Hugging Face's logo

ClOBOT
/
pick_cube_n17_abs

Safetensors
Gr00tN1d7
Model card Files
xet
 Community
minkyuNoah commited on
Commit
83b4e27
·
verified ·
1 Parent(s): 33550d2

Upload checkpoint-10000

Browse files
Files changed (19) hide show
  1. config.json +88 -0
  2. embodiment_id.json +55 -0
  3. experiment_cfg/conf.yaml +246 -0
  4. experiment_cfg/config.yaml +273 -0
  5. experiment_cfg/dataset_statistics.json +0 -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. optimizer.pt +3 -0
  13. processor_config.json +1156 -0
  14. rng_state.pth +3 -0
  15. scheduler.pt +3 -0
  16. statistics.json +0 -0
  17. trainer_state.json +0 -0
  18. training_args.bin +3 -0
  19. 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": 0.2,
63
+ "state_gaussian_noise_std": 0.0,
64
+ "transformers_version": "4.57.3",
65
+ "tune_diffusion_model": true,
66
+ "tune_linear": true,
67
+ "tune_llm": false,
68
+ "tune_projector": true,
69
+ "tune_top_llm_layers": 0,
70
+ "tune_visual": false,
71
+ "tune_vlln": true,
72
+ "use_albumentations": true,
73
+ "use_alternate_vl_dit": true,
74
+ "use_flash_attention": true,
75
+ "use_future_tokens": false,
76
+ "use_mean_std": false,
77
+ "use_percentiles": true,
78
+ "use_vl_self_attention": true,
79
+ "use_vlln": true,
80
+ "vl_self_attention_cfg": {
81
+ "attention_head_dim": 64,
82
+ "dropout": 0.2,
83
+ "final_dropout": true,
84
+ "num_attention_heads": 32,
85
+ "num_layers": 4,
86
+ "positional_embeddings": null
87
+ }
88
+ }
embodiment_id.json ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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 +1,246 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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/joshua/Datasets/pick_cube_on_basket/
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_ego_view
86
+ - cam_left_wrist
87
+ - cam_right_wrist
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
+ - left_arm
96
+ - right_arm
97
+ - left_hand
98
+ - right_hand
99
+ sin_cos_embedding_keys: null
100
+ mean_std_embedding_keys: null
101
+ action_configs: null
102
+ action:
103
+ delta_indices:
104
+ - 0
105
+ - 1
106
+ - 2
107
+ - 3
108
+ - 4
109
+ - 5
110
+ - 6
111
+ - 7
112
+ - 8
113
+ - 9
114
+ - 10
115
+ - 11
116
+ - 12
117
+ - 13
118
+ - 14
119
+ - 15
120
+ - 16
121
+ - 17
122
+ - 18
123
+ - 19
124
+ - 20
125
+ - 21
126
+ - 22
127
+ - 23
128
+ - 24
129
+ - 25
130
+ - 26
131
+ - 27
132
+ - 28
133
+ - 29
134
+ - 30
135
+ - 31
136
+ modality_keys:
137
+ - left_arm
138
+ - right_arm
139
+ - left_hand
140
+ - right_hand
141
+ sin_cos_embedding_keys: null
142
+ mean_std_embedding_keys: null
143
+ action_configs:
144
+ - rep: ABSOLUTE
145
+ type: NON_EEF
146
+ format: DEFAULT
147
+ state_key: null
148
+ - rep: ABSOLUTE
149
+ type: NON_EEF
150
+ format: DEFAULT
151
+ state_key: null
152
+ - rep: ABSOLUTE
153
+ type: NON_EEF
154
+ format: DEFAULT
155
+ state_key: null
156
+ - rep: ABSOLUTE
157
+ type: NON_EEF
158
+ format: DEFAULT
159
+ state_key: null
160
+ language:
161
+ delta_indices:
162
+ - 0
163
+ modality_keys:
164
+ - annotation.human.task_description
165
+ sin_cos_embedding_keys: null
166
+ mean_std_embedding_keys: null
167
+ action_configs: null
168
+ download_cache: false
