [
134
+ 0.1297818273305893,
135
+ 0.11738239973783493,
136
+ 0.11905454844236374,
137
+ 0.10788065940141678,
138
+ 0.09944233298301697,
139
+ 0.6733852624893188
140
+ ],
141
+ "std": [
142
+ 0.21148353815078735,
143
+ 0.21655233204364777,
144
+ 0.20444288849830627,
145
+ 0.19637329876422882,
146
+ 0.20284469425678253,
147
+ 0.2246493250131607
148
+ ],
149
+ "q01": [
150
+ -0.0010000000474974513,
151
+ -0.0005000000237487257,
152
+ 0.0,
153
+ -0.0010000000474974513,
154
+ -0.003000000026077032,
155
+ 0.07050000131130219
156
+ ],
157
+ "q99": [
158
+ 0.6449999809265137,
159
+ 0.6345000267028809,
160
+ 0.6065000295639038,
161
+ 0.5899999737739563,
162
+ 0.8355000019073486,
163
+ 0.9700000286102295
164
+ ]
165
+ },
166
+ "right_hand": {
167
+ "min": [
168
+ -0.9714999794960022,
169
+ -0.1264999955892563,
170
+ -0.6384999752044678,
171
+ -0.6334999799728394,
172
+ -0.014499999582767487,
173
+ 0.421999990940094
174
+ ],
175
+ "max": [
176
+ 0.9679999947547913,
177
+ 0.8519999980926514,
178
+ 0.8759999871253967,
179
+ 0.8190000057220459,
180
+ 0.8234999775886536,
181
+ 0.987500011920929
182
+ ],
183
+ "mean": [
184
+ 0.0794672966003418,
185
+ 0.08586394041776657,
186
+ 0.0845729261636734,
187
+ 0.07792633771896362,
188
+ 0.0825234204530716,
189
+ 0.6479356288909912
190
+ ],
191
+ "std": [
192
+ 0.18681903183460236,
193
+ 0.19215397536754608,
194
+ 0.18897227942943573,
195
+ 0.17260189354419708,
196
+ 0.18959352374076843,
197
+ 0.17591802775859833
198
+ ],
199
+ "q01": [
200
+ -0.0044999998062849045,
201
+ -0.0005000000237487257,
202
+ -0.004999999888241291,
203
+ -0.0010000000474974513,
204
+ -0.004000000189989805,
205
+ 0.5199999809265137
206
+ ],
207
+ "q99": [
208
+ 0.8370000123977661,
209
+ 0.8305000066757202,
210
+ 0.6029999852180481,
211
+ 0.5400000214576721,
212
+ 0.7455000281333923,
213
+ 0.9729999899864197
214
+ ]
215
+ },
216
+ "head": {
217
+ "min": [
218
+ -0.6981316804885864,
219
+ -0.12217304855585098
220
+ ],
221
+ "max": [
222
+ 0.8552113175392151,
223
+ 0.13962633907794952
224
+ ],
225
+ "mean": [
226
+ 0.06253354251384735,
227
+ -0.0034417936112731695
228
+ ],
229
+ "std": [
230
+ 0.30650585889816284,
231
+ 0.05744078382849693
232
+ ],
233
+ "q01": [
234
+ -0.6457718014717102,
235
+ -0.12217304855585098
236
+ ],
237
+ "q99": [
238
+ 0.7853981852531433,
239
+ 0.12217304855585098
240
+ ]
241
+ },
242
+ "waist": {
243
+ "min": [
244
+ -0.7402533292770386
245
+ ],
246
+ "max": [
247
+ 0.541886031627655
248
+ ],
249
+ "mean": [
250
+ -0.04419296979904175
251
+ ],
252
+ "std": [
253
+ 0.23224681615829468
254
+ ],
255
+ "q01": [
256
+ -0.6261920928955078
257
+ ],
258
+ "q99": [
259
+ 0.4083695709705353
260
+ ]
261
+ }
262
+ },
263
+ "action": {
264
+ "left_arm": {
265
+ "min": [
266
+ -0.8499535322189331,
267
+ -0.25394493341445923,
268
+ -1.2947511672973633,
269
+ -2.356194019317627,
270
+ -1.5707954168319702,
271
+ -0.6283181309700012,
272
+ -0.8141002058982849
273
+ ],
274
+ "max": [
275
+ 0.4256761372089386,
276
+ 0.7752004861831665,
277
+ 0.6364154815673828,
278
+ -0.5632686018943787,
279
+ 0.7934372425079346,
280
+ 0.6283181309700012,
281
+ 1.047196865081787
282
+ ],
283
+ "mean": [
284
+ -0.2504061758518219,
285
+ 0.16202592849731445,
286
+ -0.4816209673881531,
287
+ -1.684709906578064,
288
+ -0.184758722782135,
289
+ 0.5335910320281982,
290
+ 0.25200003385543823
291
+ ],
292
+ "std": [
293
+ 0.18740738928318024,
294
+ 0.13428013026714325,
295
+ 0.279079794883728,
296
+ 0.3418758809566498,
297
+ 0.4204392731189728,
298
+ 0.23198546469211578,
299
+ 0.254506379365921
300
+ ],
301
+ "q01": [
302
+ -0.7154676306247711,
303
+ -0.16645458683371545,
304
+ -0.9364646309614182,
305
+ -2.213869159221649,
306
+ -1.5707954168319702,
307
+ -0.3343559408187866,
308
+ -0.6008876466751099
309
+ ],
310
+ "q99": [
311
+ 0.2805453842878344,
312
+ 0.5947800517082216,
313
+ 0.4674780580401421,
314
+ -0.7327411317825316,
315
+ 0.5889149415493011,
316
+ 0.6283181309700012,
317
+ 0.8024686759710313
318
+ ]
319
+ },
320
+ "right_arm": {
321
+ "min": [
322
+ -1.042810082435608,
323
+ -0.3474646210670471,
324
+ -0.10725196450948715,
325
+ -2.3385281562805176,
326
+ -0.34228163957595825,
327
+ -0.6283181309700012,
328
+ -0.09903465956449509
329
+ ],
330
+ "max": [
331
+ 0.4206028878688812,
332
+ 0.3490656316280365,
333
+ 1.653267502784729,
334
+ -0.4729674756526947,
335
+ 1.5707954168319702,
336
+ -0.04136136919260025,
337
+ 1.047196865081787
