Hugging Face's logo

Whalswp
/
robocasa-kitchen

Safetensors
Model card Files
xet
 Community
Whalswp commited on
Commit
181d94c
·
verified ·
1 Parent(s): d773bcf

Add CSN MarginHuber 60K checkpoint

Browse files
Files changed (20) hide show
  1. .gitattributes +1 -0
  2. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/config.json +134 -0
  3. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/dataset_state.json +1 -0
  4. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/embodiment_id.json +57 -0
  5. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/conf.yaml +234 -0
  6. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/config.yaml +232 -0
  7. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/dataset_statistics.json +317 -0
  8. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/final_model_config.json +56 -0
  9. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/final_processor_config.json +0 -0
  10. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/model-00001-of-00003.safetensors +3 -0
  11. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/model-00002-of-00003.safetensors +3 -0
  12. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/model-00003-of-00003.safetensors +3 -0
  13. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/model.safetensors.index.json +0 -0
  14. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/processor_config.json +1148 -0
  15. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/rng_state.pth +3 -0
  16. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/scheduler.pt +3 -0
  17. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/statistics.json +0 -0
  18. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/trainer_state.json +3 -0
  19. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/training_args.bin +3 -0
  20. objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/wandb_config.json +1 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/config.json ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "csn_action_distance_temperature": 0.3,
20
+ "csn_action_mask_seed": 43,
21
+ "csn_action_target": "normalized_eef_flat_rms_v1",
22
+ "csn_action_weight": 0.5,
23
+ "csn_aux_seed": 10042,
24
+ "csn_contract_version": "csn_state_action_v2",
25
+ "csn_enabled": true,
26
+ "csn_mask_l1_reduction": "condition_mean",
27
+ "csn_mask_l1_weight": 0.0,
28
+ "csn_mask_learnable": false,
29
+ "csn_mask_location": "projector_output",
30
+ "csn_mask_num_conditions": 2,
31
+ "csn_mask_scale_anchor_weight": 0.0,
32
+ "csn_mask_type": "fixed_disjoint",
33
+ "csn_mode": "joint_csn",
34
+ "csn_objective_config": {
35
+ "action": {
36
+ "student_score_scale": 0.02
37
+ },
38
+ "huber_delta": 1.0,
39
+ "name": "margin_huber",
40
+ "pair_construction": "all_unordered",
41
+ "standardize_eps": 1e-06,
42
+ "state": {
43
+ "student_score_scale": 0.02
44
+ },
45
+ "student_score_scale": 0.02,
46
+ "tie_epsilon": 1e-06
47
+ },
48
+ "csn_objective_name": "margin_huber",
49
+ "csn_relation_action_dimension": 7,
50
+ "csn_relation_modality_dataset": "robocasa_kitchen",
51
+ "csn_relation_modality_schema": "clvla_relation_modality_v1",
52
+ "csn_relation_modality_sha256": "a8de1380600263e45b361e7038971f33bfa29990d673803616c768c4aa222e9d",
53
+ "csn_relation_state_dimension": 11,
54
+ "csn_similarity": "normalized_masked_dot_v1",
55
+ "csn_similarity_temperature": 0.02,
56
+ "csn_state_distance_temperature": 1.0,
57
