Hugging Face's logo

RobotisSW
/
Fixed_PickGreenCoffeeCan_WhiteShelf_byGrootModel_80K

Safetensors
Gr00tN1d7
Model card Files
xet
 Community
RobotisSW commited on
Commit
c7ab751
·
verified ·
1 Parent(s): 0472a2d

Upload folder using huggingface_hub

Browse files
Files changed (18) hide show
  1. config.json +88 -0
  2. embodiment_id.json +57 -0
  3. experiment_cfg/conf.yaml +237 -0
  4. experiment_cfg/config.yaml +262 -0
  5. experiment_cfg/dataset_statistics.json +257 -0
  6. experiment_cfg/final_model_config.json +55 -0
  7. experiment_cfg/final_processor_config.json +0 -0
  8. model-00001-of-00002.safetensors +3 -0
  9. model-00002-of-00002.safetensors +3 -0
  10. model.safetensors.index.json +0 -0
  11. optimizer.pt +3 -0
  12. processor_config.json +1146 -0
  13. rng_state.pth +3 -0
  14. scheduler.pt +3 -0
  15. statistics.json +0 -0
  16. trainer_state.json +0 -0
  17. training_args.bin +3 -0
  18. wandb_config.json +1 -0
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": "bfloat16",
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": true,
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
+ }
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
+ "unitree_g1_sonic": 11,
55
+ "new_embodiment": 10,
56
+ "robocasa_gr1_tabletop": 10
57
+ }
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,237 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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: true
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
+ extra_augmentation_config: null
29
+ formalize_language: true
30
+ apply_sincos_state_encoding: false
31
+ use_percentiles: true
32
+ use_relative_action: true
33
+ max_state_dim: 132
34
+ max_action_dim: 132
35
+ action_horizon: 40
36
+ hidden_size: 1024
37
+ input_embedding_dim: 1536
38
+ state_history_length: 1
39
+ add_pos_embed: true
40
+ attn_dropout: 0.2
41
+ use_vlln: true
42
+ max_seq_len: 1024
43
+ use_alternate_vl_dit: true
44
+ attend_text_every_n_blocks: 2
45
+ diffusion_model_cfg:
46
+ positional_embeddings: null
47
+ num_layers: 16
48
+ num_attention_heads: 32
49
+ attention_head_dim: 48
50
+ norm_type: ada_norm
51
+ dropout: 0.2
52
+ final_dropout: true
53
+ output_dim: 1024
54
+ interleave_self_attention: true
55
+ num_inference_timesteps: 4
56
+ noise_beta_alpha: 1.5
57
+ noise_beta_beta: 1.0
58
+ noise_s: 0.999
59
+ num_timestep_buckets: 1000
60
+ tune_projector: true
61
+ tune_diffusion_model: true
62
+ tune_vlln: true
63
+ state_dropout_prob: 0.2
64
+ exclude_state: false
65
+ use_mean_std: false
66
+ max_num_embodiments: 32
67
+ data:
68
+ datasets:
69
+ - dataset_paths:
70
+ - /root/groot_workspace/datasets/Task_000373_000379_Pick_GreenCoffeeCan_WhiteShelf_Lift0_KHM_lerobot
71
+ embodiment_tag: new_embodiment
72
+ mix_ratio: 1.0
73
+ dataset_type: physical_embodiment
74
+ val_dataset_path: null
75
+ modality_configs:
76
+ new_embodiment:
77
+ video:
78
+ delta_indices:
79
+ - 0
80
+ modality_keys:
81
+ - cam_left_head
82
+ sin_cos_embedding_keys: null
83
+ mean_std_embedding_keys: null
