Hugging Face's logo

EbotsVLA
/
cable_harness_real_plus_syn_cart

Safetensors
Model card Files
xet
 Community
gayatrid-ebots commited on
Commit
b0ddc55
·
verified ·
1 Parent(s): 77307e5

Uploaded a new model: gr00t/n1p7_real_data_cart/checkpoint-30000

Browse files
Files changed (19) hide show
  1. gr00t/n1p7_real_data_cart/checkpoint-30000/config.json +88 -0
  2. gr00t/n1p7_real_data_cart/checkpoint-30000/embodiment_id.json +56 -0
  3. gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/conf.yaml +218 -0
  4. gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/config.yaml +244 -0
  5. gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/dataset_statistics.json +932 -0
  6. gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/final_model_config.json +55 -0
  7. gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/final_processor_config.json +0 -0
  8. gr00t/n1p7_real_data_cart/checkpoint-30000/model-00001-of-00003.safetensors +3 -0
  9. gr00t/n1p7_real_data_cart/checkpoint-30000/model-00002-of-00003.safetensors +3 -0
  10. gr00t/n1p7_real_data_cart/checkpoint-30000/model-00003-of-00003.safetensors +3 -0
  11. gr00t/n1p7_real_data_cart/checkpoint-30000/model.safetensors.index.json +0 -0
  12. gr00t/n1p7_real_data_cart/checkpoint-30000/optimizer.pt +3 -0
  13. gr00t/n1p7_real_data_cart/checkpoint-30000/processor_config.json +1124 -0
  14. gr00t/n1p7_real_data_cart/checkpoint-30000/rng_state.pth +3 -0
  15. gr00t/n1p7_real_data_cart/checkpoint-30000/scheduler.pt +3 -0
  16. gr00t/n1p7_real_data_cart/checkpoint-30000/statistics.json +0 -0
  17. gr00t/n1p7_real_data_cart/checkpoint-30000/trainer_state.json +0 -0
  18. gr00t/n1p7_real_data_cart/checkpoint-30000/training_args.bin +3 -0
  19. gr00t/n1p7_real_data_cart/checkpoint-30000/wandb_config.json +1 -0
gr00t/n1p7_real_data_cart/checkpoint-30000/config.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_horizon": 40,
3
+ "add_pos_embed": true,
4
+ "apply_sincos_state_encoding": false,
5
+ "architectures": [
6
+ "Gr00tN1d7"
7
+ ],
8
+ "attn_dropout": 0.2,
9
+ "attn_implementation": null,
10
+ "backbone_embedding_dim": 2048,
11
+ "backbone_trainable_params_fp32": true,
12
+ "color_jitter_params": {
13
+ "brightness": 0.3,
14
+ "contrast": 0.4,
15
+ "hue": 0.08,
16
+ "saturation": 0.5
17
+ },
18
+ "crop_fraction": 0.95,
19
+ "diffusion_model_cfg": {
20
+ "attention_head_dim": 48,
21
+ "dropout": 0.2,
22
+ "final_dropout": true,
23
+ "interleave_self_attention": true,
24
+ "norm_type": "ada_norm",
25
+ "num_attention_heads": 32,
26
+ "num_layers": 32,
27
+ "output_dim": 1024,
28
+ "positional_embeddings": null
29
+ },
30
+ "dtype": "float32",
31
+ "exclude_state": false,
32
+ "formalize_language": true,
33
+ "hidden_size": 1024,
34
+ "image_crop_size": [
35
+ 230,
36
+ 230
37
+ ],
38
+ "image_target_size": [
39
+ 256,
40
+ 256
41
+ ],
42
+ "letter_box_transform": false,
43
+ "load_bf16": false,
44
+ "max_action_dim": 132,
45
+ "max_num_embodiments": 32,
46
+ "max_seq_len": 1024,
47
+ "max_state_dim": 132,
48
+ "model_dtype": "bfloat16",
49
+ "model_name": "nvidia/Cosmos-Reason2-2B",
50
+ "model_type": "Gr00tN1d7",
51
+ "noise_beta_alpha": 1.5,
52
+ "noise_beta_beta": 1.0,
53
+ "noise_s": 0.999,
54
+ "num_inference_timesteps": 4,
55
+ "num_timestep_buckets": 1000,
56
+ "random_history_crop": true,
57
+ "random_rotation_angle": 0,
58
+ "reproject_vision": false,
59
+ "rtc_ramp_rate": 6.0,
60
+ "select_layer": 16,
61
+ "shortest_image_edge": 256,
62
+ "state_dropout_prob": 0.2,
63
+ "state_gaussian_noise_std": 0.0,
64
+ "transformers_version": "4.57.3",
65
+ "tune_diffusion_model": true,
66
+ "tune_linear": true,
67
+ "tune_llm": false,
68
+ "tune_projector": true,
69
+ "tune_top_llm_layers": 0,
70
+ "tune_visual": false,
71
+ "tune_vlln": true,
72
+ "use_albumentations": true,
73
+ "use_alternate_vl_dit": true,
74
+ "use_flash_attention": true,
75
+ "use_future_tokens": false,
76
+ "use_mean_std": false,
77
+ "use_percentiles": true,
78
+ "use_vl_self_attention": true,
79
+ "use_vlln": true,
80
+ "vl_self_attention_cfg": {
81
+ "attention_head_dim": 64,
82
+ "dropout": 0.2,
83
+ "final_dropout": true,
84
+ "num_attention_heads": 32,
85
+ "num_layers": 4,
86
+ "positional_embeddings": null
87
+ }
88
+ }
gr00t/n1p7_real_data_cart/checkpoint-30000/embodiment_id.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ }
gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,218 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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:
27
+ brightness: 0.3
28
+ contrast: 0.4
29
+ saturation: 0.5
30
+ hue: 0.08
31
+ use_albumentations_transforms: true
32
+ extra_augmentation_config: null
33
+ formalize_language: true
34
+ apply_sincos_state_encoding: false
35
+ use_percentiles: true
36
+ use_relative_action: true
37
+ max_state_dim: 132
38
+ max_action_dim: 132
39
+ action_horizon: 40
40
+ hidden_size: 1024
41
+ input_embedding_dim: 1536
42
+ state_history_length: 1
43
+ add_pos_embed: true
44
+ attn_dropout: 0.2
45
+ use_vlln: true
46
+ max_seq_len: 1024
