Hugging Face's logo

H2Ozone
/
smoke_400_2

Safetensors
Gr00tN1d7
Model card Files
xet
 Community
H2Ozone commited on
Commit
7d0af8f
·
verified ·
1 Parent(s): 24246aa

Publish

Browse files
Files changed (16) hide show
  1. config.json +88 -0
  2. embodiment_id.json +55 -0
  3. experiment_cfg/conf.yaml +231 -0
  4. experiment_cfg/config.yaml +259 -0
  5. experiment_cfg/dataset_statistics.json +485 -0
  6. experiment_cfg/final_model_config.json +55 -0
  7. experiment_cfg/final_processor_config.json +0 -0
  8. model-00001-of-00003.safetensors +3 -0
  9. model-00002-of-00003.safetensors +3 -0
  10. model-00003-of-00003.safetensors +3 -0
  11. model.safetensors.index.json +0 -0
  12. processor_config.json +1147 -0
  13. statistics.json +0 -0
  14. trainer_state.json +94 -0
  15. training_args.bin +3 -0
  16. 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": "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": 0.05,
63
+ "state_gaussian_noise_std": 0.0,
64
+ "transformers_version": "4.57.3",
65
+ "tune_diffusion_model": false,
66
+ "tune_linear": true,
67
+ "tune_llm": false,
68
+ "tune_projector": false,
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 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ }
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,231 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ 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: false
61
+ tune_diffusion_model: false
62
+ tune_vlln: true
63
+ state_dropout_prob: 0.05
64
+ exclude_state: false
65
+ use_mean_std: false
66
+ max_num_embodiments: 32
67
+ data:
68
+ datasets:
69
+ - dataset_paths:
70
+ - /dataset0
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_zed_left
82
+ - cam_zed_right
83
+ sin_cos_embedding_keys: null
84
+ mean_std_embedding_keys: null
85
+ action_configs: null
86
+ state:
87
+ delta_indices:
88
+ - 0
89
+ modality_keys:
90
+ - left_gripper
91
+ - right_gripper
92
+ - neck
93
+ - left_hand
94
+ - right_hand
95
+ sin_cos_embedding_keys: null
96
+ mean_std_embedding_keys: null
97
+ action_configs: null
98
+ action:
99
+ delta_indices:
100
+ - 0
101
+ - 1
102
+ - 2
103
+ - 3
104
+ - 4
105
+ - 5
106
+ - 6
107
+ - 7
108
+ - 8
109
+ - 9
110
+ - 10
111
+ - 11
112
+ - 12
113
+ - 13
114
+ - 14
115
+ - 15
116
+ modality_keys:
117
+ - left_gripper
118
+ - right_gripper
119
+ - neck
120
+ - left_hand
121
+ - right_hand
122
+ sin_cos_embedding_keys: null
123
+ mean_std_embedding_keys: null
124
+ action_configs:
125
+ - rep: ABSOLUTE
126
+ type: NON_EEF
127
+ format: DEFAULT
128
+ state_key: left_gripper
129
+ - rep: ABSOLUTE
130
+ type: NON_EEF
131
+ format: DEFAULT
132
+ state_key: right_gripper
133
+ - rep: ABSOLUTE
134
+ type: NON_EEF
135
+ format: DEFAULT
136
+ state_key: neck
137
+ - rep: ABSOLUTE
138
+ type: NON_EEF
139
+ format: DEFAULT
140
+ state_key: left_hand
141
+ - rep: ABSOLUTE
142
+ type: NON_EEF
143
+ format: DEFAULT
144
+ state_key: right_hand
145
+ language:
146
+ delta_indices:
147
+ - 0
148
+ modality_keys:
149
+ - annotation.human.action.task_description
150
+ sin_cos_embedding_keys: null
151
+ mean_std_embedding_keys: null
152
+ action_configs: null
153
+ download_cache: false
154
+ shard_size: 1024
155
+ episode_sampling_rate: 0.1
156
+ num_shards_per_epoch: 100000
157
+ override_pretraining_statistics: true
158