84
+ action_configs: null
85
+ state:
86
+ delta_indices:
87
+ - 0
88
+ modality_keys:
89
+ - arm_left
90
+ - arm_right
91
+ sin_cos_embedding_keys: null
92
+ mean_std_embedding_keys: null
93
+ action_configs: null
94
+ action:
95
+ delta_indices:
96
+ - 0
97
+ - 1
98
+ - 2
99
+ - 3
100
+ - 4
101
+ - 5
102
+ - 6
103
+ - 7
104
+ - 8
105
+ - 9
106
+ - 10
107
+ - 11
108
+ - 12
109
+ - 13
110
+ - 14
111
+ - 15
112
+ - 16
113
+ - 17
114
+ - 18
115
+ - 19
116
+ - 20
117
+ - 21
118
+ - 22
119
+ - 23
120
+ - 24
121
+ - 25
122
+ - 26
123
+ - 27
124
+ - 28
125
+ - 29
126
+ - 30
127
+ - 31
128
+ - 32
129
+ - 33
130
+ - 34
131
+ - 35
132
+ - 36
133
+ - 37
134
+ - 38
135
+ - 39
136
+ modality_keys:
137
+ - arm_left
138
+ - arm_right
139
+ sin_cos_embedding_keys: null
140
+ mean_std_embedding_keys: null
141
+ action_configs:
142
+ - rep: ABSOLUTE
143
+ type: NON_EEF
144
+ format: DEFAULT
145
+ state_key: null
146
+ - rep: ABSOLUTE
147
+ type: NON_EEF
148
+ format: DEFAULT
149
+ state_key: null
150
+ language:
151
+ delta_indices:
152
+ - 0
153
+ modality_keys:
154
+ - sub_task
155
+ sin_cos_embedding_keys: null
156
+ mean_std_embedding_keys: null
157
+ action_configs: null
158
+ download_cache: false
159
+ shard_size: 1024
160
+ episode_sampling_rate: 0.1
161
+ num_shards_per_epoch: 100000
162
+ override_pretraining_statistics: true
163
+ mode: single_turn
164
+ random_chop: 0.0
165
+ mock_dataset_mode: false
166
+ shuffle: true
167
+ seed: 42
168
+ multiprocessing_context: fork
169
+ allow_padding: false
170
+ subsample_ratio: 1.0
171
+ image_crop_size:
172
+ - 244
173
+ - 244
174
+ image_target_size:
175
+ - 224
176
+ - 224
177
+ video_backend: torchcodec
178
+ training:
179
+ output_dir: /root/groot_workspace/checkpoints/Groot-n17-Pick-GreenCoffeeCan_WhiteShelf_Groot_2
180
+ experiment_name: null
181
+ max_steps: 100000
182
+ global_batch_size: 64
183
+ batch_size: null
184
+ gradient_accumulation_steps: 1
185
+ learning_rate: 0.0001
186
+ lr_scheduler_type: cosine
187
+ weight_decay: 1.0e-05
188
+ warmup_ratio: 0.05
189
+ warmup_steps: 0
190
+ max_grad_norm: 1.0
191
+ optim: adamw_torch
192
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
193
+ skip_weight_loading: false
194
+ vlm_backbone_checkpoint_path: null
195
+ tf32: true
196
+ fp16: false
197
+ bf16: true
198
+ eval_bf16: true
199
+ logging_steps: 10
200
+ save_steps: 20000
201
+ save_total_limit: 5
202
+ save_vl_model: false
203
+ save_only_model: false
204
+ upload_checkpoints: false
205
+ upload_every: 1000
206
+ upload_last_n_checkpoints: 5
207
+ max_concurrent_uploads: 2
208
+ eval_strategy: 'no'
209
+ eval_steps: 500
210
+ eval_set_split_ratio: 0.1
211
+ eval_batch_size: 2
212
+ save_best_eval_metric_name: ''
213
+ save_best_eval_metric_greater_is_better: true
214
+ deepspeed_stage: 2
215
+ gradient_checkpointing: false
216
+ transformers_trust_remote_code: true
217
+ transformers_local_files_only: false
218
+ transformers_cache_dir: null
219
+ transformers_access_token: null
220
+ use_ddp: false
221