47
+ use_alternate_vl_dit: true
48
+ attend_text_every_n_blocks: 2
49
+ diffusion_model_cfg:
50
+ positional_embeddings: null
51
+ num_layers: 16
52
+ num_attention_heads: 32
53
+ attention_head_dim: 48
54
+ norm_type: ada_norm
55
+ dropout: 0.2
56
+ final_dropout: true
57
+ output_dim: 1024
58
+ interleave_self_attention: true
59
+ num_inference_timesteps: 4
60
+ noise_beta_alpha: 1.5
61
+ noise_beta_beta: 1.0
62
+ noise_s: 0.999
63
+ num_timestep_buckets: 1000
64
+ tune_projector: true
65
+ tune_diffusion_model: true
66
+ tune_vlln: true
67
+ state_dropout_prob: 0.2
68
+ exclude_state: false
69
+ use_mean_std: false
70
+ max_num_embodiments: 32
71
+ data:
72
+ datasets:
73
+ - dataset_paths:
74
+ - /home/gayatri-davuluri/datasets/cableHarness_v2v3_v3rec13_upd_prompts_Config
75
+ embodiment_tag: new_embodiment
76
+ mix_ratio: 1.0
77
+ dataset_type: physical_embodiment
78
+ val_dataset_path: null
79
+ modality_configs:
80
+ new_embodiment:
81
+ video:
82
+ delta_indices:
83
+ - 0
84
+ modality_keys:
85
+ - cam_high
86
+ - cam_left_wrist
87
+ - cam_side
88
+ sin_cos_embedding_keys: null
89
+ mean_std_embedding_keys: null
90
+ action_configs: null
91
+ state:
92
+ delta_indices:
93
+ - 0
94
+ modality_keys:
95
+ - arm_l
96
+ - arm_l_finger
97
+ sin_cos_embedding_keys: null
98
+ mean_std_embedding_keys: null
99
+ action_configs: null
100
+ action:
101
+ delta_indices:
102
+ - 0
103
+ - 1
104
+ - 2
105
+ - 3
106
+ - 4
107
+ - 5
108
+ - 6
109
+ - 7
110
+ - 8
111
+ - 9
112
+ - 10
113
+ - 11
114
+ - 12
115
+ - 13
116
+ - 14
117
+ - 15
118
+ modality_keys:
119
+ - arm_l
120
+ - arm_l_finger
121
+ sin_cos_embedding_keys: null
122
+ mean_std_embedding_keys: null
123
+ action_configs:
124
+ - rep: RELATIVE
125
+ type: NON_EEF
126
+ format: DEFAULT
127
+ state_key: null
128
+ - rep: ABSOLUTE
129
+ type: NON_EEF
130
+ format: DEFAULT
131
+ state_key: null
132
+ language:
133
+ delta_indices:
134
+ - 0
135
+ modality_keys:
136
+ - annotation.human.task_description
137
+ sin_cos_embedding_keys: null
138
+ mean_std_embedding_keys: null
139
+ action_configs: null
140
+ download_cache: false
141
+ shard_size: 1024
142
+ episode_sampling_rate: 0.1
143
+ num_shards_per_epoch: 100000
144
+ override_pretraining_statistics: true
145
+ mode: single_turn
146
+ random_chop: 0.0
147
+ mock_dataset_mode: false
148
+ shuffle: true
149
+ seed: 42
150
+ multiprocessing_context: fork
151
+ allow_padding: false
152
+ subsample_ratio: 1.0
153
+ image_crop_size:
154
+ - 244
155
+ - 244
156
+ image_target_size:
157
+ - 224
158
+ - 224
159
+ video_backend: torchcodec
160
+ training:
161
+ output_dir: /home/gayatri-davuluri/.cache/huggingface/lerobot/EbotsVLA/gr00t_checkpoints/run_001
162
+ experiment_name: null
163
+ max_steps: 30000
164
+ global_batch_size: 32
165
+ batch_size: null
166
+ gradient_accumulation_steps: 1
167
+ learning_rate: 0.0001
168
+ lr_scheduler_type: cosine
169
+ weight_decay: 1.0e-05
170
+ warmup_ratio: 0.05
171
+ warmup_steps: 0
172
+ max_grad_norm: 1.0
173
+ optim: adamw_torch
174
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
175
+ skip_weight_loading: false
176
+ tf32: true
177
+ fp16: false
178
+ bf16: true
179
+ eval_bf16: true
180
+ logging_steps: 10
181
+ save_steps: 5000
182
+ save_total_limit: 30
183
+ save_vl_model: false
184
+ save_only_model: false
185
+ upload_checkpoints: false
186
+ upload_every: 1000
187
+ upload_last_n_checkpoints: 5
188
+ max_concurrent_uploads: 2
189
+ eval_strategy: 'no'
190
+ eval_steps: 500
191
+ eval_set_split_ratio: 0.1
192
+ eval_batch_size: 2
193
+ save_best_eval_metric_name: ''
194
+ save_best_eval_metric_greater_is_better: true
195
+ deepspeed_stage: 2
196
+ gradient_checkpointing: false
197
+ transformers_trust_remote_code: true
198
+ transformers_local_files_only: false
199
+ transformers_cache_dir: null
200
+ transformers_access_token: null
201
+ use_ddp: false
202
+ ddp_bucket_cap_mb: 100
203
+ num_gpus: 1
204
+ dataloader_num_workers: 8
205
+ remove_unused_columns: false
206
+ use_wandb: true
207
+ wandb_project: finetune-gr00t-n1d7
208
+ enable_profiling: false
209
+ max_retries: 3
210
+ assert_loss_less_than: null
211
+ add_rl_callback: false
212
+ enable_open_loop_eval: false
213
+ open_loop_eval_traj_ids:
214
+ - 0
215
+ open_loop_eval_steps_per_traj: 100
216
+ open_loop_eval_plot_indices: null
217
+ max_steps: 30000
218
+ save_steps: 5000
gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/config.yaml ADDED
@@ -0,0 +1,244 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ - /home/gayatri-davuluri/datasets/cableHarness_v2v3_v3rec13_upd_prompts_Config
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: !!python/object/apply:gr00t.data.types.ActionRepresentation
29
+ - relative
30
+ state_key: null
31
+ type: &id002 !!python/object/apply:gr00t.data.types.ActionType
32
+ - non_eef
33
+ - !!python/object:gr00t.data.types.ActionConfig
34
+ format: *id001
35
+ rep: !!python/object/apply:gr00t.data.types.ActionRepresentation
36
+ - absolute
37
+ state_key: null
38
+ type: *id002
39
+ delta_indices:
40
+ - 0
41
+ - 1
42
+ - 2
43
+ - 3
44
+ - 4
45
+ - 5
46
+ - 6
47
+ - 7
48
+ - 8
49
+ - 9
50
+ - 10
51
+ - 11
52
+ - 12
53
+ - 13
54
+ - 14
55
+ - 15
56
+ mean_std_embedding_keys: null
57
+ modality_keys:
58
+ - arm_l
59
+ - arm_l_finger
60
+ sin_cos_embedding_keys: null
61
+ language: !!python/object:gr00t.data.types.ModalityConfig
62
+ action_configs: null
63
+ delta_indices:
64
+ - 0
65
+ mean_std_embedding_keys: null
66
+ modality_keys:
67
+ - annotation.human.task_description
68
+ sin_cos_embedding_keys: null
69
+ state: !!python/object:gr00t.data.types.ModalityConfig
70
+ action_configs: null
71
+ delta_indices:
72
+ - 0
73
+ mean_std_embedding_keys: null
74
+ modality_keys:
75
+ - arm_l
76
+ - arm_l_finger
77
+ sin_cos_embedding_keys: null
78
+ video: !!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
+ - cam_high
85
+ - cam_left_wrist
86
+ - cam_side
87
+ sin_cos_embedding_keys: null
88
+ mode: single_turn
89
+ multiprocessing_context: fork
90
+ num_shards_per_epoch: 100000
91
+ override_pretraining_statistics: true
92
+ random_chop: 0.0
93
+ seed: 42
94
+ shard_size: 1024
95
+ shuffle: true
96
+ subsample_ratio: 1.0
97
+ video_backend: torchcodec
98
+ load_config_path: null
99
+ model: !!python/object:gr00t.configs.model.gr00t_n1d7.Gr00tN1d7Config
100
+ _attn_implementation_internal: null
101
+ _commit_hash: null
102
+ _name_or_path: ''
103
+ _output_attentions: false
104
+ add_cross_attention: false
105
+ architectures: null
106
+ backbone_trainable_params_fp32: true
107
+ bad_words_ids: null
108
+ begin_suppress_tokens: null
109
+ bos_token_id: null
110
+ chunk_size_feed_forward: 0
111
+ color_jitter_params:
112
+ brightness: 0.3
113
+ contrast: 0.4
114
+ hue: 0.08
115
+ saturation: 0.5
116
+ cross_attention_hidden_size: null
117
+ decoder_start_token_id: null
118
+ diffusion_model_cfg:
119
+ attention_head_dim: 48
120
+ dropout: 0.2
121
+ final_dropout: true
122
+ interleave_self_attention: true
123
+ norm_type: ada_norm
124
+ num_attention_heads: 32
125
+ num_layers: 16
126
+ output_dim: 1024
127
+ positional_embeddings: null
128
+ diversity_penalty: 0.0
129
+ do_sample: false
130
+ dtype: null
131
+ early_stopping: false
132
+ encoder_no_repeat_ngram_size: 0
133
+ eos_token_id: null
134
+ exponential_decay_length_penalty: null
135
+ extra_augmentation_config: null
136
+ finetuning_task: null
137
+ forced_bos_token_id: null
138
+ forced_eos_token_id: null
139
+ id2label:
140
+ 0: LABEL_0
141
+ 1: LABEL_1
142
+ is_decoder: false
143
+ is_encoder_decoder: false
144
+ label2id:
145
+ LABEL_0: 0
146
+ LABEL_1: 1
147
+ length_penalty: 1.0
148
+ load_bf16: false
149
+ max_length: 20
150
+ min_length: 0
151
+ model_name: nvidia/Cosmos-Reason2-2B
152
+ no_repeat_ngram_size: 0
153
+ num_beam_groups: 1
154
+ num_beams: 1
155
+ num_return_sequences: 1
156
+ output_hidden_states: false
157
+ output_scores: false
158
+ pad_token_id: null
159
+ prefix: null
160
+ problem_type: null
161
+ pruned_heads: {}
162
+ random_rotation_angle: null
163
+ remove_invalid_values: false
164
+ repetition_penalty: 1.0
165
+ reproject_vision: false
166
+ return_dict: true
167
+ return_dict_in_generate: false
168
+ sep_token_id: null
169
+ state_dropout_prob: 0.2
170
+ suppress_tokens: null
171
+ task_specific_params: null
172
+ temperature: 1.0
173
+ tf_legacy_loss: false
174
+ tie_encoder_decoder: false
175
+ tie_word_embeddings: true
176
+ tokenizer_class: null
177
+ top_k: 50
178
+ top_p: 1.0
179
+ torchscript: false
180
+ transformers_version: null
181
+ tune_diffusion_model: true
182
+ tune_llm: false
183
+ tune_projector: true
184
+ tune_visual: false
185
+ typical_p: 1.0
186
+ use_bfloat16: false
187
+ use_relative_action: true
188
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
189
+ add_rl_callback: false
190
+ assert_loss_less_than: null
191
+ batch_size: null
192
+ bf16: true
193
+ dataloader_num_workers: 8
194
+ ddp_bucket_cap_mb: 100
195
+ deepspeed_stage: 2
196
+ enable_open_loop_eval: false
197
+ enable_profiling: false
198
+ eval_batch_size: 2
199
+ eval_bf16: true
200
+ eval_set_split_ratio: 0.1
201
+ eval_steps: 500
202
+ eval_strategy: 'no'
203
+ experiment_name: null
204
+ fp16: false
205
+ global_batch_size: 32
206
+ gradient_accumulation_steps: 1
207
+ gradient_checkpointing: false
208
+ learning_rate: 0.0001
209
+ logging_steps: 10
210
+ lr_scheduler_type: cosine
211
+ max_concurrent_uploads: 2
212
+ max_grad_norm: 1.0
213
+ max_retries: 3
214
+ max_steps: 30000
215
+ num_gpus: 1
216
+ open_loop_eval_plot_indices: null
217
+ open_loop_eval_steps_per_traj: 100
218
+ open_loop_eval_traj_ids:
219
+ - 0
220
+ optim: adamw_torch
221
+ output_dir: /home/gayatri-davuluri/.cache/huggingface/lerobot/EbotsVLA/gr00t_checkpoints/run_001
222
+ remove_unused_columns: false
223
+ save_best_eval_metric_greater_is_better: true
224
+ save_best_eval_metric_name: ''
225
+ save_only_model: false
226
+ save_steps: 5000
227
+ save_total_limit: 30
228
+ save_vl_model: false
229
+ skip_weight_loading: false
230
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