+ mode: single_turn
159
+ random_chop: 0.0
160
+ mock_dataset_mode: false
161
+ shuffle: true
162
+ seed: 42
163
+ multiprocessing_context: fork
164
+ allow_padding: false
165
+ subsample_ratio: 1.0
166
+ image_crop_size:
167
+ - 244
168
+ - 244
169
+ image_target_size:
170
+ - 224
171
+ - 224
172
+ video_backend: torchcodec
173
+ training:
174
+ output_dir: /outputs/clean-rdx-smoke100-20260802133441
175
+ experiment_name: clean-rdx-smoke100-20260802133441
176
+ max_steps: 100
177
+ global_batch_size: 8
178
+ batch_size: null
179
+ gradient_accumulation_steps: 1
180
+ learning_rate: 0.0001
181
+ lr_scheduler_type: cosine
182
+ weight_decay: 1.0e-05
183
+ warmup_ratio: 0.05
184
+ warmup_steps: 0
185
+ max_grad_norm: 1.0
186
+ optim: adamw_torch
187
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
188
+ skip_weight_loading: false
189
+ tf32: true
190
+ fp16: false
191
+ bf16: true
192
+ eval_bf16: true
193
+ logging_steps: 10
194
+ save_steps: 100
195
+ save_total_limit: 1
196
+ save_vl_model: false
197
+ save_only_model: true
198
+ upload_checkpoints: false
199
+ upload_every: 1000
200
+ upload_last_n_checkpoints: 5
201
+ max_concurrent_uploads: 2
202
+ eval_strategy: 'no'
203
+ eval_steps: 500
204
+ eval_set_split_ratio: 0.1
205
+ eval_batch_size: 2
206
+ save_best_eval_metric_name: ''
207
+ save_best_eval_metric_greater_is_better: true
208
+ deepspeed_stage: 2
209
+ gradient_checkpointing: false
210
+ transformers_trust_remote_code: true
211
+ transformers_local_files_only: false
212
+ transformers_cache_dir: null
213
+ transformers_access_token: null
214
+ use_ddp: false
215
+ ddp_bucket_cap_mb: 100
216
+ num_gpus: 1
217
+ dataloader_num_workers: 2
218
+ remove_unused_columns: false
219
+ use_wandb: false
220
+ wandb_project: finetune-gr00t-n1d7
221
+ enable_profiling: false
222
+ max_retries: 3
223
+ assert_loss_less_than: null
224
+ add_rl_callback: false
225
+ enable_open_loop_eval: false
226
+ open_loop_eval_traj_ids:
227
+ - 0
228
+ open_loop_eval_steps_per_traj: 100
229
+ open_loop_eval_plot_indices: null
230
+ max_steps: 100
231
+ save_steps: 100
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,259 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ - /dataset0
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: left_gripper
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: right_gripper
37
+ type: *id003
38
+ - !!python/object:gr00t.data.types.ActionConfig
39
+ format: *id001
40
+ rep: *id002
41
+ state_key: neck
42
+ type: *id003
43
+ - !!python/object:gr00t.data.types.ActionConfig
44
+ format: *id001
45
+ rep: *id002
46
+ state_key: left_hand
47
+ type: *id003
48
+ - !!python/object:gr00t.data.types.ActionConfig
49
+ format: *id001
50
+ rep: *id002
51
+ state_key: right_hand
52
+ type: *id003
53
+ delta_indices:
54
+ - 0
55
+ - 1
56
+ - 2
57
+ - 3
58
+ - 4
59
+ - 5
60
+ - 6
61
+ - 7
62
+ - 8
63
+ - 9
64
+ - 10
65
+ - 11
66
+ - 12
67
+ - 13
68
+ - 14
69
+ - 15
70
+ mean_std_embedding_keys: null
71
+ modality_keys:
72
+ - left_gripper
73
+ - right_gripper
74
+ - neck
75
+ - left_hand
76
+ - right_hand
77
+ sin_cos_embedding_keys: null
78
+ language: !!python/object:gr00t.data.types.ModalityConfig
79
+ action_configs: null
80
+ delta_indices:
81
+ - 0
82
+ mean_std_embedding_keys: null
83
+ modality_keys:
84
+ - annotation.human.action.task_description