169
+ shard_size: 1024
170
+ episode_sampling_rate: 0.1
171
+ num_shards_per_epoch: 100000
172
+ override_pretraining_statistics: true
173
+ mode: single_turn
174
+ random_chop: 0.0
175
+ mock_dataset_mode: false
176
+ shuffle: true
177
+ seed: 42
178
+ multiprocessing_context: fork
179
+ allow_padding: false
180
+ subsample_ratio: 1.0
181
+ image_crop_size:
182
+ - 244
183
+ - 244
184
+ image_target_size:
185
+ - 224
186
+ - 224
187
+ video_backend: torchcodec
188
+ training:
189
+ output_dir: /home/joshua/checkpoints/pick_cube_n17_abs
190
+ experiment_name: null
191
+ max_steps: 10000
192
+ global_batch_size: 128
193
+ batch_size: null
194
+ gradient_accumulation_steps: 1
195
+ learning_rate: 0.0001
196
+ lr_scheduler_type: cosine
197
+ weight_decay: 1.0e-05
198
+ warmup_ratio: 0.05
199
+ warmup_steps: 0
200
+ max_grad_norm: 1.0
201
+ optim: adamw_torch
202
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
203
+ skip_weight_loading: false
204
+ tf32: true
205
+ fp16: false
206
+ bf16: true
207
+ eval_bf16: true
208
+ logging_steps: 10
209
+ save_steps: 1000
210
+ save_total_limit: 5
211
+ save_vl_model: false
212
+ save_only_model: false
213
+ upload_checkpoints: false
214
+ upload_every: 1000
215
+ upload_last_n_checkpoints: 5
216
+ max_concurrent_uploads: 2
217
+ eval_strategy: 'no'
218
+ eval_steps: 500
219
+ eval_set_split_ratio: 0.1
220
+ eval_batch_size: 2
221
+ save_best_eval_metric_name: ''
222
+ save_best_eval_metric_greater_is_better: true
223
+ deepspeed_stage: 2
224
+ gradient_checkpointing: false
225
+ transformers_trust_remote_code: true
226
+ transformers_local_files_only: false
227
+ transformers_cache_dir: null
228
+ transformers_access_token: null
229
+ use_ddp: false
230
+ ddp_bucket_cap_mb: 100
231
+ num_gpus: 1
232
+ dataloader_num_workers: 8
233
+ remove_unused_columns: false
234
+ use_wandb: false
235
+ wandb_project: finetune-gr00t-n1d7
236
+ enable_profiling: false
237
+ max_retries: 3
238
+ assert_loss_less_than: null
239
+ add_rl_callback: false
240
+ enable_open_loop_eval: false
241
+ open_loop_eval_traj_ids:
242
+ - 0
243
+ open_loop_eval_steps_per_traj: 100
244
+ open_loop_eval_plot_indices: null
245
+ max_steps: 10000
246
+ save_steps: 1000
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,273 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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/joshua/Datasets/pick_cube_on_basket/
8
+ dataset_type: physical_embodiment
9
+ embodiment_tag: new_embodiment
10
+ mix_ratio: 1.0
11
+ val_dataset_path: null
12
+ download_cache: false
13
+ episode_sampling_rate: 0.1
14
+ image_crop_size:
15
+ - 244
16
+ - 244
17
+ image_target_size:
18
+ - 224
19
+ - 224
20
+ mock_dataset_mode: false
21
+ modality_configs:
22
+ new_embodiment:
23
+ action: !!python/object:gr00t.data.types.ModalityConfig
24
+ action_configs:
25
+ - !!python/object:gr00t.data.types.ActionConfig
26
+ format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
27
+ - default
28
+ rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
29
+ - absolute
30
+ state_key: null
31
+ type: &id003 !!python/object/apply:gr00t.data.types.ActionType
32
+ - non_eef
33
+ - !!python/object:gr00t.data.types.ActionConfig
34