+ "csn_state_mask_seed": 42,
58
+ "csn_state_statistics_source": "processor_dataset_q01_q99",
59
+ "csn_state_target": "n1d7_eef_q01q99_rot6d_v1",
60
+ "csn_state_target_dimension": 11,
61
+ "csn_state_weight": 0.5,
62
+ "csn_summary_position": "append",
63
+ "csn_train_seed": 42,
64
+ "csn_view_cutoff_mode": "one_full_view",
65
+ "diffusion_model_cfg": {
66
+ "attention_head_dim": 48,
67
+ "dropout": 0.2,
68
+ "final_dropout": true,
69
+ "interleave_self_attention": true,
70
+ "norm_type": "ada_norm",
71
+ "num_attention_heads": 32,
72
+ "num_layers": 32,
73
+ "output_dim": 1024,
74
+ "positional_embeddings": null
75
+ },
76
+ "dtype": "float32",
77
+ "exclude_state": false,
78
+ "formalize_language": true,
79
+ "hidden_size": 1024,
80
+ "image_crop_size": [
81
+ 230,
82
+ 230
83
+ ],
84
+ "image_target_size": [
85
+ 256,
86
+ 256
87
+ ],
88
+ "letter_box_transform": false,
89
+ "load_bf16": false,
90
+ "max_action_dim": 132,
91
+ "max_num_embodiments": 32,
92
+ "max_seq_len": 1024,
93
+ "max_state_dim": 132,
94
+ "model_dtype": "bfloat16",
95
+ "model_name": "nvidia/Cosmos-Reason2-2B",
96
+ "model_type": "Gr00tN1d7",
97
+ "noise_beta_alpha": 1.5,
98
+ "noise_beta_beta": 1.0,
99
+ "noise_s": 0.999,
100
+ "num_inference_timesteps": 4,
101
+ "num_timestep_buckets": 1000,
102
+ "random_history_crop": true,
103
+ "random_rotation_angle": 0,
104
+ "reproject_vision": false,
105
+ "rtc_ramp_rate": 6.0,
106
+ "select_layer": 16,
107
+ "shortest_image_edge": 256,
108
+ "state_dropout_prob": 0.0,
109
+ "state_gaussian_noise_std": 0.0,
110
+ "transformers_version": "4.57.3",
111
+ "tune_diffusion_model": true,
112
+ "tune_linear": true,
113
+ "tune_llm": false,
114
+ "tune_projector": true,
115
+ "tune_top_llm_layers": 0,
116
+ "tune_visual": false,
117
+ "tune_vlln": true,
118
+ "use_albumentations": true,
119
+ "use_alternate_vl_dit": true,
120
+ "use_flash_attention": true,
121
+ "use_future_tokens": false,
122
+ "use_mean_std": false,
123
+ "use_percentiles": true,
124
+ "use_vl_self_attention": true,
125
+ "use_vlln": true,
126
+ "vl_self_attention_cfg": {
127
+ "attention_head_dim": 64,
128
+ "dropout": 0.2,
129
+ "final_dropout": true,
130
+ "num_attention_heads": 32,
131
+ "num_layers": 4,
132
+ "positional_embeddings": null
133
+ }
134
+ }
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/dataset_state.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"epoch": 0, "curr_shard_index": null}
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/embodiment_id.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "robocasa_panda_omron": 13,
3
+ "oxe_droid": 17,
4
+ "oxe_fractal": 18,
5
+ "oxe_language_table": 19,
6
+ "oxe_bridge": 20,
7
+ "unknown": 22,
8
+ "gr1_unified": 20,
9
+ "agibot": 26,
10
+ "sim_behavior_r1_pro": 23,
11
+ "xdof": 24,
12
+ "xdof_oss_data": 25,
13
+ "unitree_g1_full_body_with_waist_height_nav_cmd": 25,
14
+ "real_r1_pro_sharpa": 27,
15
+ "real_r1_pro_sharpa_add_view": 27,
16
+ "real_r1_pro_sharpa_relative_arm_joint": 26,
17
+ "real_r1_pro_sharpa_delta_eef": 26,
18
+ "real_r1_pro_sharpa_absolute_eef": 26,
19
+ "real_r1_pro_sharpa_meanstd": 26,
20
+ "real_r1_pro_sharpa_relative_eef": 26,
21
+ "real_r1_pro_sharpa_relative_eef_add_view": 26,
22
+ "real_r1_pro_sharpa_relative_eef_relative_hand": 26,
23
+ "real_r1_pro_sharpa_relative_eef_human": 26,
24
+ "real_r1_pro_sharpa_relative_eef_human_add_view": 26,
25
+ "real_r1_pro_sharpa_relative_eef_human_relative_hand": 26,