85
+ sin_cos_embedding_keys: null
86
+ state: !!python/object:gr00t.data.types.ModalityConfig
87
+ action_configs: null
88
+ delta_indices:
89
+ - 0
90
+ mean_std_embedding_keys: null
91
+ modality_keys:
92
+ - left_gripper
93
+ - right_gripper
94
+ - neck
95
+ - left_hand
96
+ - right_hand
97
+ sin_cos_embedding_keys: null
98
+ video: !!python/object:gr00t.data.types.ModalityConfig
99
+ action_configs: null
100
+ delta_indices:
101
+ - 0
102
+ mean_std_embedding_keys: null
103
+ modality_keys:
104
+ - cam_zed_left
105
+ - cam_zed_right
106
+ sin_cos_embedding_keys: null
107
+ mode: single_turn
108
+ multiprocessing_context: fork
109
+ num_shards_per_epoch: 100000
110
+ override_pretraining_statistics: true
111
+ random_chop: 0.0
112
+ seed: 42
113
+ shard_size: 1024
114
+ shuffle: true
115
+ subsample_ratio: 1.0
116
+ video_backend: torchcodec
117
+ load_config_path: null
118
+ model: !!python/object:gr00t.configs.model.gr00t_n1d7.Gr00tN1d7Config
119
+ _attn_implementation_internal: null
120
+ _commit_hash: null
121
+ _name_or_path: ''
122
+ _output_attentions: false
123
+ add_cross_attention: false
124
+ architectures: null
125
+ backbone_trainable_params_fp32: true
126
+ bad_words_ids: null
127
+ begin_suppress_tokens: null
128
+ bos_token_id: null
129
+ chunk_size_feed_forward: 0
130
+ color_jitter_params: null
131
+ cross_attention_hidden_size: null
132
+ decoder_start_token_id: null
133
+ diffusion_model_cfg:
134
+ attention_head_dim: 48
135
+ dropout: 0.2
136
+ final_dropout: true
137
+ interleave_self_attention: true
138
+ norm_type: ada_norm
139
+ num_attention_heads: 32
140
+ num_layers: 16
141
+ output_dim: 1024
142
+ positional_embeddings: null
143
+ diversity_penalty: 0.0
144
+ do_sample: false
145
+ dtype: null
146
+ early_stopping: false
147
+ encoder_no_repeat_ngram_size: 0
148
+ eos_token_id: null
149
+ exponential_decay_length_penalty: null
150
+ extra_augmentation_config: null
151
+ finetuning_task: null
152
+ forced_bos_token_id: null
153
+ forced_eos_token_id: null
154
+ id2label:
155
+ 0: LABEL_0
156
+ 1: LABEL_1
157
+ is_decoder: false
158
+ is_encoder_decoder: false
159
+ label2id:
160
+ LABEL_0: 0
161
+ LABEL_1: 1
162
+ length_penalty: 1.0
163
+ load_bf16: false
164
+ max_length: 20
165
+ min_length: 0
166
+ model_name: nvidia/Cosmos-Reason2-2B
167
+ no_repeat_ngram_size: 0
168
+ num_beam_groups: 1
169
+ num_beams: 1
170
+ num_return_sequences: 1
171
+ output_hidden_states: false
172
+ output_scores: false
173
+ pad_token_id: null
174
+ prefix: null
175
+ problem_type: null
176
+ pruned_heads: {}
177
+ random_rotation_angle: null
178
+ remove_invalid_values: false
179
+ repetition_penalty: 1.0
180
+ reproject_vision: false
181
+ return_dict: true
182
+ return_dict_in_generate: false
183
+ sep_token_id: null
184
+ state_dropout_prob: 0.05
185
+ suppress_tokens: null
186
+ task_specific_params: null
187
+ temperature: 1.0
188
+ tf_legacy_loss: false
189
+ tie_encoder_decoder: false
190
+ tie_word_embeddings: true
191
+ tokenizer_class: null
192
+ top_k: 50
193
+ top_p: 1.0
194
+ torchscript: false
195
+ transformers_version: null
196
+ tune_diffusion_model: false
197
+ tune_llm: false
198
+ tune_projector: false
199
+ tune_visual: false
200
+ typical_p: 1.0
201
+ use_bfloat16: false
202
+ use_relative_action: true