+ ddp_bucket_cap_mb: 100
222
+ num_gpus: 1
223
+ dataloader_num_workers: 8
224
+ remove_unused_columns: false
225
+ use_wandb: false
226
+ wandb_project: finetune-gr00t-n1d7
227
+ enable_profiling: false
228
+ max_retries: 3
229
+ assert_loss_less_than: null
230
+ add_rl_callback: false
231
+ enable_open_loop_eval: false
232
+ open_loop_eval_traj_ids:
233
+ - 0
234
+ open_loop_eval_steps_per_traj: 100
235
+ open_loop_eval_plot_indices: null
236
+ max_steps: 100000
237
+ save_steps: 20000
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,262 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ !!python/object:gr00t.configs.base_config.Config
2
+ data: !!python/object:gr00t.configs.data.data_config.DataConfig
3
+ allow_padding: false
4
+ datasets:
5
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
6
+ dataset_paths:
7
+ - /root/groot_workspace/datasets/Task_000373_000379_Pick_GreenCoffeeCan_WhiteShelf_Lift0_KHM_lerobot
8
+ dataset_type: physical_embodiment
9
+ embodiment_tag: new_embodiment
10
+ mix_ratio: 1.0
11
+ val_dataset_path: null
12
+ download_cache: false
13
+ episode_sampling_rate: 0.1
14
+ image_crop_size:
15
+ - 244
16
+ - 244
17
+ image_target_size:
18
+ - 224
19
+ - 224
20
+ mock_dataset_mode: false
21
+ modality_configs:
22
+ new_embodiment:
23
+ action: !!python/object:gr00t.data.types.ModalityConfig
24
+ action_configs:
25
+ - !!python/object:gr00t.data.types.ActionConfig
26
+ format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
27
+ - default
28
+ rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
29
+ - absolute
30
+ state_key: null
31
+ type: &id003 !!python/object/apply:gr00t.data.types.ActionType
32
+ - non_eef
33
+ - !!python/object:gr00t.data.types.ActionConfig
34
+ format: *id001
35
+ rep: *id002
36
+ state_key: null
37
+ type: *id003
38
+ delta_indices:
39
+ - 0
40
+ - 1
41
+ - 2
42
+ - 3
43
+ - 4
44
+ - 5
45
+ - 6
46
+ - 7
47
+ - 8
48
+ - 9
49
+ - 10
50
+ - 11
51
+ - 12
52
+ - 13
53
+ - 14
54
+ - 15
55
+ - 16
56
+ - 17
57
+ - 18
58
+ - 19
59
+ - 20
60
+ - 21
61
+ - 22
62
+ - 23
63
+ - 24
64
+ - 25
65
+ - 26
66
+ - 27
67
+ - 28
68
+ - 29
69
+ - 30
70
+ - 31
71
+ - 32
72
+ - 33
73
+ - 34
74
+ - 35
75
+ - 36
76
+ - 37
77
+ - 38
78
+ - 39
79
+ mean_std_embedding_keys: null
80
+ modality_keys:
81
+ - arm_left
82
+ - arm_right
83
+ sin_cos_embedding_keys: null
84
+ language: !!python/object:gr00t.data.types.ModalityConfig
85
+ action_configs: null
86
+ delta_indices:
87
+ - 0
88
+ mean_std_embedding_keys: null
89
+ modality_keys:
90
+ - sub_task
91
+ sin_cos_embedding_keys: null
92
+ state: !!python/object:gr00t.data.types.ModalityConfig
93
+ action_configs: null
94
+ delta_indices:
95
+ - 0
96
+ mean_std_embedding_keys: null
97
+ modality_keys:
98
+ - arm_left
99
+ - arm_right
100
+ sin_cos_embedding_keys: null
101
+ video: !!python/object:gr00t.data.types.ModalityConfig
102
+ action_configs: null
103
+ delta_indices:
104
+ - 0
105
+ mean_std_embedding_keys: null
106
+ modality_keys:
107
+ - cam_left_head
108
+ sin_cos_embedding_keys: null