26
+ "real_r1_pro_sharpa_relative_eef_egodex": 26,
27
+ "real_r1_pro_sharpa_relative_eef_egodex_relative_hand": 26,
28
+ "real_r1_pro_sharpa_relative_eef_egodex_wrist_only": 26,
29
+ "real_r1_pro_sharpa_relative_eef_maxinsights": 26,
30
+ "real_r1_pro_sharpa_relative_eef_maxinsights_relative_hand": 26,
31
+ "real_r1_pro_sharpa_relative_eef_mecka": 26,
32
+ "real_r1_pro_sharpa_relative_eef_mecka_relative_hand": 26,
33
+ "real_g1_relative_eef_absolute_joints": 25,
34
+ "real_g1_relative_eef_absolute_joints_wrist_cam": 25,
35
+ "real_g1_relative_eef_relative_joints": 25,
36
+ "real_r1_pro_sharpa_relative_eef_relative_hand_relative_joint": 26,
37
+ "real_r1_pro_sharpa_relative_joint": 29,
38
+ "oxe_droid_relative_eef_relative_joint": 24,
39
+ "oxe_droid_relative_eef_relative_joint_swapped": 24,
40
+ "oxe_droid_relative_eef_relative_joint_upweight_z": 24,
41
+ "oxe_droid_relative_eef_relative_joint_upweight_z_swapped": 24,
42
+ "oxe_droid_relative_eef_relative_joint_3view": 24,
43
+ "oxe_droid_relative_eef_relative_joint_3view_swapped": 24,
44
+ "oxe_droid_relative_eef": 24,
45
+ "oxe_droid_joint_position_relative": 24,
46
+ "xdof_relative_eef_relative_joint": 27,
47
+ "xdof_relative_eef_relative_joint_subtask": 27,
48
+ "xdof_relative_eef": 27,
49
+ "xdof_relative_joint": 28,
50
+ "simpler_env_google": 0,
51
+ "simpler_env_widowx": 1,
52
+ "libero_sim": 2,
53
+ "droid_sim": 3,
54
+ "robocasa_gr1_tabletop": 10,
55
+ "new_embodiment": 10,
56
+ "unitree_g1_sonic": 11
57
+ }
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d7
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Cosmos-Reason2-2B
6
+ backbone_model_type: qwen
7
+ model_revision: null
8
+ tune_top_llm_layers: 0
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: false
12
+ select_layer: 12
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ backbone_trainable_params_fp32: true
17
+ image_crop_size:
18
+ - 230
19
+ - 230
20
+ image_target_size:
21
+ - 256
22
+ - 256
23
+ shortest_image_edge: null
24
+ crop_fraction: null
25
+ random_rotation_angle: null
26
+ color_jitter_params: null
27
+ use_albumentations_transforms: true
28
+ letter_box_transform: false
29
+ extra_augmentation_config: null
30
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_percentiles: true
33
+ use_relative_action: true
34
+ max_state_dim: 132
35
+ max_action_dim: 132
36
+ action_horizon: 40
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ state_history_length: 1
40
+ add_pos_embed: true
41
+ attn_dropout: 0.2
42
+ use_vlln: true
43
+ max_seq_len: 1024
44
+ use_alternate_vl_dit: true
45
+ attend_text_every_n_blocks: 2
46
+ diffusion_model_cfg:
47
+ positional_embeddings: null
48
+ num_layers: 16
49
+ num_attention_heads: 32
50
+ attention_head_dim: 48
51
+ norm_type: ada_norm
52
+ dropout: 0.2
53
+ final_dropout: true
54
+ output_dim: 1024
55
+ interleave_self_attention: true
56
+ num_inference_timesteps: 4
57
+ noise_beta_alpha: 1.5
58
+ noise_beta_beta: 1.0
59
+ noise_s: 0.999
60
+ num_timestep_buckets: 1000
61
+ tune_projector: true
62
+ tune_diffusion_model: true
63
+ tune_vlln: true
64
+ state_dropout_prob: 0.0
65
+ exclude_state: false
66
+ use_mean_std: false
67
+ max_num_embodiments: 32
68
+ data:
69
+ datasets:
70
+ - dataset_paths:
71
+ - /home/work/workspace/test/seonho/robocasa_mg_gr00t_300
72
+ embodiment_tag: new_embodiment
73
+ mix_ratio: 1.0
74
+ dataset_type: physical_embodiment
75
+ val_dataset_path: null