338
+ ],
339
+ "mean": [
340
+ -0.2439965158700943,
341
+ -0.1011262983083725,
342
+ 0.6109874248504639,
343
+ -1.7618523836135864,
344
+ 0.2710442841053009,
345
+ -0.6060145497322083,
346
+ 0.5119675397872925
347
+ ],
348
+ "std": [
349
+ 0.22085259854793549,
350
+ 0.10317016392946243,
351
+ 0.27652469277381897,
352
+ 0.2890540063381195,
353
+ 0.3692925274372101,
354
+ 0.07110434770584106,
355
+ 0.23511721193790436
356
+ ],
357
+ "q01": [
358
+ -0.7713390624523163,
359
+ -0.3190961357951164,
360
+ 0.01663753580302,
361
+ -2.2107937860488893,
362
+ -0.21337952315807343,
363
+ -0.6283181309700012,
364
+ 0.1675674919784069
365
+ ],
366
+ "q99": [
367
+ 0.2990204724669457,
368
+ 0.23830342411994943,
369
+ 1.362928637266159,
370
+ -0.900239096879959,
371
+ 1.5707954168319702,
372
+ -0.22352868109941482,
373
+ 1.047196865081787
374
+ ]
375
+ },
376
+ "left_hand": {
377
+ "min": [
378
+ 0.0,
379
+ 0.0,
380
+ 0.0,
381
+ 0.0,
382
+ 0.0,
383
+ 0.5
384
+ ],
385
+ "max": [
386
+ 1.0,
387
+ 1.0,
388
+ 1.0,
389
+ 1.0,
390
+ 1.0,
391
+ 1.0
392
+ ],
393
+ "mean": [
394
+ 0.22313059866428375,
395
+ 0.22313059866428375,
396
+ 0.22313059866428375,
397
+ 0.22313059866428375,
398
+ 0.22313059866428375,
399
+ 0.6710783839225769
400
+ ],
401
+ "std": [
402
+ 0.41156288981437683,
403
+ 0.41156288981437683,
404
+ 0.41156288981437683,
405
+ 0.41156288981437683,
406
+ 0.41156288981437683,
407
+ 0.22488965094089508
408
+ ],
409
+ "q01": [
410
+ 0.0,
411
+ 0.0,
412
+ 0.0,
413
+ 0.0,
414
+ 0.0,
415
+ 0.5
416
+ ],
417
+ "q99": [
418
+ 1.0,
419
+ 1.0,
420
+ 1.0,
421
+ 1.0,
422
+ 1.0,
423
+ 0.9999073147773743
424
+ ]
425
+ },
426
+ "right_hand": {
427
+ "min": [
428
+ 0.0,
429
+ 0.0,
430
+ 0.0,
431
+ 0.0,
432
+ 0.0,
433
+ 0.5
434
+ ],
435
+ "max": [
436
+ 1.0,
437
+ 1.0,
438
+ 1.0,
439
+ 1.0,
440
+ 1.0,
441
+ 1.0
442
+ ],
443
+ "mean": [
444
+ 0.16685084998607635,
445
+ 0.16685084998607635,
446
+ 0.16685084998607635,
447
+ 0.16685084998607635,
448
+ 0.16685084998607635,
449
+ 0.6144992709159851
450
+ ],
451
+ "std": [
452
+ 0.36956050992012024,
453
+ 0.36956050992012024,
454
+ 0.36956050992012024,
455
+ 0.36956050992012024,
456
+ 0.36956050992012024,
457
+ 0.20083360373973846
458
+ ],
459
+ "q01": [
460
+ 0.0,
461
+ 0.0,
462
+ 0.0,
463
+ 0.0,
464
+ 0.0,
465
+ 0.5
466
+ ],
467
+ "q99": [
468
+ 1.0,
469
+ 1.0,
470
+ 1.0,
471
+ 1.0,
472
+ 1.0,
473
+ 0.999963641166687
474
+ ]
475
+ },
476
+ "head": {
477
+ "min": [
478
+ -0.7581518888473511,
479
+ -0.11258255690336227
480
+ ],
481
+ "max": [
482
+ 0.8933978080749512,
483
+ 0.1990974098443985
484
+ ],
485
+ "mean": [
486
+ 0.0644804909825325,
487
+ 0.02685731090605259
488
+ ],
489
+ "std": [
490
+ 0.3190660774707794,
491
+ 0.062273312360048294
492
+ ],
493
+ "q01": [
494
+ -0.685428461432457,
495
+ -0.10292920261621476
496
+ ],
497
+ "q99": [
498
+ 0.807757226228714,
499
+ 0.15959951758384708
500
+ ]
501
+ },
502
+ "waist": {
503
+ "min": [
504
+ -0.7449859976768494
505
+ ],
506
+ "max": [
507
+ 0.5428807735443115
508
+ ],
509
+ "mean": [
510
+ -0.04446377605199814
511
+ ],
512
+ "std": [
513
+ 0.2328842282295227
514
+ ],
515
+ "q01": [
516
+ -0.6253079611063004
517
+ ],
518
+ "q99": [
519
+ 0.41605783462524415
520
+ ]
521
+ },
522
+ "mob_comp": {
523
+ "min": [
524
+ 0.0,
525
+ 0.0,
526
+ 0.0,
527
+ 0.0,
528
+ 0.0,
529
+ 0.0
530
+ ],
531
+ "max": [
532
+ 1.0,
533
+ 0.0,
534
+ 1.0,
535
+ 0.0,
536
+ 1.0,
537
+ 1.0
538
+ ],
539
+ "mean": [
540
+ 0.03673145920038223,
541
+ 0.0,
542
+ 0.1621939092874527,
543
+ 0.0,
544
+ 0.03323742002248764,
545
+ 0.12208585441112518
546
+ ],
547
+ "std": [
548
+ 0.18808671832084656,
549
+ 0.0,
550
+ 0.36869359016418457,
551
+ 0.0,
552
+ 0.17924492061138153,
553
+ 0.32738250494003296
554
+ ],
555
+ "q01": [
556
+ 0.0,
557
+ 0.0,
558
+ 0.0,
559
+ 0.0,
560
+ 0.0,
561
+ 0.0
562
+ ],
563
+ "q99": [
564
+ 1.0,
565
+ 0.0,
566
+ 1.0,
567
+ 0.0,
568
+ 1.0,
569
+ 1.0
570
+ ]
571
+ },
572
+ "task_goal": {
573
+ "min": [
574
+ 0.0
575
+ ],
576
+ "max": [
577
+ 1.0
578
+ ],
579
+ "mean": [
580
+ 0.07475482672452927
581
+ ],
582
+ "std": [
583
+ 0.2630324363708496
584
+ ],
585
+ "q01": [
586
+ 0.0
587
+ ],
588
+ "q99": [
589
+ 1.0
590
+ ]
591
+ }
592
+ },
593
+ "relative_action": {}
594
+ }
595
+ }
experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "Gr00tN1d7",
3
+ "model_dtype": "bfloat16",
4
+ "model_name": "nvidia/Cosmos-Reason2-2B",
5
+ "backbone_model_type": "qwen",
6
+ "model_revision": null,