+ format: *id001
35
+ rep: *id002
36
+ state_key: null
37
+ type: *id003
38
+ - !!python/object:gr00t.data.types.ActionConfig
39
+ format: *id001
40
+ rep: *id002
41
+ state_key: null
42
+ type: *id003
43
+ - !!python/object:gr00t.data.types.ActionConfig
44
+ format: *id001
45
+ rep: *id002
46
+ state_key: null
47
+ type: *id003
48
+ delta_indices:
49
+ - 0
50
+ - 1
51
+ - 2
52
+ - 3
53
+ - 4
54
+ - 5
55
+ - 6
56
+ - 7
57
+ - 8
58
+ - 9
59
+ - 10
60
+ - 11
61
+ - 12
62
+ - 13
63
+ - 14
64
+ - 15
65
+ - 16
66
+ - 17
67
+ - 18
68
+ - 19
69
+ - 20
70
+ - 21
71
+ - 22
72
+ - 23
73
+ - 24
74
+ - 25
75
+ - 26
76
+ - 27
77
+ - 28
78
+ - 29
79
+ - 30
80
+ - 31
81
+ mean_std_embedding_keys: null
82
+ modality_keys:
83
+ - left_arm
84
+ - right_arm
85
+ - left_hand
86
+ - right_hand
87
+ sin_cos_embedding_keys: null
88
+ language: !!python/object:gr00t.data.types.ModalityConfig
89
+ action_configs: null
90
+ delta_indices:
91
+ - 0
92
+ mean_std_embedding_keys: null
93
+ modality_keys:
94
+ - annotation.human.task_description
95
+ sin_cos_embedding_keys: null
96
+ state: !!python/object:gr00t.data.types.ModalityConfig
97
+ action_configs: null
98
+ delta_indices:
99
+ - 0
100
+ mean_std_embedding_keys: null
101
+ modality_keys:
102
+ - left_arm
103
+ - right_arm
104
+ - left_hand
105
+ - right_hand
106
+ sin_cos_embedding_keys: null
107
+ video: !!python/object:gr00t.data.types.ModalityConfig
108
+ action_configs: null
109
+ delta_indices:
110
+ - 0
111
+ mean_std_embedding_keys: null
112
+ modality_keys:
113
+ - cam_ego_view
114
+ - cam_left_wrist
115
+ - cam_right_wrist
116
+ sin_cos_embedding_keys: null
117
+ mode: single_turn
118
+ multiprocessing_context: fork
119
+ num_shards_per_epoch: 100000
120
+ override_pretraining_statistics: true
121
+ random_chop: 0.0
122
+ seed: 42
123
+ shard_size: 1024
124
+ shuffle: true
125
+ subsample_ratio: 1.0
126
+ video_backend: torchcodec
127
+ load_config_path: null
128
+ model: !!python/object:gr00t.configs.model.gr00t_n1d7.Gr00tN1d7Config
129
+ _attn_implementation_internal: null
130
+ _commit_hash: null
131
+ _name_or_path: ''
132
+ _output_attentions: false
133
+ add_cross_attention: false
134
+ architectures: null
135
+ backbone_trainable_params_fp32: true
136
+ bad_words_ids: null
137
+ begin_suppress_tokens: null
138
+ bos_token_id: null
139
+ chunk_size_feed_forward: 0
140
+ color_jitter_params:
141
+ brightness: 0.3
142
+ contrast: 0.4
143
+ hue: 0.08
144
+ saturation: 0.5
145
+ cross_attention_hidden_size: null
146
+ decoder_start_token_id: null
147
+ diffusion_model_cfg:
148
+ attention_head_dim: 48
149
+ dropout: 0.2
150
+ final_dropout: true
151
+ interleave_self_attention: true
152
+ norm_type: ada_norm
153
+ num_attention_heads: 32
154
+ num_layers: 16
155
+ output_dim: 1024
156
+ positional_embeddings: null
157
+ diversity_penalty: 0.0
158
+ do_sample: false
159
+ dtype: null
160
+ early_stopping: false
161
+ encoder_no_repeat_ngram_size: 0
162
+ eos_token_id: null
163
+ exponential_decay_length_penalty: null
164
+ extra_augmentation_config: null
165
+ finetuning_task: null
166
+ forced_bos_token_id: null
167
+ forced_eos_token_id: null
168
+ id2label:
169
+ 0: LABEL_0
170
+ 1: LABEL_1
171
+ is_decoder: false
172