109
+ mode: single_turn
110
+ multiprocessing_context: fork
111
+ num_shards_per_epoch: 100000
112
+ override_pretraining_statistics: true
113
+ random_chop: 0.0
114
+ seed: 42
115
+ shard_size: 1024
116
+ shuffle: true
117
+ subsample_ratio: 1.0
118
+ video_backend: torchcodec
119
+ load_config_path: null
120
+ model: !!python/object:gr00t.configs.model.gr00t_n1d7.Gr00tN1d7Config
121
+ _attn_implementation_internal: null
122
+ _commit_hash: null
123
+ _name_or_path: ''
124
+ _output_attentions: false
125
+ add_cross_attention: false
126
+ architectures: null
127
+ backbone_trainable_params_fp32: true
128
+ bad_words_ids: null
129
+ begin_suppress_tokens: null
130
+ bos_token_id: null
131
+ chunk_size_feed_forward: 0
132
+ color_jitter_params: null
133
+ cross_attention_hidden_size: null
134
+ decoder_start_token_id: null
135
+ diffusion_model_cfg:
136
+ attention_head_dim: 48
137
+ dropout: 0.2
138
+ final_dropout: true
139
+ interleave_self_attention: true
140
+ norm_type: ada_norm
141
+ num_attention_heads: 32
142
+ num_layers: 16
143
+ output_dim: 1024
144
+ positional_embeddings: null
145
+ diversity_penalty: 0.0
146
+ do_sample: false
147
+ dtype: null
148
+ early_stopping: false
149
+ encoder_no_repeat_ngram_size: 0
150
+ eos_token_id: null
151
+ exponential_decay_length_penalty: null
152
+ extra_augmentation_config: null
153
+ finetuning_task: null
154
+ forced_bos_token_id: null
155
+ forced_eos_token_id: null
156
+ id2label:
157
+ 0: LABEL_0
158
+ 1: LABEL_1
159
+ is_decoder: false
160
+ is_encoder_decoder: false
161
+ label2id:
162
+ LABEL_0: 0
163
+ LABEL_1: 1
164
+ length_penalty: 1.0
165
+ load_bf16: true
166
+ max_length: 20
167
+ min_length: 0
168
+ model_name: nvidia/Cosmos-Reason2-2B
169
+ no_repeat_ngram_size: 0
170
+ num_beam_groups: 1
171
+ num_beams: 1
172
+ num_return_sequences: 1
173
+ output_hidden_states: false
174
+ output_scores: false
175
+ pad_token_id: null
176
+ prefix: null
177
+ problem_type: null
178
+ pruned_heads: {}
179
+ random_rotation_angle: null
180
+ remove_invalid_values: false
181
+ repetition_penalty: 1.0
182
+ reproject_vision: false
183
+ return_dict: true
184
+ return_dict_in_generate: false
185
+ sep_token_id: null
186
+ state_dropout_prob: 0.2
187
+ suppress_tokens: null
188
+ task_specific_params: null
189
+ temperature: 1.0
190
+ tf_legacy_loss: false
191
+ tie_encoder_decoder: false
192
+ tie_word_embeddings: true
193
+ tokenizer_class: null
194
+ top_k: 50
195
+ top_p: 1.0
196
+ torchscript: false
197
+ transformers_version: null
198
+ tune_diffusion_model: true
199
+ tune_llm: false
200
+ tune_projector: true
201
+ tune_visual: false
202
+ typical_p: 1.0
203
+ use_bfloat16: false
204
+ use_relative_action: true
205
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
206
+ add_rl_callback: false
207
+ assert_loss_less_than: null
208
+ batch_size: null
209
+ bf16: true
210
+ dataloader_num_workers: 8
211
+ ddp_bucket_cap_mb: 100
212
+ deepspeed_stage: 2
213
+ enable_open_loop_eval: false
214
+ enable_profiling: false