7
+ "tune_top_llm_layers": 0,
8
+ "backbone_embedding_dim": 2048,
9
+ "tune_llm": false,
10
+ "tune_visual": false,
11
+ "select_layer": 16,
12
+ "reproject_vision": false,
13
+ "use_flash_attention": true,
14
+ "load_bf16": false,
15
+ "backbone_trainable_params_fp32": true,
16
+ "letter_box_transform": false,
17
+ "extra_augmentation_config": null,
18
+ "apply_sincos_state_encoding": false,
19
+ "use_percentiles": true,
20
+ "use_relative_action": false,
21
+ "max_state_dim": 132,
22
+ "max_action_dim": 132,
23
+ "action_horizon": 40,
24
+ "hidden_size": 1024,
25
+ "input_embedding_dim": 1536,
26
+ "state_history_length": 1,
27
+ "add_pos_embed": true,
28
+ "attn_dropout": 0.2,
29
+ "use_vlln": true,
30
+ "max_seq_len": 1024,
31
+ "use_alternate_vl_dit": true,
32
+ "attend_text_every_n_blocks": 2,
33
+ "diffusion_model_cfg": {
34
+ "attention_head_dim": 48,
35
+ "dropout": 0.2,
36
+ "final_dropout": true,
37
+ "interleave_self_attention": true,
38
+ "norm_type": "ada_norm",
39
+ "num_attention_heads": 32,
40
+ "num_layers": 32,
41
+ "output_dim": 1024,
42
+ "positional_embeddings": null
43
+ },
44
+ "num_inference_timesteps": 4,
45
+ "noise_beta_alpha": 1.5,
46
+ "noise_beta_beta": 1.0,
47
+ "noise_s": 0.999,
48
+ "num_timestep_buckets": 1000,
49
+ "tune_projector": true,
50
+ "tune_diffusion_model": true,
51
+ "tune_vlln": true,
52
+ "state_dropout_prob": 0.2,
53
+ "exclude_state": false,
54
+ "use_mean_std": false,
55
+ "max_num_embodiments": 32
56
+ }
experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbcaea5ee88f1e0f1465043920a2647c67e7de17d24adfd1c477742a6168edec
3
+ size 4986649584
model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca72f26be25f0bf67d1e8c7d8cccdb1a1efb5e9fcc623262f900dd72ea924b52
3
+ size 4970792616
model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ebf7872ce1490869f9660b4c00750db30840739bfb020bb7ad0718e297a4992
3
+ size 2618758696
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
processor/embodiment_id.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "robocasa_panda_omron": 13,
3
+ "oxe_droid": 17,
4
+ "oxe_fractal": 18,
5
+ "oxe_language_table": 19,
6
+ "oxe_bridge": 20,
7
+ "unknown": 22,
8
+ "gr1_unified": 20,
9
+ "agibot": 26,
10
+ "sim_behavior_r1_pro": 23,
11
+ "xdof": 24,
12
+ "xdof_oss_data": 25,
13
+ "unitree_g1_full_body_with_waist_height_nav_cmd": 25,
14
+ "real_r1_pro_sharpa": 27,
15
+ "real_r1_pro_sharpa_add_view": 27,
16
+ "real_r1_pro_sharpa_relative_arm_joint": 26,
17
+ "real_r1_pro_sharpa_delta_eef": 26,
18
+ "real_r1_pro_sharpa_absolute_eef": 26,
19
+ "real_r1_pro_sharpa_meanstd": 26,
20
+ "real_r1_pro_sharpa_relative_eef": 26,
21
+ "real_r1_pro_sharpa_relative_eef_add_view": 26,
22
+ "real_r1_pro_sharpa_relative_eef_relative_hand": 26,
23
+ "real_r1_pro_sharpa_relative_eef_human": 26,
24
+ "real_r1_pro_sharpa_relative_eef_human_add_view": 26,
25
+ "real_r1_pro_sharpa_relative_eef_human_relative_hand": 26,
26
+ "real_r1_pro_sharpa_relative_eef_egodex": 26,
27
+ "real_r1_pro_sharpa_relative_eef_egodex_relative_hand": 26,
28
+ "real_r1_pro_sharpa_relative_eef_egodex_wrist_only": 26,
29
+ "real_r1_pro_sharpa_relative_eef_maxinsights": 26,
30
+ "real_r1_pro_sharpa_relative_eef_maxinsights_relative_hand": 26,
31
+ "real_r1_pro_sharpa_relative_eef_mecka": 26,
32
+ "real_r1_pro_sharpa_relative_eef_mecka_relative_hand": 26,
33
+ "real_g1_relative_eef_absolute_joints": 25,
34
+ "real_g1_relative_eef_absolute_joints_wrist_cam": 25,
35
+ "real_g1_relative_eef_relative_joints": 25,
36
+ "real_r1_pro_sharpa_relative_eef_relative_hand_relative_joint": 26,
37
+ "real_r1_pro_sharpa_relative_joint": 29,
38
+ "oxe_droid_relative_eef_relative_joint": 24,
39
+ "oxe_droid_relative_eef_relative_joint_swapped": 24,
40
+ "oxe_droid_relative_eef_relative_joint_upweight_z": 24,
41
+ "oxe_droid_relative_eef_relative_joint_upweight_z_swapped": 24,
42
+ "oxe_droid_relative_eef_relative_joint_3view": 24,
43
+ "oxe_droid_relative_eef_relative_joint_3view_swapped": 24,
44
+ "oxe_droid_relative_eef": 24,
45
+ "oxe_droid_joint_position_relative": 24,
46
+ "xdof_relative_eef_relative_joint": 27,
47
+ "xdof_relative_eef_relative_joint_subtask": 27,
48
+ "xdof_relative_eef": 27,
49
+ "xdof_relative_joint": 28,
50
+ "simpler_env_google": 0,
51
+ "simpler_env_widowx": 1,
52
+ "libero_sim": 2,
53
+ "droid_sim": 3,
54
+ "new_embodiment": 10,
55
+ "robocasa_gr1_tabletop": 10,
56