231
+ tf32: true
232
+ transformers_access_token: null
233
+ transformers_cache_dir: null
234
+ transformers_local_files_only: false
235
+ transformers_trust_remote_code: true
236
+ upload_checkpoints: false
237
+ upload_every: 1000
238
+ upload_last_n_checkpoints: 5
239
+ use_ddp: false
240
+ use_wandb: true
241
+ wandb_project: finetune-gr00t-n1d7
242
+ warmup_ratio: 0.05
243
+ warmup_steps: 0
244
+ weight_decay: 1.0e-05
gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,932 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "arm_l": {
5
+ "min": [
6
+ -0.1006389930844307,
7
+ -0.3946418762207031,
8
+ -0.08660126477479935,
9
+ -3.053096294403076,
10
+ -2.5050132274627686,
11
+ -2.2086756229400635
12
+ ],
13
+ "max": [
14
+ 0.3803256154060364,
15
+ -0.03152116760611534,
16
+ 0.11409957706928253,
17
+ 3.0605881214141846,
18
+ 1.1357500553131104,
19
+ 2.0965044498443604
20
+ ],
21
+ "mean": [
22
+ 0.21154533326625824,
23
+ -0.19236719608306885,
24
+ 0.0059371208772063255,
25
+ 0.41348111629486084,
26
+ -0.10213359445333481,
27
+ 0.12139146775007248
28
+ ],
29
+ "std": [
30
+ 0.14255768060684204,
31
+ 0.10641054064035416,
32
+ 0.041256923228502274,
33
+ 2.2815706729888916,
34
+ 0.6122822761535645,
35
+ 1.5502700805664062
36
+ ],
37
+ "q01": [
38
+ -0.05833661928772926,
39
+ -0.3710125744342804,
40
+ -0.07111530750989914,
41
+ -2.5719981670379637,
42
+ -1.9289566040039063,
43
+ -1.981496000289917
44
+ ],
45
+ "q99": [
46
+ 0.3668590366840363,
47
+ -0.04091229885816572,
48
+ 0.08768809437751793,
49
+ 2.7495419025421146,
50
+ 1.0022219896316529,
51
+ 1.962279200553894
52
+ ]
53
+ },
54
+ "arm_l_finger": {
55
+ "min": [
56
+ 0.0
57
+ ],
58
+ "max": [
59
+ 1.0
60
+ ],
61
+ "mean": [
62
+ 0.44890621304512024
63
+ ],
64
+ "std": [
65
+ 0.49714404344558716
66
+ ],
67
+ "q01": [
68
+ 0.0
69
+ ],
70
+ "q99": [
71
+ 1.0
72
+ ]
73
+ }
74
+ },
75
+ "action": {
76
+ "arm_l": {
77
+ "min": [
78
+ -0.1001979187130928,
79
+ -0.3946418762207031,
80
+ -0.08660126477479935,
81
+ -3.053096294403076,
82
+ -2.5050132274627686,
83
+ -2.2086756229400635
84
+ ],
85
+ "max": [
86
+ 0.3803256154060364,
87
+ -0.03152116760611534,
88
+ 0.11409957706928253,
89
+ 3.0605881214141846,
90
+ 1.1357500553131104,
91
+ 2.0965044498443604
92
+ ],
93
+ "mean": [
94
+ 0.21177242696285248,
95
+ -0.19229459762573242,
96
+ 0.005872717592865229,
97
+ 0.41525575518608093,
98
+ -0.0998842641711235,
99
+ 0.1232220008969307
100
+ ],
101
+ "std": [
102
+ 0.1424476057291031,
103
+ 0.1063520759344101,
104
+ 0.04125887155532837,
105
+ 2.281841993331909,
106
+ 0.6101807951927185,
107
+ 1.5508856773376465
108
+ ],
109
+ "q01": [
110
+ -0.058252960443496704,
111
+ -0.3710835099220276,
112
+ -0.07111530750989914,
113
+ -2.5720277786254884,
114
+ -1.9249901294708252,
115
+ -1.9815640687942504
116
+ ],
117
+ "q99": [
118
+ 0.36717885732650757,
119
+ -0.04090133011341095,
120
+ 0.08725636303424847,
121
+ 2.7495419025421146,
122
+ 1.0022656202316285,
123
+ 1.9637924432754517
124
+ ]
125
+ },
126
+ "arm_l_finger": {
127
+ "min": [
128
+ 0.0
129
+ ],
130
+ "max": [
131
+ 1.0
132
+ ],
133
+ "mean": [
134
+ 0.4476431906223297
135
+ ],
136
+ "std": [
137
+ 0.49733617901802063
138
+ ],
139
+ "q01": [
140
+ 0.0
141
+ ],
142
+ "q99": [
143
+ 1.0
144
+ ]
145
+ }
146
+ },
147
+ "relative_action": {
148
+ "arm_l": {
149
+ "min": [
150
+ [
151
+ -0.004235490225255489,
152
+ -0.0022258460521698,
153
+ -0.0029641762375831604,
154
+ -6.11368465423584,
155
+ -1.0007593631744385,
156
+ -4.141870498657227
157
+ ],
158
+ [
159
+ -0.007333104033023119,
160
+ -0.0038002878427505493,
161
+ -0.005079630762338638,
162
+ -6.109963417053223,
163
+ -1.0420904159545898,
164
+ -4.141870498657227
165
+ ],
166
+ [
167
+ -0.010181523859500885,
168
+ -0.005613610148429871,
169
+ -0.007330911699682474,
170
+ -6.105661869049072,
171
+ -1.0420892238616943,
172
+ -4.141870498657227
173
+ ],
174
+ [
175
+ -0.013596647419035435,
176
+ -0.007086321711540222,
177
+ -0.009969491511583328,
178
+ -6.101454734802246,
179
+ -1.0416295528411865,
180
+ -4.141870498657227
181
+ ],
182
+ [
183
+ -0.016582434996962547,
184
+ -0.008699402213096619,
185
+ -0.012323632836341858,
186
+ -6.097495079040527,
187
+ -1.0410830974578857,
188
+ -4.141870498657227
189
+ ],
190
+ [
191
+ -0.01968994550406933,
192
+ -0.010646328330039978,
193
+ -0.014502190053462982,
194
+ -6.093442916870117,
195
+ -1.0447039604187012,
196
+ -4.141870498657227
197
+ ],
198
+ [
199
+ -0.02321077696979046,
200
+ -0.012145861983299255,
201
+ -0.01675165817141533,
202
+ -6.090091705322266,
203
+ -1.0529625415802002,
204
+ -4.141870498657227
205
+ ],
206
+ [
207
+ -0.026251966133713722,
208
+ -0.013671144843101501,
209
+ -0.018930215388536453,
210
+ -6.086215019226074,
211
+ -1.062105417251587,
212
+ -4.094995021820068
213
+ ],
214
+ [
215
+ -0.029011275619268417,
216
+ -0.015162065625190735,
217
+ -0.021075669676065445,
218
+ -6.0817484855651855,
219
+ -1.0701377391815186,
220
+ -4.094995021820068
221
+ ],
222
+ [
223
+ -0.030858267098665237,
224
+ -0.016241326928138733,
225