76
+ modality_configs:
77
+ new_embodiment:
78
+ video:
79
+ delta_indices:
80
+ - 0
81
+ modality_keys:
82
+ - left_view
83
+ - right_view
84
+ - wrist_view
85
+ sin_cos_embedding_keys: null
86
+ mean_std_embedding_keys: null
87
+ action_configs: null
88
+ state:
89
+ delta_indices:
90
+ - 0
91
+ modality_keys:
92
+ - end_effector_position_relative
93
+ - end_effector_rotation_relative
94
+ - gripper_qpos
95
+ - base_position
96
+ - base_rotation
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
+ - end_effector_position
120
+ - end_effector_rotation
121
+ - gripper_close
122
+ - base_motion
123
+ - control_mode
124
+ sin_cos_embedding_keys: null
125
+ mean_std_embedding_keys: null
126
+ action_configs:
127
+ - rep: DELTA
128
+ type: EEF
129
+ format: DEFAULT
130
+ state_key: null
131
+ - rep: DELTA
132
+ type: EEF
133
+ format: DEFAULT
134
+ state_key: null
135
+ - rep: ABSOLUTE
136
+ type: NON_EEF
137
+ format: DEFAULT
138
+ state_key: null
139
+ - rep: DELTA
140
+ type: NON_EEF
141
+ format: DEFAULT
142
+ state_key: null
143
+ - rep: ABSOLUTE
144
+ type: NON_EEF
145
+ format: DEFAULT
146
+ state_key: null
147
+ language:
148
+ delta_indices:
149
+ - 0
150
+ modality_keys:
151
+ - annotation.human.action.task_description
152
+ sin_cos_embedding_keys: null
153
+ mean_std_embedding_keys: null
154
+ action_configs: null
155
+ download_cache: false
156
+ shard_size: 1024
157
+ episode_sampling_rate: 0.1
158
+ num_shards_per_epoch: 100000
159
+ ds_weights_alpha: null
160
+ override_pretraining_statistics: true
161
+ mode: single_turn
162
+ random_chop: 0.0
163
+ mock_dataset_mode: false
164
+ shuffle: true
165
+ seed: 42
166
+ multiprocessing_context: fork
167
+ allow_padding: false
168
+ subsample_ratio: 1.0
169
+ image_crop_size:
170
+ - 244
171
+ - 244
172
+ image_target_size:
173
+ - 224
174
+ - 224
175
+ training:
176
+ output_dir: /home/work/workspace/test/seonho/groot_robocasa-kitchen/train_ckpt/Objective/CSN_MarginHuber_N1d7_Full_BS64_60K
177
+ experiment_name: CSN_MarginHuber_N1d7_Full_BS64_60K
178
+ max_steps: 60000
179
+ global_batch_size: 64
180
+ per_gpu_batch_size: null
181
+ gradient_accumulation_steps: 1
182
+ learning_rate: 0.0001
183
+ lr_scheduler_type: cosine
184
+ weight_decay: 1.0e-05
185
+ warmup_ratio: 0.05
186
+ warmup_steps: 0
187
+ max_grad_norm: 1.0
188
+ optim: adamw_torch
189
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
190
+ skip_weight_loading: false
191
+ tf32: true
192
+ fp16: false
193
+ bf16: true
194
+ eval_bf16: true
195
+ logging_steps: 10
196
+ save_steps: 5000
197
+ save_total_limit: 1
198
+ save_vl_model: false
199
+ save_only_model: false
200
+ resume_from_checkpoint: true
201
+ upload_checkpoints: false
202
+ upload_every: 1000
203
+ upload_last_n_checkpoints: 5
204
+ max_concurrent_uploads: 2
205
+ eval_strategy: 'no'
206
+ eval_steps: 500
207
+ eval_set_split_ratio: 0.1
208
+ eval_batch_size: 2
209
+ save_best_eval_metric_name: ''
210
+ save_best_eval_metric_greater_is_better: true
211
+ deepspeed_stage: 2
212
+ gradient_checkpointing: false
213
+ transformers_trust_remote_code: true
214
+ transformers_local_files_only: false
215
+ transformers_cache_dir: null
216
+ transformers_access_token: null
217
+ use_ddp: false
218
+ ddp_bucket_cap_mb: 100
219
+ num_gpus: 1
220
+ dataloader_num_workers: 8
221
+ remove_unused_columns: false
222
+ use_wandb: true
223
+ wandb_project: finetune-gr00t-n1d7
224
+ enable_profiling: false