+ "unitree_g1_sonic": 11
57
+ }
processor/processor_config.json ADDED
@@ -0,0 +1,1171 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "processor_class": "Gr00tN1d7Processor",
3
+ "processor_kwargs": {
4
+ "modality_configs": {
5
+ "real_g1_relative_eef_relative_joints": {
6
+ "video": {
7
+ "delta_indices": [
8
+ -20,
9
+ 0
10
+ ],
11
+ "modality_keys": [
12
+ "ego_view"
13
+ ],
14
+ "sin_cos_embedding_keys": null,
15
+ "mean_std_embedding_keys": null,
16
+ "action_configs": null
17
+ },
18
+ "state": {
19
+ "delta_indices": [
20
+ 0
21
+ ],
22
+ "modality_keys": [
23
+ "left_wrist_eef_9d",
24
+ "right_wrist_eef_9d",
25
+ "left_hand",
26
+ "right_hand",
27
+ "left_arm",
28
+ "right_arm",
29
+ "waist"
30
+ ],
31
+ "sin_cos_embedding_keys": null,
32
+ "mean_std_embedding_keys": null,
33
+ "action_configs": null
34
+ },
35
+ "action": {
36
+ "delta_indices": [
37
+ 0,
38
+ 1,
39
+ 2,
40
+ 3,
41
+ 4,
42
+ 5,
43
+ 6,
44
+ 7,
45
+ 8,
46
+ 9,
47
+ 10,
48
+ 11,
49
+ 12,
50
+ 13,
51
+ 14,
52
+ 15,
53
+ 16,
54
+ 17,
55
+ 18,
56
+ 19,
57
+ 20,
58
+ 21,
59
+ 22,
60
+ 23,
61
+ 24,
62
+ 25,
63
+ 26,
64
+ 27,
65
+ 28,
66
+ 29,
67
+ 30,
68
+ 31,
69
+ 32,
70
+ 33,
71
+ 34,
72
+ 35,
73
+ 36,
74
+ 37,
75
+ 38,
76
+ 39
77
+ ],
78
+ "modality_keys": [
79
+ "left_wrist_eef_9d",
80
+ "right_wrist_eef_9d",
81
+ "left_hand",
82
+ "right_hand",
83
+ "left_arm",
84
+ "right_arm",
85
+ "waist",
86
+ "base_height_command",
87
+ "navigate_command"
88
+ ],
89
+ "sin_cos_embedding_keys": null,
90
+ "mean_std_embedding_keys": null,
91
+ "action_configs": [
92
+ {
93
+ "rep": "RELATIVE",
94
+ "type": "EEF",
95
+ "format": "XYZ_ROT6D",
96
+ "state_key": "left_wrist_eef_9d"
97
+ },
98
+ {
99
+ "rep": "RELATIVE",
100
+ "type": "EEF",
101
+ "format": "XYZ_ROT6D",
102
+ "state_key": "right_wrist_eef_9d"
103
+ },
104
+ {
105
+ "rep": "ABSOLUTE",
106
+ "type": "NON_EEF",
107
+ "format": "DEFAULT",
108
+ "state_key": "left_hand"
109
+ },
110
+ {
111
+ "rep": "ABSOLUTE",
112
+ "type": "NON_EEF",
113
+ "format": "DEFAULT",
114
+ "state_key": "right_hand"
115
+ },
116
+ {
117
+ "rep": "RELATIVE",
118
+ "type": "NON_EEF",
119
+ "format": "DEFAULT",
120
+ "state_key": "left_arm"
121
+ },
122
+ {
123
+ "rep": "RELATIVE",
124
+ "type": "NON_EEF",
125
+ "format": "DEFAULT",
126
+ "state_key": "right_arm"
127
+ },
128
+ {
129
+ "rep": "ABSOLUTE",
130
+ "type": "NON_EEF",
131
+ "format": "DEFAULT",
132
+ "state_key": "waist"
133
+ },
134
+ {
135
+ "rep": "ABSOLUTE",
136
+ "type": "NON_EEF",
137
+ "format": "DEFAULT",
138
+ "state_key": "base_height_command"
139
+ },
140
+ {
141
+ "rep": "ABSOLUTE",
142
+ "type": "NON_EEF",
143
+ "format": "DEFAULT",
144
+ "state_key": "navigate_command"
145
+ }
146
+ ]
147
+ },
148
+ "language": {
149
+ "delta_indices": [
150
+ 0
151
+ ],
152
+ "modality_keys": [
153
+ "annotation.human.task_description"
154
+ ],
155
+ "sin_cos_embedding_keys": null,
156
+ "mean_std_embedding_keys": null,
157
+ "action_configs": null
158
+ }
159
+ },
160
+ "real_r1_pro_sharpa_relative_eef_mecka": {
161
+ "video": {
162
+ "delta_indices": [
163
+ -30,
164
+ 0
165
+ ],
166
+ "modality_keys": [
167
+ "ego_view_cropratio_res320x240_freq30"
168
+ ],
169
+ "sin_cos_embedding_keys": null,
170
+ "mean_std_embedding_keys": null,
171
+ "action_configs": null
172
+ },
173
+ "state": {
174
+ "delta_indices": [
175
+ 0
176
+ ],
177
+ "modality_keys": [
178
+ "left_wrist_eef",
179
+ "right_wrist_eef",
180
+ "left_hand_joints",
181
+ "right_hand_joints"
182
+ ],
183
+ "sin_cos_embedding_keys": null,
184
+ "mean_std_embedding_keys": null,
185
+ "action_configs": null
186
+ },
187
+ "action": {
188
+ "delta_indices": [
189
+ 0,
190
+ 1,
191
+ 2,
192
+ 3,
193
+ 4,
194
+ 5,
195
+ 6,
196
+ 7,
197
+ 8,
198
+ 9,
199
+ 10,
200
+ 11,
201
+ 12,
202
+ 13,
203
+ 14,
204
+ 15,
205
+ 16,
206
+ 17,
207
+ 18,
208
+ 19,
209
+ 20,
210
+ 21,
211
+ 22,
212
+ 23,
213
+ 24,
214
+ 25,
215
+ 26,
216
+ 27,
217
+ 28,
218
+ 29,
219
+ 30,
220
+ 31,
221
+ 32,
222
+ 33,
223
+ 34,
224
+ 35,
225
+ 36,
226
+ 37,
227
+ 38,
228
+ 39
229
+ ],
230
+ "modality_keys": [
231
+ "left_wrist_eef",
232
+ "right_wrist_eef",
233
+ "left_hand_joints",
234
+ "right_hand_joints"
235
+ ],
236
+ "sin_cos_embedding_keys": null,
237
+ "mean_std_embedding_keys": null,
238
+ "action_configs": [
239
+ {
240
+ "rep": "RELATIVE",
241
+ "type": "EEF",
242
+ "format": "XYZ_ROT6D",
243
+ "state_key": "left_wrist_eef"
244
+ },
245
+ {
246