+ -0.022946612909436226,
226
+ -6.076840877532959,
227
+ -1.079089879989624,
228
+ -4.094995021820068
229
+ ],
230
+ [
231
+ -0.030858267098665237,
232
+ -0.01686413586139679,
233
+ -0.024742793291807175,
234
+ -6.073434352874756,
235
+ -1.0620386600494385,
236
+ -4.094995021820068
237
+ ],
238
+ [
239
+ -0.030665498226881027,
240
+ -0.01831965148448944,
241
+ -0.027076484635472298,
242
+ -6.071956634521484,
243
+ -1.03946053981781,
244
+ -4.094995021820068
245
+ ],
246
+ [
247
+ -0.030776353552937508,
248
+ -0.019868552684783936,
249
+ -0.029044806957244873,
250
+ -6.076470375061035,
251
+ -1.0394726991653442,
252
+ -4.094995021820068
253
+ ],
254
+ [
255
+ -0.030368803068995476,
256
+ -0.021434292197227478,
257
+ -0.030379850417375565,
258
+ -6.072749137878418,
259
+ -1.0644713640213013,
260
+ -4.094995021820068
261
+ ],
262
+ [
263
+ -0.028749702498316765,
264
+ -0.022919461131095886,
265
+ -0.03225233405828476,
266
+ -6.068447589874268,
267
+ -1.0686957836151123,
268
+ -4.094995021820068
269
+ ],
270
+ [
271
+ -0.028726020827889442,
272
+ -0.024467021226882935,
273
+ -0.03399868309497833,
274
+ -6.064240455627441,
275
+ -1.0724029541015625,
276
+ -4.094995021820068
277
+ ]
278
+ ],
279
+ "max": [
280
+ [
281
+ 0.004725892096757889,
282
+ 0.0027886033058166504,
283
+ 0.0022737495601177216,
284
+ 6.035301685333252,
285
+ 1.474894642829895,
286
+ 4.174610614776611
287
+ ],
288
+ [
289
+ 0.008718788623809814,
290
+ 0.004971235990524292,
291
+ 0.003996998071670532,
292
+ 6.030788421630859,
293
+ 1.477924108505249,
294
+ 4.174603462219238
295
+ ],
296
+ [
297
+ 0.010449156165122986,
298
+ 0.007192403078079224,
299
+ 0.005943775177001953,
300
+ 6.026132106781006,
301
+ 1.4809396266937256,
302
+ 4.1746063232421875
303
+ ],
304
+ [
305
+ 0.012407854199409485,
306
+ 0.00953701138496399,
307
+ 0.00783834233880043,
308
+ 6.024982452392578,
309
+ 1.4844553470611572,
310
+ 4.17458963394165
311
+ ],
312
+ [
313
+ 0.013578694313764572,
314
+ 0.01195332407951355,
315
+ 0.009777691215276718,
316
+ 6.027261734008789,
317
+ 1.4884464740753174,
318
+ 4.174583435058594
319
+ ],
320
+ [
321
+ 0.015893515199422836,
322
+ 0.014174491167068481,
323
+ 0.011414855718612671,
324
+ 6.0297393798828125,
325
+ 1.4888871908187866,
326
+ 4.174566268920898
327
+ ],
328
+ [
329
+ 0.018296536058187485,
330
+ 0.016482263803482056,
331
+ 0.01325889304280281,
332
+ 6.031765460968018,
333
+ 1.5922274589538574,
334
+ 4.174569129943848
335
+ ],
336
+ [
337
+ 0.020938966423273087,
338
+ 0.018898576498031616,
339
+ 0.014180425554513931,
340
+ 6.03474760055542,
341
+ 1.669804573059082,
342
+ 4.174562454223633
343
+ ],
344
+ [
345
+ 0.023059602826833725,
346
+ 0.021119743585586548,
347
+ 0.016026385128498077,
348
+ 6.036848068237305,
349
+ 1.6720974445343018,
350
+ 4.174544334411621
351
+ ],
352
+ [
353
+ 0.02528684213757515,
354
+ 0.023333728313446045,
355
+ 0.01709432154893875,
356
+ 6.045999526977539,
357
+ 1.6746065616607666,
358
+ 4.174551963806152
359
+ ],
360
+ [
361
+ 0.027646947652101517,
362
+ 0.02493080496788025,
363
+ 0.017966773360967636,
364
+ 6.054683685302734,
365
+ 1.6769334077835083,
366
+ 4.1745405197143555
367
+ ],
368
+ [
369
+ 0.02964061126112938,
370
+ 0.026324838399887085,
371
+ 0.018863672390580177,
372
+ 6.063432693481445,
373
+ 1.6792851686477661,
374
+ 4.174548149108887
375
+ ],
376
+ [
377
+ 0.03165138140320778,
378
+ 0.02792191505432129,
379
+ 0.020340880379080772,
380
+ 6.072515964508057,
381
+ 1.6815807819366455,
382
+ 4.174520492553711
383
+ ],
384
+ [
385
+ 0.033690545707941055,
386
+ 0.029246509075164795,
387
+ 0.021087855100631714,
388
+ 6.050999641418457,
389
+ 1.6830500364303589,
390
+ 4.17452335357666
391
+ ],
392
+ [
393
+ 0.03488786518573761,
394
+ 0.030564427375793457,
395
+ 0.02272876352071762,
396
+ 6.043996334075928,
397
+ 1.6830500364303589,
398
+ 4.174544334411621
399
+ ],
400
+ [
401
+ 0.03697727620601654,
402
+ 0.03216150403022766,
403
+ 0.023145724087953568,
404
+ 6.033823490142822,
405
+ 1.6830500364303589,
406
+ 4.174530029296875
407
+ ]
408
+ ],
409
+ "mean": [
410
+ [
411
+ 0.0002253354905406013,
412
+ 8.495293150190264e-05,
413
+ -9.155122097581625e-05,
414
+ 0.001859949086792767,
415
+ 0.0023580819834023714,
416
+ 0.0019161647651344538
417
+ ],
418
+ [
419
+ 0.0004514687752816826,
420
+ 0.00016951623547356576,
421
+ -0.00018200863269157708,
422
+ 0.0036681420169770718,
423
+ 0.004713411908596754,
424
+ 0.0037944307550787926
425
+ ],
426
+ [
427
+ 0.0006781455012969673,
428
+ 0.0002537094405852258,
429
+ -0.00027126900386065245,
430
+ 0.005369907710701227,
431
+ 0.00705572497099638,
432
+ 0.005595036316663027
433
+ ],
434
+ [
435
+ 0.0009054223774001002,
436
+ 0.0003374245425220579,
437
+ -0.00035926984855905175,
438
+ 0.006865270901471376,
439
+ 0.009383310563862324,
440
+ 0.007242748513817787
441
+ ],
442
+ [
443
+ 0.0011333131697028875,
444
+ 0.0004206871963106096,
445
+ -0.0004459292977117002,
446
+ 0.008305021561682224,
447
+ 0.011696092784404755,
448
+ 0.008852104656398296
449
+ ],
450
+ [
451
+ 0.0013618997763842344,
452
+ 0.000503317394759506,