225
+ max_retries: 3
226
+ assert_loss_less_than: null
227
+ add_rl_callback: false
228
+ enable_open_loop_eval: false
229
+ open_loop_eval_traj_ids:
230
+ - 0
231
+ open_loop_eval_steps_per_traj: 100
232
+ open_loop_eval_plot_indices: null
233
+ max_steps: 60000
234
+ save_steps: 5000
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/config.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d7
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Cosmos-Reason2-2B
6
+ backbone_model_type: qwen
7
+ model_revision: null
8
+ tune_top_llm_layers: 0
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: false
12
+ select_layer: 12
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ backbone_trainable_params_fp32: true
17
+ image_crop_size:
18
+ - 230
19
+ - 230
20
+ image_target_size:
21
+ - 256
22
+ - 256
23
+ shortest_image_edge: null
24
+ crop_fraction: null
25
+ random_rotation_angle: null
26
+ color_jitter_params: null
27
+ use_albumentations_transforms: true
28
+ letter_box_transform: false
29
+ extra_augmentation_config: null
30
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_percentiles: true
33
+ use_relative_action: true
34
+ max_state_dim: 132
35
+ max_action_dim: 132
36
+ action_horizon: 40
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ state_history_length: 1
40
+ add_pos_embed: true
41
+ attn_dropout: 0.2
42
+ use_vlln: true
43
+ max_seq_len: 1024
44
+ use_alternate_vl_dit: true
45
+ attend_text_every_n_blocks: 2
46
+ diffusion_model_cfg:
47
+ positional_embeddings: null
48
+ num_layers: 16
49
+ num_attention_heads: 32
50
+ attention_head_dim: 48
51
+ norm_type: ada_norm
52
+ dropout: 0.2
53
+ final_dropout: true
54
+ output_dim: 1024
55
+ interleave_self_attention: true
56
+ num_inference_timesteps: 4
57
+ noise_beta_alpha: 1.5
58
+ noise_beta_beta: 1.0
59
+ noise_s: 0.999
60
+ num_timestep_buckets: 1000
61
+ tune_projector: true
62
+ tune_diffusion_model: true
63
+ tune_vlln: true
64
+ state_dropout_prob: 0.0
65
+ exclude_state: false
66
+ use_mean_std: false
67
+ max_num_embodiments: 32
68
+ data:
69
+ datasets:
70
+ - dataset_paths:
71
+ - /home/work/workspace/test/seonho/robocasa_mg_gr00t_300
72
+ embodiment_tag: new_embodiment
73
+ mix_ratio: 1.0
74
+ dataset_type: physical_embodiment
75
+ val_dataset_path: null
76
+ modality_configs:
77
+ new_embodiment:
78
+ video:
79
+ delta_indices:
80
+ - 0
81
+ modality_keys:
82
+ - left_view
83
+ - right_view
84
+ - wrist_view
85
+ sin_cos_embedding_keys: null
86
+ mean_std_embedding_keys: null
87
+ action_configs: null
88
+ state:
89
+ delta_indices:
90
+ - 0
91
+ modality_keys:
92
+ - end_effector_position_relative
93
+ - end_effector_rotation_relative
94
+ - gripper_qpos
95
+ - base_position
96
+ - base_rotation
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
+ - end_effector_position
120
+ - end_effector_rotation
121
+ - gripper_close
122
+ - base_motion
123
+ - control_mode
124
+ sin_cos_embedding_keys: null
125
+ mean_std_embedding_keys: null
126
+ action_configs:
127
+ - rep: delta
128
+ type: eef
129
+ format: default
130
+ state_key: null
131
+ - rep: delta
132
+ type: eef
133
+ format: default
134
+ state_key: null
135
+ - rep: absolute
136
+ type: non_eef
137
+ format: default
138
+ state_key: null
139
+ - rep: delta
140
+ type: non_eef
141
+ format: default
142
+ state_key: null
143
+ - rep: absolute
144
+ type: non_eef
145
+ format: default
146
+ state_key: null
147
+ language:
148