215
+ eval_batch_size: 2
216
+ eval_bf16: true
217
+ eval_set_split_ratio: 0.1
218
+ eval_steps: 500
219
+ eval_strategy: 'no'
220
+ experiment_name: null
221
+ fp16: false
222
+ global_batch_size: 64
223
+ gradient_accumulation_steps: 1
224
+ gradient_checkpointing: false
225
+ learning_rate: 0.0001
226
+ logging_steps: 10
227
+ lr_scheduler_type: cosine
228
+ max_concurrent_uploads: 2
229
+ max_grad_norm: 1.0
230
+ max_retries: 3
231
+ max_steps: 100000
232
+ num_gpus: 1
233
+ open_loop_eval_plot_indices: null
234
+ open_loop_eval_steps_per_traj: 100
235
+ open_loop_eval_traj_ids:
236
+ - 0
237
+ optim: adamw_torch
238
+ output_dir: /root/groot_workspace/checkpoints/Groot-n17-Pick-GreenCoffeeCan_WhiteShelf_Groot_2
239
+ remove_unused_columns: false
240
+ save_best_eval_metric_greater_is_better: true
241
+ save_best_eval_metric_name: ''
242
+ save_only_model: false
243
+ save_steps: 20000
244
+ save_total_limit: 5
245
+ save_vl_model: false
246
+ skip_weight_loading: false
247
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
248
+ tf32: true
249
+ transformers_access_token: null
250
+ transformers_cache_dir: null
251
+ transformers_local_files_only: false
252
+ transformers_trust_remote_code: true
253
+ upload_checkpoints: false
254
+ upload_every: 1000
255
+ upload_last_n_checkpoints: 5
256
+ use_ddp: false
257
+ use_wandb: false
258
+ vlm_backbone_checkpoint_path: null
259
+ wandb_project: finetune-gr00t-n1d7
260
+ warmup_ratio: 0.05
261
+ warmup_steps: 0
262
+ weight_decay: 1.0e-05
experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,257 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "arm_left": {
5
+ "min": [
6
+ -1.0179400444030762,
7
+ -0.000107858024421148,
8
+ -0.2951115369796753,
9
+ -2.9208791255950928,
10
+ -0.14306767284870148,
11
+ 0.5229675769805908,
12
+ -0.6291666030883789,
13
+ -0.0015339808305725455
14
+ ],
15
+ "max": [
16
+ 0.12697286903858185,
17
+ 0.36734047532081604,
18
+ 0.6430495381355286,
19
+ -1.190009593963623,
20
+ 1.1268287897109985,
21
+ 1.4312280416488647,
22
+ 0.5737504959106445,
23
+ 1.0906603336334229
24
+ ],
25
+ "mean": [
26
+ -0.08349303901195526,
27
+ 0.2696293294429779,
28
+ 0.38212642073631287,
29
+ -2.4406867027282715,
30
+ 0.5138493776321411,
31
+ 0.9316314458847046,
32
+ -0.44350045919418335,
33
+ 0.1829506754875183
34
+ ],
35
+ "std": [
36
+ 0.2761023938655853,
37
+ 0.11780083179473877,
38
+ 0.2242833524942398,
39
+ 0.25472286343574524,
40
+ 0.18124833703041077,
41
+ 0.12002525478601432,
42
+ 0.29895201325416565,
43
+ 0.3405115008354187
44
+ ],
45
+ "q01": [
46
+ -0.9474626272916794,
47
+ 0.0,
48
+ -0.19774031341075896,
49
+ -2.792319691181183,
50
+ -0.007867044559679922,
51
+ 0.6268762230873108,
52
+ -0.6280909776687622,
53
+ 0.0
54
+ ],
55
+ "q99": [
56
+ 0.0659252181649208,
57
+ 0.3420897126197815,
58
+ 0.5122896432876587,
59
+ -1.368415719270706,
60
+ 0.6547455549240113,
61
+ 1.289342600107193,
62
+ 0.3304298728704454,
63
+ 0.9986214637756348
64
+ ]
65
+ },
66
+ "arm_right": {
67
+ "min": [
68
+ -1.328391432762146,