+ "rep": "RELATIVE",
247
+ "type": "EEF",
248
+ "format": "XYZ_ROT6D",
249
+ "state_key": "right_wrist_eef"
250
+ },
251
+ {
252
+ "rep": "ABSOLUTE",
253
+ "type": "NON_EEF",
254
+ "format": "DEFAULT",
255
+ "state_key": "left_hand_joints"
256
+ },
257
+ {
258
+ "rep": "ABSOLUTE",
259
+ "type": "NON_EEF",
260
+ "format": "DEFAULT",
261
+ "state_key": "right_hand_joints"
262
+ }
263
+ ]
264
+ },
265
+ "language": {
266
+ "delta_indices": [
267
+ 0
268
+ ],
269
+ "modality_keys": [
270
+ "annotation.human.coarse_action"
271
+ ],
272
+ "sin_cos_embedding_keys": null,
273
+ "mean_std_embedding_keys": null,
274
+ "action_configs": null
275
+ }
276
+ },
277
+ "real_r1_pro_sharpa_relative_eef_human": {
278
+ "video": {
279
+ "delta_indices": [
280
+ -20,
281
+ 0
282
+ ],
283
+ "modality_keys": [
284
+ "ego_view_res320x240_freq20",
285
+ "left_wrist_view_res320x240_freq20",
286
+ "right_wrist_view_res320x240_freq20"
287
+ ],
288
+ "sin_cos_embedding_keys": null,
289
+ "mean_std_embedding_keys": null,
290
+ "action_configs": null
291
+ },
292
+ "state": {
293
+ "delta_indices": [
294
+ 0
295
+ ],
296
+ "modality_keys": [
297
+ "left_wrist_eef",
298
+ "right_wrist_eef",
299
+ "left_hand_joints",
300
+ "right_hand_joints"
301
+ ],
302
+ "sin_cos_embedding_keys": null,
303
+ "mean_std_embedding_keys": null,
304
+ "action_configs": null
305
+ },
306
+ "action": {
307
+ "delta_indices": [
308
+ 0,
309
+ 1,
310
+ 2,
311
+ 3,
312
+ 4,
313
+ 5,
314
+ 6,
315
+ 7,
316
+ 8,
317
+ 9,
318
+ 10,
319
+ 11,
320
+ 12,
321
+ 13,
322
+ 14,
323
+ 15,
324
+ 16,
325
+ 17,
326
+ 18,
327
+ 19,
328
+ 20,
329
+ 21,
330
+ 22,
331
+ 23,
332
+ 24,
333
+ 25,
334
+ 26,
335
+ 27,
336
+ 28,
337
+ 29,
338
+ 30,
339
+ 31,
340
+ 32,
341
+ 33,
342
+ 34,
343
+ 35,
344
+ 36,
345
+ 37,
346
+ 38,
347
+ 39
348
+ ],
349
+ "modality_keys": [
350
+ "left_wrist_eef",
351
+ "right_wrist_eef",
352
+ "left_hand_joints",
353
+ "right_hand_joints"
354
+ ],
355
+ "sin_cos_embedding_keys": null,
356
+ "mean_std_embedding_keys": null,
357
+ "action_configs": [
358
+ {
359
+ "rep": "RELATIVE",
360
+ "type": "EEF",
361
+ "format": "XYZ_ROT6D",
362
+ "state_key": "left_wrist_eef"
363
+ },
364
+ {
365
+ "rep": "RELATIVE",
366
+ "type": "EEF",
367
+ "format": "XYZ_ROT6D",
368
+ "state_key": "right_wrist_eef"
369
+ },
370
+ {
371
+ "rep": "ABSOLUTE",
372
+ "type": "NON_EEF",
373
+ "format": "DEFAULT",
374
+ "state_key": "left_hand_joints"
375
+ },
376
+ {
377
+ "rep": "ABSOLUTE",
378
+ "type": "NON_EEF",
379
+ "format": "DEFAULT",
380
+ "state_key": "right_hand_joints"
381
+ }
382
+ ]
383
+ },
384
+ "language": {
385
+ "delta_indices": [
386
+ 0
387
+ ],
388
+ "modality_keys": [
389
+ "annotation.human.coarse_action"
390
+ ],
391
+ "sin_cos_embedding_keys": null,
392
+ "mean_std_embedding_keys": null,
393
+ "action_configs": null
394
+ }
395
+ },
396
+ "real_r1_pro_sharpa_relative_eef": {
397
+ "video": {
398
+ "delta_indices": [
399
+ -20,
400
+ 0
401
+ ],
402
+ "modality_keys": [
403
+ "ego_view_res320x240_freq20",
404
+ "left_wrist_view_res320x240_freq20",
405
+ "right_wrist_view_res320x240_freq20"
406
+ ],
407
+ "sin_cos_embedding_keys": null,
408
+ "mean_std_embedding_keys": null,
409
+ "action_configs": null
410
+ },
411
+ "state": {
412
+ "delta_indices": [
413
+ 0
414
+ ],
415
+ "modality_keys": [
416
+ "left_wrist_eef",
417
+ "right_wrist_eef",
418
+ "left_hand_joints",
419
+ "right_hand_joints"
420
+ ],
421
+ "sin_cos_embedding_keys": null,
422
+ "mean_std_embedding_keys": null,
423
+ "action_configs": null
424
+ },
425
+ "action": {
426
+ "delta_indices": [
427
+ 0,
428
+ 1,
429
+ 2,
430
+ 3,
431
+ 4,
432
+ 5,
433
+ 6,
434
+ 7,
435
+ 8,
436
+ 9,
437
+ 10,
438
+ 11,
439
+ 12,
440
+ 13,
441
+ 14,
442
+ 15,
443
+ 16,
444
+ 17,
445
+ 18,
446
+ 19,
447
+ 20,
448
+ 21,
449
+ 22,
450
+ 23,
451
+ 24,
452
+ 25,
453
+ 26,
454
+ 27,
455
+ 28,
456
+ 29,
457
+ 30,
458
+ 31,
459
+ 32,
460
+ 33,
461
+ 34,
462
+ 35,
463
+ 36,
464
+ 37,
465
+ 38,
466
+ 39
467
+ ],
468
+ "modality_keys": [
469
+ "left_wrist_eef",
470
+ "right_wrist_eef",
471
+ "left_hand_joints",
472
+ "right_hand_joints"
473
+ ],
474
+ "sin_cos_embedding_keys": null,
475
+ "mean_std_embedding_keys": null,
476
+ "action_configs": [
477
+ {
478
+ "rep": "RELATIVE",
479
+ "type": "EEF",
480
+ "format": "XYZ_ROT6D",
481
+ "state_key": "left_wrist_eef"
482
+ },
483
+ {
484
+ "rep": "RELATIVE",
485
+ "type": "EEF",
486
+ "format": "XYZ_ROT6D",