453
+ -0.0005310149863362312,
454
+ 0.009689897298812866,
455
+ 0.013994180597364902,
456
+ 0.010424453765153885
457
+ ],
458
+ [
459
+ 0.0015911730006337166,
460
+ 0.0005851986934430897,
461
+ -0.0006143946666270494,
462
+ 0.011067060753703117,
463
+ 0.016274532303214073,
464
+ 0.011996821500360966
465
+ ],
466
+ [
467
+ 0.001821166486479342,
468
+ 0.0006663412787020206,
469
+ -0.0006958885933272541,
470
+ 0.012386387214064598,
471
+ 0.018535438925027847,
472
+ 0.013530325144529343
473
+ ],
474
+ [
475
+ 0.0020517874509096146,
476
+ 0.0007468481780961156,
477
+ -0.0007755015976727009,
478
+ 0.013644667342305183,
479
+ 0.020774630829691887,
480
+ 0.015025388449430466
481
+ ],
482
+ [
483
+ 0.0022829726804047823,
484
+ 0.0008266648510470986,
485
+ -0.0008532237261533737,
486
+ 0.014741025865077972,
487
+ 0.022989491000771523,
488
+ 0.016405878588557243
489
+ ],
490
+ [
491
+ 0.0025146545376628637,
492
+ 0.000905906199477613,
493
+ -0.000929144851397723,
494
+ 0.01582816243171692,
495
+ 0.02518182247877121,
496
+ 0.017787287011742592
497
+ ],
498
+ [
499
+ 0.0027467007748782635,
500
+ 0.0009845839813351631,
501
+ -0.001003241864964366,
502
+ 0.016854768618941307,
503
+ 0.027351414784789085,
504
+ 0.019132064655423164
505
+ ],
506
+ [
507
+ 0.002978826640173793,
508
+ 0.0010626959847286344,
509
+ -0.001075556967407465,
510
+ 0.01782034896314144,
511
+ 0.02949974685907364,
512
+ 0.02044137567281723
513
+ ],
514
+ [
515
+ 0.0032108735758811235,
516
+ 0.0011402673553675413,
517
+ -0.0011462272377684712,
518
+ 0.01867625303566456,
519
+ 0.0316278412938118,
520
+ 0.021675975993275642
521
+ ],
522
+ [
523
+ 0.0034426709171384573,
524
+ 0.001217387616634369,
525
+ -0.0012154292780905962,
526
+ 0.019469983875751495,
527
+ 0.03373156115412712,
528
+ 0.022874409332871437
529
+ ],
530
+ [
531
+ 0.0036739015486091375,
532
+ 0.0012943618930876255,
533
+ -0.0012838816037401557,
534
+ 0.02025437355041504,
535
+ 0.03581295162439346,
536
+ 0.02407418005168438
537
+ ]
538
+ ],
539
+ "std": [
540
+ [
541
+ 0.0004978042561560869,
542
+ 0.0004314902180340141,
543
+ 0.0004494226013775915,
544
+ 0.21372829377651215,
545
+ 0.021313343197107315,
546
+ 0.1533714383840561
547
+ ],
548
+ [
549
+ 0.0009576921002008021,
550
+ 0.0008236853173002601,
551
+ 0.0008601380977779627,
552
+ 0.29378029704093933,
553
+ 0.03246835991740227,
554
+ 0.21048009395599365
555
+ ],
556
+ [
557
+ 0.001402944908477366,
558
+ 0.0012000853894278407,
559
+ 0.0012554135173559189,
560
+ 0.3575674295425415,
561
+ 0.04277355596423149,
562
+ 0.2562621831893921
563
+ ],
564
+ [
565
+ 0.0018342790426686406,
566
+ 0.0015627779066562653,
567
+ 0.0016366345807909966,
568
+ 0.40969276428222656,
569
+ 0.05243466794490814,
570
+ 0.293476402759552
571
+ ],
572
+ [
573
+ 0.002252009464427829,
574
+ 0.0019135181792080402,
575
+ 0.0020054858177900314,
576
+ 0.45436400175094604,
577
+ 0.061222851276397705,
578
+ 0.3255850672721863
579
+ ],
580
+ [
581
+ 0.0026586749590933323,
582
+ 0.0022550737485289574,
583
+ 0.0023641837760806084,
584
+ 0.4916367828845978,
585
+ 0.06939713656902313,
586
+ 0.3522043228149414
587
+ ],
588
+ [
589
+ 0.003055528737604618,
590
+ 0.0025886634830385447,
591
+ 0.002714423695579171,
592
+ 0.5251379013061523,
593
+ 0.077008917927742,
594
+ 0.37645071744918823
595
+ ],
596
+ [
597
+ 0.0034438674338161945,
598
+ 0.0029149476904422045,
599
+ 0.003057266352698207,
600
+ 0.5554863214492798,
601
+ 0.08409731090068817,
602
+ 0.3983165919780731
603
+ ],
604
+ [
605
+ 0.003822925966233015,
606
+ 0.003234784584492445,
607
+ 0.0033933527301996946,
608
+ 0.5840228199958801,
609
+ 0.09070045500993729,
610
+ 0.41887974739074707
611
+ ],
612
+ [
613
+ 0.004194746259599924,
614
+ 0.0035491236485540867,
615
+ 0.003723748493939638,
616
+ 0.6106894016265869,
617
+ 0.09698854386806488,
618
+ 0.4379228353500366
619
+ ],
620
+ [
621
+ 0.004558791872113943,
622
+ 0.003858424024656415,
623
+ 0.00404907064512372,
624
+ 0.6352975368499756,
625
+ 0.1027967780828476,
626
+ 0.4557587504386902
627
+ ],
628
+ [
629
+ 0.004915650002658367,
630
+ 0.004162661265581846,
631
+ 0.0043701897375285625,
632
+ 0.6579070687294006,
633
+ 0.10835184901952744,
634
+ 0.4723413586616516
635
+ ],
636
+ [
637
+ 0.005265559069812298,
638
+ 0.004462769255042076,
639
+ 0.004688119515776634,
640
+ 0.6795804500579834,
641
+ 0.1137050986289978,
642
+ 0.48824542760849
643
+ ],
644
+ [
645
+ 0.005609154235571623,
646
+ 0.0047590043395757675,
647
+ 0.005002682562917471,
648
+ 0.6993448734283447,
649
+ 0.11882363259792328,
650
+ 0.5030515789985657
651
+ ],
652
+ [
653
+ 0.005946788936853409,
654
+ 0.005051432177424431,
655
+ 0.00531320134177804,
656
+ 0.7176821827888489,
657
+ 0.12371432781219482,
658
+ 0.5168843269348145
659
+ ],
660
+ [
661
+ 0.006278567481786013,
662
+ 0.005339807365089655,
663
+ 0.005618891678750515,
664
+ 0.7348477244377136,
665
+ 0.12838014960289001,
666
+ 0.5298192501068115
667
+ ]
668
+ ],
669
+ "q01": [
670
+ [
671
+ -0.0008204290643334389,
672
+ -0.0011303499341011047,
673
+ -0.0017389055341482163,
674
+ -0.012860989570617676,