+ delta_indices:
149
+ - 0
150
+ modality_keys:
151
+ - annotation.human.action.task_description
152
+ sin_cos_embedding_keys: null
153
+ mean_std_embedding_keys: null
154
+ action_configs: null
155
+ download_cache: false
156
+ shard_size: 1024
157
+ episode_sampling_rate: 0.1
158
+ num_shards_per_epoch: 100000
159
+ ds_weights_alpha: null
160
+ override_pretraining_statistics: true
161
+ mode: single_turn
162
+ random_chop: 0.0
163
+ mock_dataset_mode: false
164
+ shuffle: true
165
+ seed: 42
166
+ multiprocessing_context: fork
167
+ allow_padding: false
168
+ subsample_ratio: 1.0
169
+ image_crop_size:
170
+ - 244
171
+ - 244
172
+ image_target_size:
173
+ - 224
174
+ - 224
175
+ training:
176
+ output_dir: /home/work/workspace/test/seonho/groot_robocasa-kitchen/train_ckpt/Objective/CSN_MarginHuber_N1d7_Full_BS64_60K
177
+ experiment_name: CSN_MarginHuber_N1d7_Full_BS64_60K
178
+ max_steps: 60000
179
+ global_batch_size: 64
180
+ per_gpu_batch_size: null
181
+ gradient_accumulation_steps: 1
182
+ learning_rate: 0.0001
183
+ lr_scheduler_type: cosine
184
+ weight_decay: 1.0e-05
185
+ warmup_ratio: 0.05
186
+ warmup_steps: 0
187
+ max_grad_norm: 1.0
188
+ optim: adamw_torch
189
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
190
+ skip_weight_loading: false
191
+ tf32: true
192
+ fp16: false
193
+ bf16: true
194
+ eval_bf16: true
195
+ logging_steps: 10
196
+ save_steps: 5000
197
+ save_total_limit: 1
198
+ save_vl_model: false
199
+ save_only_model: false
200
+ resume_from_checkpoint: true
201
+ upload_checkpoints: false
202
+ upload_every: 1000
203
+ upload_last_n_checkpoints: 5
204
+ max_concurrent_uploads: 2
205
+ eval_strategy: 'no'
206
+ eval_steps: 500
207
+ eval_set_split_ratio: 0.1
208
+ eval_batch_size: 2
209
+ save_best_eval_metric_name: ''
210
+ save_best_eval_metric_greater_is_better: true
211
+ deepspeed_stage: 2
212
+ gradient_checkpointing: false
213
+ transformers_trust_remote_code: true
214
+ transformers_local_files_only: false
215
+ transformers_cache_dir: null
216
+ transformers_access_token: null
217
+ use_ddp: false
218
+ ddp_bucket_cap_mb: 100
219
+ num_gpus: 1
220
+ dataloader_num_workers: 8
221
+ remove_unused_columns: false
222
+ use_wandb: true
223
+ wandb_project: finetune-gr00t-n1d7
224
+ enable_profiling: false
225
+ max_retries: 3
226
+ assert_loss_less_than: null
227
+ add_rl_callback: false
228
+ enable_open_loop_eval: false
229
+ open_loop_eval_traj_ids:
230
+ - 0
231
+ open_loop_eval_steps_per_traj: 100
232
+ open_loop_eval_plot_indices: null
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,317 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "end_effector_position_relative": {
5
+ "min": [
6
+ -0.46594083309173584,
7
+ -0.8278615474700928,
8
+ 0.011948669329285622
9
+ ],
10
+ "max": [
11
+ 0.8734620809555054,
12
+ 0.8441421985626221,
13
+ 1.1348456144332886
14
+ ],
15
+ "mean": [
16
+ 0.266306072473526,
17
+ -0.04002884030342102,
18
+ 0.5537278652191162
19
+ ],
20
+ "std": [
21
+ 0.19139817357063293,
22
+ 0.2582354247570038,
23
+ 0.2461009919643402
24
+ ],
25
+ "q01": [
26
+ -0.251964715719223,
27
+ -0.6177875232696534,
28
+ 0.07541620314121246
29
+ ],
30
+ "q99": [
31
+ 0.736240336894989,
32
+ 0.6068667078018188,
33
+ 0.9772176861763
34
+ ]
35
+ },
36
+ "end_effector_rotation_relative": {
37
+ "min": [
38
+ -0.9999996423721313,
39
+ -0.9910850524902344,
40
+ -0.9571061134338379,
41
+ 6.351022108219695e-08
42
+ ],
43
+ "max": [