69
+ -0.9925814270973206,
70
+ -0.6948214173316956,
71
+ -2.8949573040008545,
72
+ -1.3027331829071045,
73
+ 0.16631707549095154,
74
+ -1.3498246669769287,
75
+ 0.0015339808305725455
76
+ ],
77
+ "max": [
78
+ 0.1341274380683899,
79
+ 0.00014381069922819734,
80
+ 1.0354130268096924,
81
+ -1.039931058883667,
82
+ 0.0935608446598053,
83
+ 1.292762279510498,
84
+ 0.7704581618309021,
85
+ 1.0937283039093018
86
+ ],
87
+ "mean": [
88
+ -0.21940943598747253,
89
+ -0.27110999822616577,
90
+ -0.1844315081834793,
91
+ -2.3738045692443848,
92
+ -0.546974241733551,
93
+ 0.8536734580993652,
94
+ 0.19668342173099518,
95
+ 0.3265850841999054
96
+ ],
97
+ "std": [
98
+ 0.3627884089946747,
99
+ 0.15213322639465332,
100
+ 0.4376094341278076,
101
+ 0.3168535828590393,
102
+ 0.2433137744665146,
103
+ 0.1658935248851776,
104
+ 0.5454496145248413,
105
+ 0.4119211435317993
106
+ ],
107
+ "q01": [
108
+ -1.1754349768161774,
109
+ -0.7025392055511475,
110
+ -0.5733924269676208,
111
+ -2.746969485282898,
112
+ -1.0947781145572661,
113
+ 0.3063940778374672,
114
+ -1.156208485364914,
115
+ 0.0015339808305725455
116
+ ],
117
+ "q99": [
118
+ 0.0659012496471405,
119
+ 1.1984225238848012e-05,
120
+ 0.9540114164352418,
121
+ -1.2299230217933652,
122
+ -0.027975374646484835,
123
+ 1.1693840801715851,
124
+ 0.6283562839031219,
125
+ 1.0461748838424683
126
+ ]
127
+ }
128
+ },
129
+ "action": {
130
+ "arm_left": {
131
+ "min": [
132
+ -1.0185632705688477,
133
+ 0.0,
134
+ -0.2989911735057831,
135
+ -2.9251444339752197,
136
+ -0.14658840000629425,
137
+ 0.5230874419212341,
138
+ -0.6293507218360901,
139
+ 0.0
140
+ ],
141
+ "max": [
142
+ 0.12857358157634735,
143
+ 0.36815065145492554,
144
+ 0.6476532816886902,
145
+ -1.1888351440429688,
146
+ 1.1319832801818848,
147
+ 1.431204080581665,
148
+ 0.5737088322639465,
149
+ 1.100000023841858
150
+ ],
151
+ "mean": [
152
+ -0.08348958194255829,
153
+ 0.26964664459228516,
154
+ 0.3821379542350769,
155
+ -2.440678834915161,
156
+ 0.5138611197471619,
157
+ 0.9316415786743164,
158
+ -0.44351497292518616,
159
+ 0.22973352670669556
160
+ ],
161
+ "std": [
162
+ 0.27622321248054504,
163
+ 0.11787990480661392,
164
+ 0.22440168261528015,
165
+ 0.25500747561454773,
166
+ 0.18135865032672882,
167
+ 0.12023748457431793,
168
+ 0.298767626285553,
169
+ 0.4284822642803192
170
+ ],
171
+ "q01": [
172
+ -0.9480001330375671,
173
+ 0.0,
174
+ -0.19941750168800354,
175
+ -2.794218695163727,
176
+ -0.008726346166804401,
177
+ 0.6264010429382325,
178
+ -0.6280999779701233,
179
+ 0.0
180
+ ],
181
+ "q99": [
182
+ 0.06589999794960022,
183
+ 0.34209999442100525,
184
+ 0.5123000144958496,
185
+ -1.368310809135437,
186
+ 0.6551887005567552,
187
+ 1.290077805519104,
188
+ 0.3309912174940124,
189
+ 1.100000023841858
190
+ ]
191
+ },
192
+ "arm_right": {
193
+ "min": [
194
+ -1.3284273147583008,
195
+ -0.9980451464653015,
196
+ -0.6979562640190125,
197
+ -2.897864580154419,
198
+ -1.3056524991989136,
199
+ 0.16566991806030273,
200
+ -1.3499031066894531,
201
+ 0.0
202
+ ],
203
+ "max": [