675
+ -0.01610390217974782,
676
+ -0.028635123372077943
677
+ ],
678
+ [
679
+ -0.0015023794025182725,
680
+ -0.0021064475178718565,
681
+ -0.003339956887066364,
682
+ -0.02503448724746704,
683
+ -0.03360688611865044,
684
+ -0.053917679190635684
685
+ ],
686
+ [
687
+ -0.0021307291463017464,
688
+ -0.0029923141002655028,
689
+ -0.004898873530328274,
690
+ -0.037855756282806394,
691
+ -0.0729036569595337,
692
+ -0.0768855020403862
693
+ ],
694
+ [
695
+ -0.002726387791335583,
696
+ -0.0038163319230079653,
697
+ -0.006357023306190967,
698
+ -0.05010286569595337,
699
+ -0.10365459620952606,
700
+ -0.10222572833299637
701
+ ],
702
+ [
703
+ -0.0032588220201432705,
704
+ -0.004503662884235382,
705
+ -0.00773897310718894,
706
+ -0.061943423748016355,
707
+ -0.12861181795597076,
708
+ -0.12782760113477706
709
+ ],
710
+ [
711
+ -0.003741525113582611,
712
+ -0.0052297510206699375,
713
+ -0.009075716324150562,
714
+ -0.07252004146575927,
715
+ -0.14615237712860107,
716
+ -0.15432221293449402
717
+ ],
718
+ [
719
+ -0.0042274147272109985,
720
+ -0.005866098403930664,
721
+ -0.010316292708739638,
722
+ -0.08401930332183838,
723
+ -0.16078983545303344,
724
+ -0.18014225363731384
725
+ ],
726
+ [
727
+ -0.0046352967619895935,
728
+ -0.006521512567996979,
729
+ -0.011614927370101214,
730
+ -0.09547014236450195,
731
+ -0.1721242845058441,
732
+ -0.20546859055757524
733
+ ],
734
+ [
735
+ -0.00501422774977982,
736
+ -0.007206776738166809,
737
+ -0.012884581089019775,
738
+ -0.10706287622451782,
739
+ -0.17999446392059326,
740
+ -0.23390269577503206
741
+ ],
742
+ [
743
+ -0.005326274037361145,
744
+ -0.007821504771709443,
745
+ -0.014078816398978233,
746
+ -0.11749796867370606,
747
+ -0.18854969143867492,
748
+ -0.265094630420208
749
+ ],
750
+ [
751
+ -0.00570037174038589,
752
+ -0.00846915990114212,
753
+ -0.015224083513021468,
754
+ -0.13562406301498414,
755
+ -0.19591855332255365,
756
+ -0.3018211334943771
757
+ ],
758
+ [
759
+ -0.005989532545208931,
760
+ -0.009087064862251281,
761
+ -0.01645280374214053,
762
+ -0.15705504417419433,
763
+ -0.20118582919239997,
764
+ -0.3393710017204285
765
+ ],
766
+ [
767
+ -0.006305968761444092,
768
+ -0.009709213674068452,
769
+ -0.01762270824983716,
770
+ -0.18729126453399658,
771
+ -0.20703091248869895,
772
+ -0.38604599982500076
773
+ ],
774
+ [
775
+ -0.006614441052079201,
776
+ -0.01031842976808548,
777
+ -0.018795522022992374,
778
+ -0.23054687976837163,
779
+ -0.21339795887470245,
780
+ -0.44858020544052124
781
+ ],
782
+ [
783
+ -0.0069403484463691715,
784
+ -0.010942097008228301,
785
+ -0.019896874949336053,
786
+ -0.3410740494728089,
787
+ -0.21772613152861595,
788
+ -0.5584201753139496
789
+ ],
790
+ [
791
+ -0.007210374297574163,
792
+ -0.011553259193897247,
793
+ -0.020932440273463727,
794
+ -4.7844929695129395,
795
+ -0.22433793544769287,
796
+ -2.0315409183502204
797
+ ]
798
+ ],
799
+ "q99": [
800
+ [
801
+ 0.0018875066190958045,
802
+ 0.001507572829723359,
803
+ 0.0012135540135204802,
804
+ 0.03193371295928958,
805
+ 0.07497657090425523,
806
+ 0.02459929361939432
807
+ ],
808
+ [
809
+ 0.003629810363054278,
810
+ 0.0028538174927234673,
811
+ 0.002308204211294654,
812
+ 0.05978170037269595,
813
+ 0.13217666968703304,
814
+ 0.04792829751968387
815
+ ],
816
+ [
817
+ 0.005342817306518562,
818
+ 0.004173709452152253,
819
+ 0.0033831471577286765,
820
+ 0.08642190694808999,
821
+ 0.1821535259485245,
822
+ 0.0700379624962807
823
+ ],
824
+ [
825
+ 0.0070166064426302925,
826
+ 0.005508905649185183,
827
+ 0.004381483420729651,
828
+ 0.11254231929779084,
829
+ 0.22703098356723847,
830
+ 0.09310215115547187
831
+ ],
832
+ [
833
+ 0.00869046375155449,
834
+ 0.006822316348552707,
835
+ 0.0053444372490048615,
836
+ 0.1439155519008637,
837
+ 0.2758188933134099,
838
+ 0.11646757125854502
839
+ ],
840
+ [
841
+ 0.010331999883055708,
842
+ 0.008125563710927963,
843
+ 0.006281240656971951,
844
+ 0.17151316404342729,
845
+ 0.3156844452023522,
846
+ 0.14013778865337395
847
+ ],
848
+ [
849
+ 0.011938411463052053,
850
+ 0.009407697618007665,
851
+ 0.0072881287895143225,
852
+ 0.20220767259597838,
853
+ 0.3535799562931061,
854
+ 0.16554524600505843
855
+ ],
856
+ [
857
+ 0.013510413141921163,
858
+ 0.010672427713871006,
859
+ 0.00807656031101945,
860
+ 0.23618569374084514,
861
+ 0.381205736100674,
862
+ 0.19086208939552332
863
+ ],
864
+ [
865
+ 0.015041159372776754,
866
+ 0.011941570043563854,
867
+ 0.009000060334801729,
868
+ 0.2714147567749042,
869
+ 0.40975958108902033,
870
+ 0.2197139568626883
871
+ ],
872
+ [
873
+ 0.016608255170285705,
874
+ 0.013171589374542238,
875
+ 0.009889581240713602,
876
+ 0.3098403930664082,
877
+ 0.4377403214573874,
878
+ 0.24962427020073147
879
+ ],
880
+ [
881
+ 0.018111442215740684,
882
+ 0.01435973420739174,
883
+ 0.010811550915241285,
884
+ 0.35505937039852736,
885
+ 0.460975171625615,
886
+ 0.2836328268051164
887
+ ],
888
+ [
889
+ 0.019602144137024914,
890
+ 0.015561373531818395,
891
+ 0.011795567814260754,
892
+ 0.4108364462852569,
893
+ 0.4815591961145407,
894
+ 0.3245323121547729