487
+ "state_key": "right_wrist_eef"
488
+ },
489
+ {
490
+ "rep": "ABSOLUTE",
491
+ "type": "NON_EEF",
492
+ "format": "DEFAULT",
493
+ "state_key": "left_hand_joints"
494
+ },
495
+ {
496
+ "rep": "ABSOLUTE",
497
+ "type": "NON_EEF",
498
+ "format": "DEFAULT",
499
+ "state_key": "right_hand_joints"
500
+ }
501
+ ]
502
+ },
503
+ "language": {
504
+ "delta_indices": [
505
+ 0
506
+ ],
507
+ "modality_keys": [
508
+ "annotation.human.coarse_action"
509
+ ],
510
+ "sin_cos_embedding_keys": null,
511
+ "mean_std_embedding_keys": null,
512
+ "action_configs": null
513
+ }
514
+ },
515
+ "xdof_relative_eef_relative_joint": {
516
+ "video": {
517
+ "delta_indices": [
518
+ -30,
519
+ 0
520
+ ],
521
+ "modality_keys": [
522
+ "top_camera-images-rgb_320_240",
523
+ "left_camera-images-rgb_320_240",
524
+ "right_camera-images-rgb_320_240"
525
+ ],
526
+ "sin_cos_embedding_keys": null,
527
+ "mean_std_embedding_keys": null,
528
+ "action_configs": null
529
+ },
530
+ "state": {
531
+ "delta_indices": [
532
+ 0
533
+ ],
534
+ "modality_keys": [
535
+ "left_wrist_eef",
536
+ "right_wrist_eef",
537
+ "left_gripper_pos",
538
+ "right_gripper_pos",
539
+ "left_joint_pos",
540
+ "right_joint_pos"
541
+ ],
542
+ "sin_cos_embedding_keys": null,
543
+ "mean_std_embedding_keys": null,
544
+ "action_configs": null
545
+ },
546
+ "action": {
547
+ "delta_indices": [
548
+ 0,
549
+ 1,
550
+ 2,
551
+ 3,
552
+ 4,
553
+ 5,
554
+ 6,
555
+ 7,
556
+ 8,
557
+ 9,
558
+ 10,
559
+ 11,
560
+ 12,
561
+ 13,
562
+ 14,
563
+ 15,
564
+ 16,
565
+ 17,
566
+ 18,
567
+ 19,
568
+ 20,
569
+ 21,
570
+ 22,
571
+ 23,
572
+ 24,
573
+ 25,
574
+ 26,
575
+ 27,
576
+ 28,
577
+ 29,
578
+ 30,
579
+ 31,
580
+ 32,
581
+ 33,
582
+ 34,
583
+ 35,
584
+ 36,
585
+ 37,
586
+ 38,
587
+ 39
588
+ ],
589
+ "modality_keys": [
590
+ "left_wrist_eef",
591
+ "right_wrist_eef",
592
+ "left_gripper_pos",
593
+ "right_gripper_pos",
594
+ "left_joint_pos",
595
+ "right_joint_pos"
596
+ ],
597
+ "sin_cos_embedding_keys": null,
598
+ "mean_std_embedding_keys": null,
599
+ "action_configs": [
600
+ {
601
+ "rep": "RELATIVE",
602
+ "type": "EEF",
603
+ "format": "XYZ_ROT6D",
604
+ "state_key": "left_wrist_eef"
605
+ },
606
+ {
607
+ "rep": "RELATIVE",
608
+ "type": "EEF",
609
+ "format": "XYZ_ROT6D",
610
+ "state_key": "right_wrist_eef"
611
+ },
612
+ {
613
+ "rep": "ABSOLUTE",
614
+ "type": "NON_EEF",
615
+ "format": "DEFAULT",
616
+ "state_key": "left_gripper_pos"
617
+ },
618
+ {
619
+ "rep": "ABSOLUTE",
620
+ "type": "NON_EEF",
621
+ "format": "DEFAULT",
622
+ "state_key": "right_gripper_pos"
623
+ },
624
+ {
625
+ "rep": "RELATIVE",
626
+ "type": "NON_EEF",
627
+ "format": "DEFAULT",
628
+ "state_key": "left_joint_pos"
629
+ },
630
+ {
631
+ "rep": "RELATIVE",
632
+ "type": "NON_EEF",
633
+ "format": "DEFAULT",
634
+ "state_key": "right_joint_pos"
635
+ }
636
+ ]
637
+ },
638
+ "language": {
639
+ "delta_indices": [
640
+ 0
641
+ ],
642
+ "modality_keys": [
643
+ "annotation.task"
644
+ ],
645
+ "sin_cos_embedding_keys": null,
646
+ "mean_std_embedding_keys": null,
647
+ "action_configs": null
648
+ }
649
+ },
650
+ "real_r1_pro_sharpa_relative_eef_maxinsights": {
651
+ "video": {
652
+ "delta_indices": [
653
+ -30,
654
+ 0
655
+ ],
656
+ "modality_keys": [
657
+ "ego_view_cropratio_res320x240_freq30"
658
+ ],
659
+ "sin_cos_embedding_keys": null,
660
+ "mean_std_embedding_keys": null,
661
+ "action_configs": null
662
+ },
663
+ "state": {
664
+ "delta_indices": [
665
+ 0
666
+ ],
667
+ "modality_keys": [
668
+ "left_wrist_eef",
669
+ "right_wrist_eef",
670
+ "left_hand_joints",
671
+ "right_hand_joints"
672
+ ],
673
+ "sin_cos_embedding_keys": null,
674
+ "mean_std_embedding_keys": null,
675
+ "action_configs": null
676
+ },
677
+ "action": {
678
+ "delta_indices": [
679
+ 0,
680
+ 1,
681
+ 2,
682
+ 3,
683
+ 4,
684
+ 5,
685
+ 6,
686
+ 7,
687
+ 8,
688
+ 9,
689
+ 10,
690
+ 11,
691
+ 12,
692
+ 13,
693
+ 14,
694
+ 15,
695
+ 16,
696
+ 17,
697
+ 18,
698
+ 19,
699
+ 20,
700
+ 21,
701
+ 22,
702
+ 23,
703
+ 24,
704
+ 25,
705
+ 26,
706
+ 27,
707
+ 28,
708
+ 29,
709
+ 30,
710
+ 31,
711
+ 32,
712
+ 33,
713
+ 34,
714
+ 35,
715
+ 36,
716
+ 37,
717
+ 38,
718
+ 39
719
+ ],
720
+ "modality_keys": [
721
+ "left_wrist_eef",
722
+ "right_wrist_eef",
723
+ "left_hand_joints",
724
+ "right_hand_joints"
725
+ ],
726
+ "sin_cos_embedding_keys": null,
727
+ "mean_std_embedding_keys": null,
728
+ "action_configs": [