204
+ 0.13472957909107208,
205
+ 0.0,
206
+ 1.035436987876892,
207
+ -1.0400389432907104,
208
+ 0.09357282519340515,
209
+ 1.2962563037872314,
210
+ 0.7700583338737488,
211
+ 1.100000023841858
212
+ ],
213
+ "mean": [
214
+ -0.21939951181411743,
215
+ -0.2711280584335327,
216
+ -0.18442992866039276,
217
+ -2.373810291290283,
218
+ -0.5469835996627808,
219
+ 0.8536958694458008,
220
+ 0.1966881901025772,
221
+ 0.38944774866104126
222
+ ],
223
+ "std": [
224
+ 0.36294928193092346,
225
+ 0.15225516259670258,
226
+ 0.43781426548957825,
227
+ 0.3171001970767975,
228
+ 0.24347245693206787,
229
+ 0.16601185500621796,
230
+ 0.5455403327941895,
231
+ 0.4952579438686371
232
+ ],
233
+ "q01": [
234
+ -1.1760263681411742,
235
+ -0.7025632262229919,
236
+ -0.5743336498737335,
237
+ -2.7483887910842895,
238
+ -1.0952622890472412,
239
+ 0.3067961633205414,
240
+ -1.1566214561462402,
241
+ 0.0
242
+ ],
243
+ "q99": [
244
+ 0.06589999794960022,
245
+ 0.0,
246
+ 0.954136073589325,
247
+ -1.2292555332183817,
248
+ -0.02761165425181389,
249
+ 1.1704273223876953,
250
+ 0.6280999779701233,
251
+ 1.100000023841858
252
+ ]
253
+ }
254
+ },
255
+ "relative_action": {}
256
+ }
257
+ }
experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": true,
15
+ "backbone_trainable_params_fp32": true,
16
+ "extra_augmentation_config": null,
17
+ "apply_sincos_state_encoding": false,
18
+ "use_percentiles": true,
19
+ "use_relative_action": false,
20
+ "max_state_dim": 132,
21
+ "max_action_dim": 132,
22
+ "action_horizon": 40,
23
+ "hidden_size": 1024,
24
+ "input_embedding_dim": 1536,
25
+ "state_history_length": 1,
26
+ "add_pos_embed": true,
27
+ "attn_dropout": 0.2,
28
+ "use_vlln": true,
29
+ "max_seq_len": 1024,
30
+ "use_alternate_vl_dit": true,
31
+ "attend_text_every_n_blocks": 2,
32
+ "diffusion_model_cfg": {
33
+ "attention_head_dim": 48,
34
+ "dropout": 0.2,
35
+ "final_dropout": true,
36
+ "interleave_self_attention": true,
37
+ "norm_type": "ada_norm",
38
+ "num_attention_heads": 32,
39
+ "num_layers": 32,
40
+ "output_dim": 1024,
41
+ "positional_embeddings": null
42
+ },
43
+ "num_inference_timesteps": 4,
44
+ "noise_beta_alpha": 1.5,
45
+ "noise_beta_beta": 1.0,
46
+ "noise_s": 0.999,
47
+ "num_timestep_buckets": 1000,
48
+ "tune_projector": true,
49
+ "tune_diffusion_model": true,
50
+ "tune_vlln": true,
51
+ "state_dropout_prob": 0.2,
52
+ "exclude_state": false,
53
+ "use_mean_std": false,
54
+ "max_num_embodiments": 32
55
+ }
experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2f86687a7c6219c3e2ee9925f22f211d0ea6eab1d14fd744609902ba801f375
3
+ size 4993709288
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88a5ad55650430f7aef1b4f929dffb08a9c5c60433900c944a414d264db81905
3
+ size 4535492016
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4d7f851869bedf9196cab6c68c055561f12725d9a6e8f1daa1602174a1c0c76
3
+ size 12964594710
processor_config.json ADDED
@@ -0,0 +1,1146 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "cam_left_head"
1019
+ ],
1020
+ "sin_cos_embedding_keys": null,