203
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
204
+ add_rl_callback: false
205
+ assert_loss_less_than: null
206
+ batch_size: null
207
+ bf16: true
208
+ dataloader_num_workers: 2
209
+ ddp_bucket_cap_mb: 100
210
+ deepspeed_stage: 2
211
+ enable_open_loop_eval: false
212
+ enable_profiling: false
213
+ eval_batch_size: 2
214
+ eval_bf16: true
215
+ eval_set_split_ratio: 0.1
216
+ eval_steps: 500
217
+ eval_strategy: 'no'
218
+ experiment_name: clean-rdx-smoke100-20260802133441
219
+ fp16: false
220
+ global_batch_size: 8
221
+ gradient_accumulation_steps: 1
222
+ gradient_checkpointing: false
223
+ learning_rate: 0.0001
224
+ logging_steps: 10
225
+ lr_scheduler_type: cosine
226
+ max_concurrent_uploads: 2
227
+ max_grad_norm: 1.0
228
+ max_retries: 3
229
+ max_steps: 100
230
+ num_gpus: 1
231
+ open_loop_eval_plot_indices: null
232
+ open_loop_eval_steps_per_traj: 100
233
+ open_loop_eval_traj_ids:
234
+ - 0
235
+ optim: adamw_torch
236
+ output_dir: /outputs/clean-rdx-smoke100-20260802133441
237
+ remove_unused_columns: false
238
+ save_best_eval_metric_greater_is_better: true
239
+ save_best_eval_metric_name: ''
240
+ save_only_model: true
241
+ save_steps: 100
242
+ save_total_limit: 1
243
+ save_vl_model: false
244
+ skip_weight_loading: false
245
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
246
+ tf32: true
247
+ transformers_access_token: null
248
+ transformers_cache_dir: null
249
+ transformers_local_files_only: false
250
+ transformers_trust_remote_code: true
251
+ upload_checkpoints: false
252
+ upload_every: 1000
253
+ upload_last_n_checkpoints: 5
254
+ use_ddp: false
255
+ use_wandb: false
256
+ wandb_project: finetune-gr00t-n1d7
257
+ warmup_ratio: 0.05
258
+ warmup_steps: 0
259
+ weight_decay: 1.0e-05
experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,485 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "left_gripper": {
5
+ "min": [
6
+ 0.06650379300117493,
7
+ 0.30686625838279724,
8
+ 1.015108585357666,
9
+ 0.053404130041599274,
10
+ -0.5671195983886719,
11
+ -0.11310392618179321,
12
+ 0.8140828609466553
13
+ ],
14
+ "max": [
15
+ 0.10620637238025665,
16
+ 0.34438881278038025,
17
+ 1.0170872211456299,
18
+ 0.053407713770866394,
19
+ -0.5671164989471436,
20
+ -0.11309898644685745,
21
+ 0.8140853047370911
22
+ ],
23
+ "mean": [
24
+ 0.07703523337841034,
25
+ 0.3292888104915619,
26
+ 1.0158779621124268,
27
+ 0.05342363566160202,
28
+ -0.5668591856956482,
29
+ -0.11313915997743607,
30
+ 0.8139637112617493
31
+ ],
32
+ "std": [
33
+ 0.009054359048604965,
34
+ 0.007089548278600452,
35
+ 0.000431410648099603,
36
+ 1.7881719028756088e-05,
37
+ 0.0002590171934817411,
38
+ 3.726810972896151e-05,
39
+ 0.00012026249147287637
40
+ ],
41
+ "q01": [
42
+ 0.06812608391046523,
43
+ 0.31214413344860076,
44
+ 1.0156248688697815,
45
+ 0.053405098617076874,
46
+ -0.5671187043190002,
47
+ -0.1131029725074768,
48
+ 0.8140836358070374
49
+ ],
50
+ "q99": [
51
+ 0.10512059070169925,
52
+ 0.3421141102910042,
53
+ 1.016560959815979,
54
+ 0.05340619385242462,
55
+ -0.5671178698539734,
56
+ -0.11310117691755295,
57
+ 0.8140842318534851
58
+ ]
59
+ },
60
+ "right_gripper": {
61
+ "min": [
62
+ 0.07252335548400879,
63
+ -0.6595533490180969,
64
+ 0.9555243849754333,
65
+ -0.060599904507398605,
66
+ -0.8033609390258789,
67
+ -0.038712505251169205,