895
+ ],
896
+ [
897
+ 0.0210545476526022,
898
+ 0.016788020730018637,
899
+ 0.012623235769569922,
900
+ 0.4795063257217571,
901
+ 0.502637743949891,
902
+ 0.3608253687620175
903
+ ],
904
+ [
905
+ 0.022485633753240124,
906
+ 0.01797343343496323,
907
+ 0.013562189228832755,
908
+ 0.686625534296049,
909
+ 0.522907102108002,
910
+ 0.4441104859113831
911
+ ],
912
+ [
913
+ 0.02378636598587038,
914
+ 0.01914146840572359,
915
+ 0.01462116651237015,
916
+ 4.622694611549377,
917
+ 0.5465678155422211,
918
+ 3.028123152256041
919
+ ],
920
+ [
921
+ 0.0251483490690589,
922
+ 0.02033632546663285,
923
+ 0.015623088181018837,
924
+ 4.752910733222961,
925
+ 0.5665670424699796,
926
+ 3.3086502432823393
927
+ ]
928
+ ]
929
+ }
930
+ }
931
+ }
932
+ }
gr00t/n1p7_real_data_cart/checkpoint-30000/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": 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
+ }
gr00t/n1p7_real_data_cart/checkpoint-30000/experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t/n1p7_real_data_cart/checkpoint-30000/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
gr00t/n1p7_real_data_cart/checkpoint-30000/model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1512508517d8b70023ce8fd89bd43e507608ac69f18978c4522a13f0db7ef263
3
+ size 4970792616
gr00t/n1p7_real_data_cart/checkpoint-30000/model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e24adb2bec417d3ac86187528585ae911899dde98df70107ab1def0b156cdca3
3
+ size 2618758696
gr00t/n1p7_real_data_cart/checkpoint-30000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t/n1p7_real_data_cart/checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ae9c6d19f68a99d7d1f3e8f73643eec5d8190de7e68c214633e39600f645fd4
3
+ size 12964594710
gr00t/n1p7_real_data_cart/checkpoint-30000/processor_config.json ADDED
@@ -0,0 +1,1124 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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_high",
1019
+ "cam_left_wrist",
1020
+ "cam_side"
1021
+ ],
1022
+ "sin_cos_embedding_keys": null,
1023
+ "mean_std_embedding_keys": null,
1024
+ "action_configs": null
1025
+ },
1026
+ "state": {
1027
+ "delta_indices": [
1028
+ 0
1029
+ ],
1030
+ "modality_keys": [
1031
+ "arm_l",
1032
+ "arm_l_finger"
1033
+ ],
1034
+ "sin_cos_embedding_keys": null,
1035
+ "mean_std_embedding_keys": null,
1036
+ "action_configs": null
1037
+ },
1038
+ "action": {
1039
+ "delta_indices": [
1040
+ 0,
1041
+ 1,
1042
+ 2,
1043
+ 3,
1044
+ 4,
1045
+ 5,
1046
+ 6,
1047
+ 7,
1048
+ 8,
1049
+ 9,
1050
+ 10,
1051
+ 11,
1052
+ 12,
1053
+ 13,
1054
+ 14,
1055
+ 15
1056
+ ],
1057
+ "modality_keys": [
1058
+ "arm_l",
1059
+ "arm_l_finger"
1060
+ ],
1061
+ "sin_cos_embedding_keys": null,
1062
+ "mean_std_embedding_keys": null,
1063
+ "action_configs": [
1064
+ {
1065
+ "rep": "RELATIVE",
1066
+ "type": "NON_EEF",
1067
+ "format": "DEFAULT",
1068
+ "state_key": null
1069
+ },
1070
+ {
1071
+ "rep": "ABSOLUTE",
1072
+ "type": "NON_EEF",
1073
+ "format": "DEFAULT",
1074
+ "state_key": null
1075
+ }
1076
+ ]
1077
+ },
1078
+ "language": {
1079
+ "delta_indices": [
1080
+ 0
1081
+ ],
1082
+ "modality_keys": [
1083
+ "annotation.human.task_description"
1084
+ ],
1085
+ "sin_cos_embedding_keys": null,
1086
+ "mean_std_embedding_keys": null,
1087
+ "action_configs": null
1088
+ }
1089
+ }
1090
+ },
1091
+ "image_crop_size": [
1092
+ 230,
1093
+ 230
1094
+ ],
1095
+ "image_target_size": [
1096
+ 256,
1097
+ 256
1098
+ ],
1099
+ "use_albumentations": true,
1100
+ "random_rotation_angle": 0,
1101
+ "color_jitter_params": {
1102
+ "brightness": 0.3,
1103
+ "contrast": 0.4,
1104
+ "saturation": 0.5,
1105
+ "hue": 0.08
1106
+ },
1107
+ "shortest_image_edge": 256,
1108
+ "crop_fraction": 0.95,
1109
+ "letter_box_transform": false,
1110
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1111
+ "model_type": "qwen",
1112
+ "formalize_language": true,
1113
+ "max_state_dim": 132,
1114
+ "max_action_dim": 132,
1115
+ "max_action_horizon": 40,
1116
+ "use_percentiles": true,
1117
+ "use_mean_std": false,
1118
+ "clip_outliers": true,
1119
+ "apply_sincos_state_encoding": false,
1120
+ "use_relative_action": true,
1121
+ "exclude_state": false,
1122
+ "state_dropout_prob": 0.2
1123
+ }
1124
+ }
gr00t/n1p7_real_data_cart/checkpoint-30000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2676f41e1ae15cf232df26612357da80277e17a1e8509eee71515edf0dfa4ab9
3
+ size 14645
gr00t/n1p7_real_data_cart/checkpoint-30000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:267d574a13475f9a04d019b2bb837a7b9e4f16114c2bd21c58193dd1ba7892a1
3
+ size 1465
gr00t/n1p7_real_data_cart/checkpoint-30000/statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t/n1p7_real_data_cart/checkpoint-30000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
gr00t/n1p7_real_data_cart/checkpoint-30000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1577b1a68aa4dc193c3bea7d2aa2dc27b439e5d28ce7f3999b716236664fe400
3
+ size 5969
gr00t/n1p7_real_data_cart/checkpoint-30000/wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d7", "run_id": "run_001"}