1021
+ "mean_std_embedding_keys": null,
1022
+ "action_configs": null
1023
+ },
1024
+ "state": {
1025
+ "delta_indices": [
1026
+ 0
1027
+ ],
1028
+ "modality_keys": [
1029
+ "arm_left",
1030
+ "arm_right"
1031
+ ],
1032
+ "sin_cos_embedding_keys": null,
1033
+ "mean_std_embedding_keys": null,
1034
+ "action_configs": null
1035
+ },
1036
+ "action": {
1037
+ "delta_indices": [
1038
+ 0,
1039
+ 1,
1040
+ 2,
1041
+ 3,
1042
+ 4,
1043
+ 5,
1044
+ 6,
1045
+ 7,
1046
+ 8,
1047
+ 9,
1048
+ 10,
1049
+ 11,
1050
+ 12,
1051
+ 13,
1052
+ 14,
1053
+ 15,
1054
+ 16,
1055
+ 17,
1056
+ 18,
1057
+ 19,
1058
+ 20,
1059
+ 21,
1060
+ 22,
1061
+ 23,
1062
+ 24,
1063
+ 25,
1064
+ 26,
1065
+ 27,
1066
+ 28,
1067
+ 29,
1068
+ 30,
1069
+ 31,
1070
+ 32,
1071
+ 33,
1072
+ 34,
1073
+ 35,
1074
+ 36,
1075
+ 37,
1076
+ 38,
1077
+ 39
1078
+ ],
1079
+ "modality_keys": [
1080
+ "arm_left",
1081
+ "arm_right"
1082
+ ],
1083
+ "sin_cos_embedding_keys": null,
1084
+ "mean_std_embedding_keys": null,
1085
+ "action_configs": [
1086
+ {
1087
+ "rep": "ABSOLUTE",
1088
+ "type": "NON_EEF",
1089
+ "format": "DEFAULT",
1090
+ "state_key": null
1091
+ },
1092
+ {
1093
+ "rep": "ABSOLUTE",
1094
+ "type": "NON_EEF",
1095
+ "format": "DEFAULT",
1096
+ "state_key": null
1097
+ }
1098
+ ]
1099
+ },
1100
+ "language": {
1101
+ "delta_indices": [
1102
+ 0
1103
+ ],
1104
+ "modality_keys": [
1105
+ "sub_task"
1106
+ ],
1107
+ "sin_cos_embedding_keys": null,
1108
+ "mean_std_embedding_keys": null,
1109
+ "action_configs": null
1110
+ }
1111
+ }
1112
+ },
1113
+ "image_crop_size": [
1114
+ 230,
1115
+ 230
1116
+ ],
1117
+ "image_target_size": [
1118
+ 256,
1119
+ 256
1120
+ ],
1121
+ "use_albumentations": true,
1122
+ "random_rotation_angle": 0,
1123
+ "color_jitter_params": {
1124
+ "brightness": 0.3,
1125
+ "contrast": 0.4,
1126
+ "saturation": 0.5,
1127
+ "hue": 0.08
1128
+ },
1129
+ "shortest_image_edge": 256,
1130
+ "crop_fraction": 0.95,
1131
+ "letter_box_transform": false,
1132
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1133
+ "model_type": "qwen",
1134
+ "formalize_language": true,
1135
+ "max_state_dim": 132,
1136
+ "max_action_dim": 132,
1137
+ "max_action_horizon": 40,
1138
+ "use_percentiles": true,
1139
+ "use_mean_std": false,
1140
+ "clip_outliers": true,
1141
+ "apply_sincos_state_encoding": false,
1142
+ "use_relative_action": true,
1143
+ "exclude_state": false,
1144
+ "state_dropout_prob": 0.2
1145
+ }
1146
+ }
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6b3dda28bf9c7b09c1c7d7627a67e2acc9015b93fed8ff54390fccae98fd8b2
3
+ size 14645
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95f04fade68ac8bca3e0e3a6410cecfeb6954ecb64402667bc7c4ea9d9c53c6d
3
+ size 1465
statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
trainer_state.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 sha256:628ced46cb1d1a776fa7cadb3f34304bfc63b2a12ebb33b9d6291f8dd3e24b25
3
+ size 5969
wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d7", "run_id": "Groot-n17-Pick-GreenCoffeeCan_WhiteShelf_Groot_2"}