68
+ 0.5944293141365051
69
+ ],
70
+ "max": [
71
+ 0.4935738146305084,
72
+ -0.16382907330989838,
73
+ 1.4599127769470215,
74
+ 0.15038828551769257,
75
+ -0.5671184062957764,
76
+ 0.1621238738298416,
77
+ 0.8144175410270691
78
+ ],
79
+ "mean": [
80
+ 0.3964168429374695,
81
+ -0.3828529417514801,
82
+ 1.2114816904067993,
83
+ -0.0025471041444689035,
84
+ -0.6917619109153748,
85
+ 0.017066726461052895,
86
+ 0.718989908695221
87
+ ],
88
+ "std": [
89
+ 0.1058579757809639,
90
+ 0.10126025229692459,
91
+ 0.10796082019805908,
92
+ 0.020969947800040245,
93
+ 0.038796257227659045,
94
+ 0.034196916967630386,
95
+ 0.032114662230013934
96
+ ],
97
+ "q01": [
98
+ 0.07790312059223652,
99
+ -0.6007718592882156,
100
+ 0.9929429292678833,
101
+ -0.05521200839430094,
102
+ -0.7626979351043701,
103
+ -0.019676192943006755,
104
+ 0.6431728601455688
105
+ ],
106
+ "q99": [
107
+ 0.4896747425198555,
108
+ -0.18751139044761647,
109
+ 1.4084784388542175,
110
+ 0.07251456975936892,
111
+ -0.5694491803646087,
112
+ 0.11543079726397992,
113
+ 0.8126143902540207
114
+ ]
115
+ },
116
+ "neck": {
117
+ "min": [
118
+ 1.0698148145138475e-07,
119
+ -0.9108108282089233
120
+ ],
121
+ "max": [
122
+ 0.4886775612831116,
123
+ -7.201997220818157e-09
124
+ ],
125
+ "mean": [
126
+ 0.3409970998764038,
127
+ -0.6277368664741516
128
+ ],
129
+ "std": [
130
+ 0.09556373953819275,
131
+ 0.19625218212604523
132
+ ],
133
+ "q01": [
134
+ 0.00493338811211288,
135
+ -0.8788565397262573
136
+ ],
137
+ "q99": [
138
+ 0.48858001828193665,
139
+ -0.008497750619426219
140
+ ]
141
+ },
142
+ "left_hand": {
143
+ "min": [
144
+ 0.0,
145
+ 0.0,
146
+ 0.0,
147
+ 0.0,
148
+ 0.0,
149
+ 0.0
150
+ ],
151
+ "max": [
152
+ 0.0,
153
+ 0.0,
154
+ 0.0,
155
+ 0.0,
156
+ 0.0,
157
+ 0.0
158
+ ],
159
+ "mean": [
160
+ 0.0,
161
+ 0.0,
162
+ 0.0,
163
+ 0.0,
164
+ 0.0,
165
+ 0.0
166
+ ],
167
+ "std": [
168
+ 0.0,
169
+ 0.0,
170
+ 0.0,
171
+ 0.0,
172
+ 0.0,
173
+ 0.0
174
+ ],
175
+ "q01": [
176
+ 0.0,
177
+ 0.0,
178
+ 0.0,
179
+ 0.0,
180
+ 0.0,
181
+ 0.0
182
+ ],
183
+ "q99": [
184
+ 0.0,
185
+ 0.0,
186
+ 0.0,
187
+ 0.0,
188
+ 0.0,
189
+ 0.0
190
+ ]
191
+ },
192
+ "right_hand": {
193
+ "min": [
194
+ 1.751050353050232,
195
+ 1.751050353050232,
196
+ 1.751050353050232,
197
+ 1.751050353050232,
198
+ 0.0,
199
+ 0.5093724131584167
200
+ ],
201
+ "max": [
202
+ 1.751050353050232,
203
+ 1.751050353050232,
204
+ 1.751050353050232,
205
+ 1.751050353050232,
206
+ 0.0,
207
+ 0.5093724131584167
208
+ ],
209
+ "mean": [
210
+ 1.7505443096160889,
211
+ 1.7505443096160889,
212
+ 1.7505443096160889,
213
+ 1.7505443096160889,
214
+ 0.0,
215
+ 0.5096193552017212
216
+ ],
217
+ "std": [
218
+ 0.0005060734694468678,
219
+ 0.0005060734694468678,
220
+ 0.0005060734694468678,
221
+ 0.0005060734694468678,
222
+ 0.0,
223
+ 0.0002469774918248063
224
+ ],
225
+ "q01": [
226
+ 1.751050353050232,
227
+ 1.751050353050232,
228
+ 1.751050353050232,
229
+ 1.751050353050232,
230
+ 0.0,
231
+ 0.5093724131584167
232
+ ],
233
+ "q99": [
234
+ 1.751050353050232,
235
+ 1.751050353050232,
236
+ 1.751050353050232,
237
+ 1.751050353050232,
238
+ 0.0,
239
+ 0.5093724131584167
240
+ ]
241
+ }
242
+ },
243
+ "action": {
244
+ "left_gripper": {
245
+ "min": [
246
+ 0.0015351285692304373,
247
+ 0.26026827096939087,
248
+ -0.16075946390628815,
249
+ 0.0,
250