729
+ {
730
+ "rep": "RELATIVE",
731
+ "type": "EEF",
732
+ "format": "XYZ_ROT6D",
733
+ "state_key": "left_wrist_eef"
734
+ },
735
+ {
736
+ "rep": "RELATIVE",
737
+ "type": "EEF",
738
+ "format": "XYZ_ROT6D",
739
+ "state_key": "right_wrist_eef"
740
+ },
741
+ {
742
+ "rep": "ABSOLUTE",
743
+ "type": "NON_EEF",
744
+ "format": "DEFAULT",
745
+ "state_key": "left_hand_joints"
746
+ },
747
+ {
748
+ "rep": "ABSOLUTE",
749
+ "type": "NON_EEF",
750
+ "format": "DEFAULT",
751
+ "state_key": "right_hand_joints"
752
+ }
753
+ ]
754
+ },
755
+ "language": {
756
+ "delta_indices": [
757
+ 0
758
+ ],
759
+ "modality_keys": [
760
+ "annotation.human.coarse_action"
761
+ ],
762
+ "sin_cos_embedding_keys": null,
763
+ "mean_std_embedding_keys": null,
764
+ "action_configs": null
765
+ }
766
+ },
767
+ "xdof_relative_eef_relative_joint_subtask": {
768
+ "video": {
769
+ "delta_indices": [
770
+ -30,
771
+ 0
772
+ ],
773
+ "modality_keys": [
774
+ "top_camera-images-rgb_320_240",
775
+ "left_camera-images-rgb_320_240",
776
+ "right_camera-images-rgb_320_240"
777
+ ],
778
+ "sin_cos_embedding_keys": null,
779
+ "mean_std_embedding_keys": null,
780
+ "action_configs": null
781
+ },
782
+ "state": {
783
+ "delta_indices": [
784
+ 0
785
+ ],
786
+ "modality_keys": [
787
+ "left_wrist_eef",
788
+ "right_wrist_eef",
789
+ "left_gripper_pos",
790
+ "right_gripper_pos",
791
+ "left_joint_pos",
792
+ "right_joint_pos"
793
+ ],
794
+ "sin_cos_embedding_keys": null,
795
+ "mean_std_embedding_keys": null,
796
+ "action_configs": null
797
+ },
798
+ "action": {
799
+ "delta_indices": [
800
+ 0,
801
+ 1,
802
+ 2,
803
+ 3,
804
+ 4,
805
+ 5,
806
+ 6,
807
+ 7,
808
+ 8,
809
+ 9,
810
+ 10,
811
+ 11,
812
+ 12,
813
+ 13,
814
+ 14,
815
+ 15,
816
+ 16,
817
+ 17,
818
+ 18,
819
+ 19,
820
+ 20,
821
+ 21,
822
+ 22,
823
+ 23,
824
+ 24,
825
+ 25,
826
+ 26,
827
+ 27,
828
+ 28,
829
+ 29,
830
+ 30,
831
+ 31,
832
+ 32,
833
+ 33,
834
+ 34,
835
+ 35,
836
+ 36,
837
+ 37,
838
+ 38,
839
+ 39
840
+ ],
841
+ "modality_keys": [
842
+ "left_wrist_eef",
843
+ "right_wrist_eef",
844
+ "left_gripper_pos",
845
+ "right_gripper_pos",
846
+ "left_joint_pos",
847
+ "right_joint_pos"
848
+ ],
849
+ "sin_cos_embedding_keys": null,
850
+ "mean_std_embedding_keys": null,
851
+ "action_configs": [
852
+ {
853
+ "rep": "RELATIVE",
854
+ "type": "EEF",
855
+ "format": "XYZ_ROT6D",
856
+ "state_key": "left_wrist_eef"
857
+ },
858
+ {
859
+ "rep": "RELATIVE",
860
+ "type": "EEF",
861
+ "format": "XYZ_ROT6D",
862
+ "state_key": "right_wrist_eef"
863
+ },
864
+ {
865
+ "rep": "ABSOLUTE",
866
+ "type": "NON_EEF",
867
+ "format": "DEFAULT",
868
+ "state_key": "left_gripper_pos"
869
+ },
870
+ {
871
+ "rep": "ABSOLUTE",
872
+ "type": "NON_EEF",
873
+ "format": "DEFAULT",
874
+ "state_key": "right_gripper_pos"
875
+ },
876
+ {
877
+ "rep": "RELATIVE",
878
+ "type": "NON_EEF",
879
+ "format": "DEFAULT",
880
+ "state_key": "left_joint_pos"
881
+ },
882
+ {
883
+ "rep": "RELATIVE",
884
+ "type": "NON_EEF",
885
+ "format": "DEFAULT",
886
+ "state_key": "right_joint_pos"
887
+ }
888
+ ]
889
+ },
890
+ "language": {
891
+ "delta_indices": [
892
+ 0
893
+ ],
894
+ "modality_keys": [
895
+ "annotation.sub_task"
896
+ ],
897
+ "sin_cos_embedding_keys": null,
898
+ "mean_std_embedding_keys": null,
899
+ "action_configs": null
900
+ }
901
+ },
902
+ "oxe_droid_relative_eef_relative_joint": {
903
+ "video": {
904
+ "delta_indices": [
905
+ -15,
906
+ 0
907
+ ],
908
+ "modality_keys": [
909
+ "exterior_image_1_left",
910
+ "wrist_image_left"
911
+ ],
912
+ "sin_cos_embedding_keys": null,
913
+ "mean_std_embedding_keys": null,
914
+ "action_configs": null
915
+ },
916
+ "state": {
917
+ "delta_indices": [
918
+ 0
919
+ ],
920
+ "modality_keys": [
921
+ "eef_9d",
922
+ "gripper_position",
923
+ "joint_position"
924
+ ],
925
+ "sin_cos_embedding_keys": null,
926
+ "mean_std_embedding_keys": null,
927
+ "action_configs": null
928
+ },
929
+ "action": {
930
+ "delta_indices": [
931
+ 0,
932
+ 1,
933
+ 2,
934
+ 3,
935
+ 4,
936
+ 5,
937
+ 6,
938
+ 7,
939
+ 8,
940
+ 9,
941
+ 10,
942
+ 11,
943
+ 12,
944
+ 13,
945
+ 14,
946
+ 15,
947
+ 16,
948
+ 17,
949
+ 18,
950
+ 19,
951
+ 20,
952
+ 21,
953
+ 22,
954
+ 23,
955
+ 24,
956
+ 25,
957
+ 26,
958
+ 27,
959
+ 28,
960
+ 29,
961
+ 30,
962
+ 31,
963
+ 32,
964
+ 33,
965
+ 34,
966
+ 35,
967
+ 36,
968
+ 37,
969
+ 38,
970
+ 39
971
+ ],
972
+ "modality_keys": [
973
+ "eef_9d",
974