+ is_encoder_decoder: false
173
+ label2id:
174
+ LABEL_0: 0
175
+ LABEL_1: 1
176
+ length_penalty: 1.0
177
+ load_bf16: false
178
+ max_length: 20
179
+ min_length: 0
180
+ model_name: nvidia/Cosmos-Reason2-2B
181
+ no_repeat_ngram_size: 0
182
+ num_beam_groups: 1
183
+ num_beams: 1
184
+ num_return_sequences: 1
185
+ output_hidden_states: false
186
+ output_scores: false
187
+ pad_token_id: null
188
+ prefix: null
189
+ problem_type: null
190
+ pruned_heads: {}
191
+ random_rotation_angle: null
192
+ remove_invalid_values: false
193
+ repetition_penalty: 1.0
194
+ reproject_vision: false
195
+ return_dict: true
196
+ return_dict_in_generate: false
197
+ sep_token_id: null
198
+ state_dropout_prob: 0.2
199
+ suppress_tokens: null
200
+ task_specific_params: null
201
+ temperature: 1.0
202
+ tf_legacy_loss: false
203
+ tie_encoder_decoder: false
204
+ tie_word_embeddings: true
205
+ tokenizer_class: null
206
+ top_k: 50
207
+ top_p: 1.0
208
+ torchscript: false
209
+ transformers_version: null
210
+ tune_diffusion_model: true
211
+ tune_llm: false
212
+ tune_projector: true
213
+ tune_visual: false
214
+ typical_p: 1.0
215
+ use_bfloat16: false
216
+ use_relative_action: true
217
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
218
+ add_rl_callback: false
219
+ assert_loss_less_than: null
220
+ batch_size: null
221
+ bf16: true
222
+ dataloader_num_workers: 8
223
+ ddp_bucket_cap_mb: 100
224
+ deepspeed_stage: 2
225
+ enable_open_loop_eval: false
226
+ enable_profiling: false
227
+ eval_batch_size: 2
228
+ eval_bf16: true
229
+ eval_set_split_ratio: 0.1
230
+ eval_steps: 500
231
+ eval_strategy: 'no'
232
+ experiment_name: null
233
+ fp16: false
234
+ global_batch_size: 128
235
+ gradient_accumulation_steps: 1
236
+ gradient_checkpointing: false
237
+ learning_rate: 0.0001
238
+ logging_steps: 10
239
+ lr_scheduler_type: cosine
240
+ max_concurrent_uploads: 2
241
+ max_grad_norm: 1.0
242
+ max_retries: 3
243
+ max_steps: 10000
244
+ num_gpus: 1
245
+ open_loop_eval_plot_indices: null
246
+ open_loop_eval_steps_per_traj: 100
247
+ open_loop_eval_traj_ids:
248
+ - 0
249
+ optim: adamw_torch
250
+ output_dir: /home/joshua/checkpoints/pick_cube_n17_abs
251
+ remove_unused_columns: false
252
+ save_best_eval_metric_greater_is_better: true
253
+ save_best_eval_metric_name: ''
254
+ save_only_model: false
255
+ save_steps: 1000
256
+ save_total_limit: 5
257
+ save_vl_model: false
258
+ skip_weight_loading: false
259
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
260
+ tf32: true
261
+ transformers_access_token: null
262
+ transformers_cache_dir: null
263
+ transformers_local_files_only: false
264
+ transformers_trust_remote_code: true
265
+ upload_checkpoints: false
266
+ upload_every: 1000
267
+ upload_last_n_checkpoints: 5
268
+ use_ddp: false
269
+ use_wandb: false
270
+ wandb_project: finetune-gr00t-n1d7
271
+ warmup_ratio: 0.05
272
+ warmup_steps: 0
273
+ weight_decay: 1.0e-05
experiment_cfg/dataset_statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
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
+ }
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 sha256:86eff2d1ef7c0c18bf005497e2b3b7c963275e10e1b161bebcb332a47a7e6ad4
3
+ size 4970792616