+ 0.0,
251
+ 0.0,
252
+ 1.0
253
+ ],
254
+ "max": [
255
+ 0.0015351285692304373,
256
+ 0.26026827096939087,
257
+ -0.16075946390628815,
258
+ 0.0,
259
+ 0.0,
260
+ 0.0,
261
+ 1.0
262
+ ],
263
+ "mean": [
264
+ 0.0015342573169618845,
265
+ 0.2604384124279022,
266
+ -0.16067518293857574,
267
+ 0.0,
268
+ 0.0,
269
+ 0.0,
270
+ 1.0
271
+ ],
272
+ "std": [
273
+ 8.713068382272917e-07,
274
+ 0.00017016551283718478,
275
+ 8.427217835011648e-05,
276
+ 0.0,
277
+ 0.0,
278
+ 0.0,
279
+ 0.0
280
+ ],
281
+ "q01": [
282
+ 0.0015351285692304373,
283
+ 0.26026827096939087,
284
+ -0.16075946390628815,
285
+ 0.0,
286
+ 0.0,
287
+ 0.0,
288
+ 1.0
289
+ ],
290
+ "q99": [
291
+ 0.0015351285692304373,
292
+ 0.26026827096939087,
293
+ -0.16075946390628815,
294
+ 0.0,
295
+ 0.0,
296
+ 0.0,
297
+ 1.0
298
+ ]
299
+ },
300
+ "right_gripper": {
301
+ "min": [
302
+ 0.43602389097213745,
303
+ -0.5957071185112,
304
+ 0.9515008926391602,
305
+ 0.0,
306
+ -0.7071067690849304,
307
+ 0.0,
308
+ 0.7071067690849304
309
+ ],
310
+ "max": [
311
+ 0.5346108078956604,
312
+ -0.16541779041290283,
313
+ 1.420967936515808,
314
+ 0.0,
315
+ -0.7071067690849304,
316
+ 0.0,
317
+ 0.7071067690849304
318
+ ],
319
+ "mean": [
320
+ 0.49645933508872986,
321
+ -0.3569822609424591,
322
+ 1.2123045921325684,
323
+ 0.0,
324
+ -0.7070366740226746,
325
+ 0.0,
326
+ 0.7070366740226746
327
+ ],
328
+ "std": [
329
+ 0.025103358551859786,
330
+ 0.09924318641424179,
331
+ 0.11866503208875633,
332
+ 0.0,
333
+ 7.011165858179447e-05,
334
+ 0.0,
335
+ 7.011165858179447e-05
336
+ ],
337
+ "q01": [
338
+ 0.440896213054657,
339
+ -0.5652589797973633,
340
+ 0.9792395234107971,
341
+ 0.0,
342
+ -0.7071067690849304,
343
+ 0.0,
344
+ 0.7071067690849304
345
+ ],
346
+ "q99": [
347
+ 0.5333313345909119,
348
+ -0.17929720878601074,
349
+ 1.4075326919555664,
350
+ 0.0,
351
+ -0.7071067690849304,
352
+ 0.0,
353
+ 0.7071067690849304
354
+ ]
355
+ },
356
+ "neck": {
357
+ "min": [
358
+ 0.05709997937083244,
359
+ -0.9229851961135864
360
+ ],
361
+ "max": [
362
+ 0.4886919856071472,
363
+ -0.2774601876735687
364
+ ],
365
+ "mean": [
366
+ 0.3864362835884094,
367
+ -0.6038118600845337
368
+ ],
369
+ "std": [
370
+ 0.11821556091308594,
371
+ 0.13974444568157196
372
+ ],
373
+ "q01": [
374
+ 0.10200093947350979,
375
+ -0.862481090426445
376
+ ],
377
+ "q99": [
378
+ 0.4886919856071472,
379
+ -0.32631652057170846
380
+ ]
381
+ },
382
+ "left_hand": {
383
+ "min": [
384
+ 0.0,
385
+ 0.0,
386
+ 0.0,
387
+ 0.0,
388
+ 0.0,
389
+ 0.0
390
+ ],
391
+ "max": [
392
+ 0.0,
393
+ 0.0,
394
+ 0.0,
395
+ 0.0,
396
+ 0.0,
397
+ 0.0
398
+ ],
399
+ "mean": [
400
+ 0.0,
401
+ 0.0,
402
+ 0.0,
403
+ 0.0,
404
+ 0.0,
405
+ 0.0
406
+ ],
407
+ "std": [
408
+ 0.0,
409
+ 0.0,
410
+ 0.0,
411
+ 0.0,
412
+ 0.0,
413
+ 0.0
414
+ ],
415
+ "q01": [
416
+ 0.0,
417
+ 0.0,
418
+ 0.0,
419
+ 0.0,
420
+ 0.0,
421
+ 0.0
422
+ ],
423
+ "q99": [
424
+ 0.0,
425
+ 0.0,
426
+ 0.0,
427
+ 0.0,
428
+ 0.0,
429
+ 0.0
430
+ ]
431
+ },
432
+ "right_hand": {
433
+ "min": [
434
+ 1.751050353050232,
435
+ 1.751050353050232,
436
+ 1.751050353050232,
437
+ 1.751050353050232,
438
+ 0.0,
439
+ 0.5093724131584167
440
+ ],
441
+ "max": [
442
+ 1.751050353050232,
443
+ 1.751050353050232,
444
+ 1.751050353050232,
445
+ 1.751050353050232,
446
+ 0.0,
447
+ 0.5093724131584167
448
+ ],
449
+ "mean": [
450