+ "gripper_position",
975
+ "joint_position"
976
+ ],
977
+ "sin_cos_embedding_keys": null,
978
+ "mean_std_embedding_keys": null,
979
+ "action_configs": [
980
+ {
981
+ "rep": "RELATIVE",
982
+ "type": "EEF",
983
+ "format": "XYZ_ROT6D",
984
+ "state_key": "eef_9d"
985
+ },
986
+ {
987
+ "rep": "ABSOLUTE",
988
+ "type": "NON_EEF",
989
+ "format": "DEFAULT",
990
+ "state_key": "gripper_position"
991
+ },
992
+ {
993
+ "rep": "RELATIVE",
994
+ "type": "NON_EEF",
995
+ "format": "DEFAULT",
996
+ "state_key": "joint_position"
997
+ }
998
+ ]
999
+ },
1000
+ "language": {
1001
+ "delta_indices": [
1002
+ 0
1003
+ ],
1004
+ "modality_keys": [
1005
+ "annotation.language.language_instruction"
1006
+ ],
1007
+ "sin_cos_embedding_keys": null,
1008
+ "mean_std_embedding_keys": null,
1009
+ "action_configs": null
1010
+ }
1011
+ },
1012
+ "new_embodiment": {
1013
+ "video": {
1014
+ "delta_indices": [
1015
+ 0
1016
+ ],
1017
+ "modality_keys": [
1018
+ "head_left",
1019
+ "head_right",
1020
+ "hand_left",
1021
+ "hand_right"
1022
+ ],
1023
+ "sin_cos_embedding_keys": null,
1024
+ "mean_std_embedding_keys": null,
1025
+ "action_configs": null
1026
+ },
1027
+ "state": {
1028
+ "delta_indices": [
1029
+ 0
1030
+ ],
1031
+ "modality_keys": [
1032
+ "left_arm",
1033
+ "right_arm",
1034
+ "left_hand",
1035
+ "right_hand",
1036
+ "head",
1037
+ "waist"
1038
+ ],
1039
+ "sin_cos_embedding_keys": null,
1040
+ "mean_std_embedding_keys": null,
1041
+ "action_configs": null
1042
+ },
1043
+ "action": {
1044
+ "delta_indices": [
1045
+ 0,
1046
+ 1,
1047
+ 2,
1048
+ 3,
1049
+ 4,
1050
+ 5,
1051
+ 6,
1052
+ 7,
1053
+ 8,
1054
+ 9,
1055
+ 10,
1056
+ 11,
1057
+ 12,
1058
+ 13,
1059
+ 14,
1060
+ 15
1061
+ ],
1062
+ "modality_keys": [
1063
+ "left_arm",
1064
+ "right_arm",
1065
+ "left_hand",
1066
+ "right_hand",
1067
+ "head",
1068
+ "waist",
1069
+ "mob_comp",
1070
+ "task_goal"
1071
+ ],
1072
+ "sin_cos_embedding_keys": null,
1073
+ "mean_std_embedding_keys": null,
1074
+ "action_configs": [
1075
+ {
1076
+ "rep": "DELTA",
1077
+ "type": "NON_EEF",
1078
+ "format": "DEFAULT",
1079
+ "state_key": null
1080
+ },
1081
+ {
1082
+ "rep": "DELTA",
1083
+ "type": "NON_EEF",
1084
+ "format": "DEFAULT",
1085
+ "state_key": null
1086
+ },
1087
+ {
1088
+ "rep": "DELTA",
1089
+ "type": "NON_EEF",
1090
+ "format": "DEFAULT",
1091
+ "state_key": null
1092
+ },
1093
+ {
1094
+ "rep": "DELTA",
1095
+ "type": "NON_EEF",
1096
+ "format": "DEFAULT",
1097
+ "state_key": null
1098
+ },
1099
+ {
1100
+ "rep": "DELTA",
1101
+ "type": "NON_EEF",
1102
+ "format": "DEFAULT",
1103
+ "state_key": null
1104
+ },
1105
+ {
1106
+ "rep": "DELTA",
1107
+ "type": "NON_EEF",
1108
+ "format": "DEFAULT",
1109
+ "state_key": null
1110
+ },
1111
+ {
1112
+ "rep": "DELTA",
1113
+ "type": "NON_EEF",
1114
+ "format": "DEFAULT",
1115
+ "state_key": null
1116
+ },
1117
+ {
1118
+ "rep": "DELTA",
1119
+ "type": "NON_EEF",
1120
+ "format": "DEFAULT",
1121
+ "state_key": null
1122
+ }
1123
+ ]
1124
+ },
1125
+ "language": {
1126
+ "delta_indices": [
1127
+ 0
1128
+ ],
1129
+ "modality_keys": [
1130
+ "annotation.human.task_description"
1131
+ ],
1132
+ "sin_cos_embedding_keys": null,
1133
+ "mean_std_embedding_keys": null,
1134
+ "action_configs": null
1135
+ }
1136
+ }
1137
+ },
1138
+ "image_crop_size": [
1139
+ 230,
1140
+ 230
1141
+ ],
1142
+ "image_target_size": [
1143
+ 256,
1144
+ 256
1145
+ ],
1146
+ "use_albumentations": true,
1147
+ "random_rotation_angle": 0,
1148
+ "color_jitter_params": {
1149
+ "brightness": 0.3,
1150
+ "contrast": 0.4,
1151
+ "saturation": 0.5,
1152
+ "hue": 0.08
1153
+ },
1154
+ "shortest_image_edge": 256,
1155
+ "crop_fraction": 0.95,
1156
+ "letter_box_transform": false,
1157
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1158
+ "model_type": "qwen",
1159
+ "formalize_language": true,
1160
+ "max_state_dim": 132,
1161
+ "max_action_dim": 132,
1162
+ "max_action_horizon": 40,
1163
+ "use_percentiles": true,
1164
+ "use_mean_std": false,
1165
+ "clip_outliers": true,
1166
+ "apply_sincos_state_encoding": false,
1167
+ "use_relative_action": true,
1168
+ "exclude_state": false,
1169
+ "state_dropout_prob": 0.2
1170
+ }
1171
+ }
processor/statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c06c6312892d4f8856e3cf36eda5c6087283cb9454423c024dd03521616ec9db
3
+ size 5841
wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d7", "run_id": "daru_mob_comp_minimum_0715"}