model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87b5b48948e546d968ebdf11eec8f847182b2ebceb55bd51bde570db582a24b2
3
+ size 2618758696
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9889d3cb90ebe9c689f819e8d9538b576d4873fc8fc979802c09118426db452f
3
+ size 12964594710
processor_config.json ADDED
@@ -0,0 +1,1156 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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_ego_view",
1019
+ "cam_left_wrist",
1020
+ "cam_right_wrist"
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
+ "left_arm",
1032
+ "right_arm",
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
+ 16,
1059
+ 17,
1060
+ 18,
1061
+ 19,
1062
+ 20,
1063
+ 21,
1064
+ 22,
1065
+ 23,
1066
+ 24,
1067
+ 25,
1068
+ 26,
1069
+ 27,
1070
+ 28,
1071
+ 29,
1072
+ 30,
1073
+ 31
1074
+ ],
1075
+ "modality_keys": [
1076
+ "left_arm",
1077
+ "right_arm",
1078
+ "left_hand",
1079
+ "right_hand"
1080
+ ],
1081
+ "sin_cos_embedding_keys": null,
1082
+ "mean_std_embedding_keys": null,
1083
+ "action_configs": [
1084
+ {
1085
+ "rep": "ABSOLUTE",
1086
+ "type": "NON_EEF",
1087
+ "format": "DEFAULT",
1088
+ "state_key": null
1089
+ },
1090
+ {
1091
+ "rep": "ABSOLUTE",
1092
+ "type": "NON_EEF",
1093
+ "format": "DEFAULT",
1094
+ "state_key": null
1095
+ },
1096
+ {
1097
+ "rep": "ABSOLUTE",
1098
+ "type": "NON_EEF",
1099
+ "format": "DEFAULT",
1100
+ "state_key": null
1101
+ },
1102
+ {
1103
+ "rep": "ABSOLUTE",
1104
+ "type": "NON_EEF",
1105
+ "format": "DEFAULT",
1106
+ "state_key": null
1107
+ }
1108
+ ]
1109
+ },
1110
+ "language": {
1111
+ "delta_indices": [
1112
+ 0
1113
+ ],
1114
+ "modality_keys": [
1115
+ "annotation.human.task_description"
1116
+ ],
1117
+ "sin_cos_embedding_keys": null,
1118
+ "mean_std_embedding_keys": null,
1119
+ "action_configs": null
1120
+ }
1121
+ }
1122
+ },
1123
+ "image_crop_size": [
1124
+ 230,
1125
+ 230
1126
+ ],
1127
+ "image_target_size": [
1128
+ 256,
1129
+ 256
1130
+ ],
1131
+ "use_albumentations": true,
1132
+ "random_rotation_angle": 0,
1133
+ "color_jitter_params": {
1134
+ "brightness": 0.3,
1135
+ "contrast": 0.4,
1136
+ "saturation": 0.5,
1137
+ "hue": 0.08
1138
+ },
1139
+ "shortest_image_edge": 256,
1140
+ "crop_fraction": 0.95,
1141
+ "letter_box_transform": false,
1142
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1143
+ "model_type": "qwen",
1144
+ "formalize_language": true,
1145
+ "max_state_dim": 132,
1146
+ "max_action_dim": 132,
1147
+ "max_action_horizon": 40,
1148
+ "use_percentiles": true,
1149
+ "use_mean_std": false,
1150
+ "clip_outliers": true,
1151
+ "apply_sincos_state_encoding": false,
1152
+ "use_relative_action": true,
1153
+ "exclude_state": false,
1154
+ "state_dropout_prob": 0.2
1155
+ }
1156
+ }
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2eaec8303240c5401f6e55ce923129eb141e47eeea61ac8724f003f665352857
3
+ size 14645
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d02be6d8bda4ea9c67040ed89f878acdc986bd4df3fbc60440a9d3eacca02d63
3
+ size 1465
statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b69bddb93020269d421ea945eec680d5da044566bf045b99ee9c5dd7e7514e9f
3
+ size 5841
wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d7", "run_id": "pick_cube_n17_abs"}