+ 1.7505443096160889,
451
+ 1.7505443096160889,
452
+ 1.7505443096160889,
453
+ 1.7505443096160889,
454
+ 0.0,
455
+ 0.5096193552017212
456
+ ],
457
+ "std": [
458
+ 0.0005060734694468678,
459
+ 0.0005060734694468678,
460
+ 0.0005060734694468678,
461
+ 0.0005060734694468678,
462
+ 0.0,
463
+ 0.0002469774918248063
464
+ ],
465
+ "q01": [
466
+ 1.751050353050232,
467
+ 1.751050353050232,
468
+ 1.751050353050232,
469
+ 1.751050353050232,
470
+ 0.0,
471
+ 0.5093724131584167
472
+ ],
473
+ "q99": [
474
+ 1.751050353050232,
475
+ 1.751050353050232,
476
+ 1.751050353050232,
477
+ 1.751050353050232,
478
+ 0.0,
479
+ 0.5093724131584167
480
+ ]
481
+ }
482
+ },
483
+ "relative_action": {}
484
+ }
485
+ }
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": false,
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": false,
49
+ "tune_diffusion_model": false,
50
+ "tune_vlln": true,
51
+ "state_dropout_prob": 0.05,
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-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:c39153e2c89aa5c7b6423b2fb05725d21fa1b594fe61dbebaffa7956736848e3
3
+ size 4970792616
model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b6963d0a7b05f9dcdfc9ce2b9decbca8947edb2b9427b2dd711105425553c74
3
+ size 2618758696
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
processor_config.json ADDED
@@ -0,0 +1,1147 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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_zed_left",
1019
+ "cam_zed_right"
1020
+ ],
1021
+ "sin_cos_embedding_keys": null,
1022
+ "mean_std_embedding_keys": null,
1023
+ "action_configs": null
1024
+ },
1025
+ "state": {
1026
+ "delta_indices": [
1027
+ 0
1028
+ ],
1029
+ "modality_keys": [
1030
+ "left_gripper",
1031
+ "right_gripper",
1032
+ "neck",
1033
+ "left_hand",
1034
+ "right_hand"
1035
+ ],
1036
+ "sin_cos_embedding_keys": null,
1037
+ "mean_std_embedding_keys": null,
1038
+ "action_configs": null
1039
+ },
1040
+ "action": {
1041
+ "delta_indices": [
1042
+ 0,
1043
+ 1,
1044
+ 2,
1045
+ 3,
1046
+ 4,
1047
+ 5,
1048
+ 6,
1049
+ 7,
1050
+ 8,
1051
+ 9,
1052
+ 10,
1053
+ 11,
1054
+ 12,
1055
+ 13,
1056
+ 14,
1057
+ 15
1058
+ ],
1059
+ "modality_keys": [
1060
+ "left_gripper",
1061
+ "right_gripper",
1062
+ "neck",
1063
+ "left_hand",
1064
+ "right_hand"
1065
+ ],
1066
+ "sin_cos_embedding_keys": null,
1067
+ "mean_std_embedding_keys": null,
1068
+ "action_configs": [
1069
+ {
1070
+ "rep": "ABSOLUTE",
1071
+ "type": "NON_EEF",
1072
+ "format": "DEFAULT",
1073
+ "state_key": "left_gripper"
1074
+ },
1075
+ {
1076
+ "rep": "ABSOLUTE",
1077
+ "type": "NON_EEF",
1078
+ "format": "DEFAULT",
1079
+ "state_key": "right_gripper"
1080
+ },
1081
+ {
1082
+ "rep": "ABSOLUTE",
1083
+ "type": "NON_EEF",
1084
+ "format": "DEFAULT",
1085
+ "state_key": "neck"
1086
+ },
1087
+ {
1088
+ "rep": "ABSOLUTE",
1089
+ "type": "NON_EEF",
1090
+ "format": "DEFAULT",
1091
+ "state_key": "left_hand"
1092
+ },
1093
+ {
1094
+ "rep": "ABSOLUTE",
1095
+ "type": "NON_EEF",
1096
+ "format": "DEFAULT",
1097
+ "state_key": "right_hand"
1098
+ }
1099
+ ]
1100
+ },
1101
+ "language": {
1102
+ "delta_indices": [
1103
+ 0
1104
+ ],
1105
+ "modality_keys": [
1106
+ "annotation.human.action.task_description"
1107
+ ],
1108
+ "sin_cos_embedding_keys": null,
1109
+ "mean_std_embedding_keys": null,
1110
+ "action_configs": null
1111
+ }
1112
+ }
1113
+ },
1114
+ "image_crop_size": [
1115
+ 230,
1116
+ 230
1117
+ ],
1118
+ "image_target_size": [
1119
+ 256,
1120
+ 256
1121
+ ],
1122
+ "use_albumentations": true,
1123
+ "random_rotation_angle": 0,
1124
+ "color_jitter_params": {
1125
+ "brightness": 0.3,
1126
+ "contrast": 0.4,
1127
+ "saturation": 0.5,
1128
+ "hue": 0.08
1129
+ },
1130
+ "shortest_image_edge": 256,
1131
+ "crop_fraction": 0.95,
1132
+ "letter_box_transform": false,
1133
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1134
+ "model_type": "qwen",
1135
+ "formalize_language": true,
1136
+ "max_state_dim": 132,
1137
+ "max_action_dim": 132,
1138
+ "max_action_horizon": 40,
1139
+ "use_percentiles": true,
1140
+ "use_mean_std": false,
1141
+ "clip_outliers": true,
1142
+ "apply_sincos_state_encoding": false,
1143
+ "use_relative_action": true,
1144
+ "exclude_state": false,
1145
+ "state_dropout_prob": 0.05
1146
+ }
1147
+ }
statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
trainer_state.json ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.0,
6
+ "eval_steps": 500,
7
+ "global_step": 100,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "grad_norm": 0.06216534227132797,
14
+ "learning_rate": 9.956320346634876e-05,
15
+ "loss": 1.0637,
16
+ "step": 10
17
+ },
18
+ {
19
+ "grad_norm": 0.024572907015681267,
20
+ "learning_rate": 9.473646649103818e-05,
21
+ "loss": 1.0544,
22
+ "step": 20
23
+ },
24
+ {
25
+ "grad_norm": 0.08026651293039322,
26
+ "learning_rate": 8.506183921362443e-05,
27
+ "loss": 1.062,
28
+ "step": 30
29
+ },
30
+ {
31
+ "grad_norm": 0.0559636652469635,
32
+ "learning_rate": 7.158771761692464e-05,
33
+ "loss": 1.0507,
34
+ "step": 40
35
+ },
36
+ {
37
+ "grad_norm": 0.028828831389546394,
38
+ "learning_rate": 5.577423184847932e-05,
39
+ "loss": 1.0524,
40
+ "step": 50
41
+ },
42
+ {
43
+ "grad_norm": 0.028848476707935333,
44
+ "learning_rate": 3.933501846281267e-05,
45
+ "loss": 1.0528,
46
+ "step": 60
47
+ },
48
+ {
49
+ "grad_norm": 0.05331048369407654,
50
+ "learning_rate": 2.405152131093926e-05,
51
+ "loss": 1.0421,
52
+ "step": 70
53
+ },
54
+ {
55
+ "grad_norm": 0.06458555161952972,
56
+ "learning_rate": 1.157994445715706e-05,
57
+ "loss": 1.0586,
58
+ "step": 80
59
+ },
60
+ {
61
+ "grad_norm": 0.05882468819618225,
62
+ "learning_rate": 3.271776770026963e-06,
63
+ "loss": 1.0451,
64
+ "step": 90
65
+ },
66
+ {
67
+ "grad_norm": 0.04416602477431297,
68
+ "learning_rate": 2.7337132953697554e-08,
69
+ "loss": 1.0563,
70
+ "step": 100
71
+ }
72
+ ],
73
+ "logging_steps": 10,
74
+ "max_steps": 100,
75
+ "num_input_tokens_seen": 0,
76
+ "num_train_epochs": 9223372036854775807,
77
+ "save_steps": 100,
78
+ "stateful_callbacks": {
79
+ "TrainerControl": {
80
+ "args": {
81
+ "should_epoch_stop": false,
82
+ "should_evaluate": false,
83
+ "should_log": false,
84
+ "should_save": true,
85
+ "should_training_stop": true
86
+ },
87
+ "attributes": {}
88
+ }
89
+ },
90
+ "total_flos": 0.0,
91
+ "train_batch_size": 8,
92
+ "trial_name": null,
93
+ "trial_params": null
94
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:074de6326986dd7fa5e9149f23d9832ad2fb04ba7ce759683a9fb3e80dbbc4e6
3
+ size 5905
wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d7", "run_id": "clean-rdx-smoke100-20260802133441"}