44
+ 0.9999980926513672,
45
+ 0.990567147731781,
46
+ 0.964046835899353,
47
+ 0.9740113019943237
48
+ ],
49
+ "mean": [
50
+ -0.14963096380233765,
51
+ 0.1248803362250328,
52
+ 0.0027880165725946426,
53
+ 0.2082664668560028
54
+ ],
55
+ "std": [
56
+ 0.7683019042015076,
57
+ 0.36296239495277405,
58
+ 0.37635329365730286,
59
+ 0.2322801947593689
60
+ ],
61
+ "q01": [
62
+ -0.9973158955574035,
63
+ -0.6837371301651001,
64
+ -0.7814695286750794,
65
+ 0.0013687612861394884
66
+ ],
67
+ "q99": [
68
+ 0.9957762026786804,
69
+ 0.8491807365417481,
70
+ 0.7687161755561829,
71
+ 0.802422981262207
72
+ ]
73
+ },
74
+ "gripper_qpos": {
75
+ "min": [
76
+ -0.027591193094849586,
77
+ -0.05319468677043915
78
+ ],
79
+ "max": [
80
+ 0.07896614819765091,
81
+ 0.10181591659784317
82
+ ],
83
+ "mean": [
84
+ 0.030854083597660065,
85
+ -0.03072655387222767
86
+ ],
87
+ "std": [
88
+ 0.013360796496272087,
89
+ 0.013391095213592052
90
+ ],
91
+ "q01": [
92
+ -4.0135159069905056e-05,
93
+ -0.04032127052545548
94
+ ],
95
+ "q99": [
96
+ 0.04066328033804893,
97
+ -0.00040660659200511887
98
+ ]
99
+ },
100
+ "base_position": {
101
+ "min": [
102
+ 0.31250813603401184,
103
+ -5.100343704223633,
104
+ 0.6963803172111511
105
+ ],
106
+ "max": [
107
+ 5.601607322692871,
108
+ -0.4872598946094513,
109
+ 0.7495372295379639
110
+ ],
111
+ "mean": [
112
+ 2.5315651893615723,
113
+ -1.5020122528076172,
114
+ 0.7073471546173096
115
+ ],
116
+ "std": [
117
+ 1.4670116901397705,
118
+ 0.8559004664421082,
119
+ 0.006971539929508914
120
+ ],
121
+ "q01": [
122
+ 0.5000457167625427,
123
+ -4.100119590759277,
124
+ 0.6999446749687195
125
+ ],
126
+ "q99": [
127
+ 5.5999836921691895,
128
+ -0.778001925945282,
129
+ 0.7071865797042847
130
+ ]
131
+ },
132
+ "base_rotation": {
133
+ "min": [
134
+ 0.0,
135
+ 0.0,
136
+ -1.0,
137
+ 0.0
138
+ ],
139
+ "max": [
140
+ 0.0,
141
+ 0.0,
142
+ 1.0,
143
+ 1.0
144
+ ],
145
+ "mean": [
146
+ 0.0,
147
+ 0.0,
148
+ 0.2765239179134369,
149
+ 0.6414101719856262
150
+ ],
151
+ "std": [
152
+ 0.0,
153
+ 0.0,
154
+ 0.6192871332168579,
155
+ 0.35383468866348267
156
+ ],
157
+ "q01": [
158
+ 0.0,
159
+ 0.0,
160
+ -1.0,
161
+ 5.569559434661642e-06
162
+ ],
163
+ "q99": [
164
+ 0.0,
165
+ 0.0,
166
+ 1.0,
167
+ 1.0
168
+ ]
169
+ }
170
+ },
171
+ "action": {
172
+ "end_effector_position": {
173
+ "min": [
174
+ -1.0,
175
+ -1.0,
176
+ -1.0
177
+ ],
178
+ "max": [
179
+ 1.0,
180
+ 1.0,
181
+ 1.0
182
+ ],
183
+ "mean": [
184
+ 0.029150759801268578,
185
+ 0.007607606239616871,
186
+ -0.037477411329746246
187
+ ],
188
+ "std": [
189
+ 0.4175023138523102,
190
+ 0.41343387961387634,
191
+ 0.36864909529685974
192
+ ],
193
+ "q01": [
194
+ -1.0,
195
+ -1.0,
196
+ -1.0
197
+ ],
198
+ "q99": [
199
+ 1.0,
200
+ 1.0,
201
+ 1.0
202
+ ]
203
+ },
204
+ "end_effector_rotation": {
205
+ "min": [
206
+ -1.0,
207
+ -1.0,
208
+ -1.0
209
+ ],
210
+ "max": [
211
+ 1.0,
212
+ 1.0,
213
+ 1.0
214
+ ],
215
+ "mean": [
216
+ 0.007235987111926079,
217
+ -0.023321982473134995,
218
+ -0.0024119403678923845
219
+ ],
220
+ "std": [
221
+ 0.147710919380188,
222
+ 0.15399158000946045,
223
+ 0.13929523527622223
224
+ ],
225
+ "q01": [
226
+ -0.42321433901786804,
227
+ -0.5219977021217346,
228
+ -0.4101182746887207
229
+ ],
230
+ "q99": [
231
+ 0.47503985762596074,
232
+ 0.38744416117668146,
233
+ 0.4242807936668394
234
+ ]
235
+ },
236
+ "gripper_close": {
237
+ "min": [
238
+ 0.0
239
+ ],
240
+ "max": [
241
+ 1.0
242
+ ],
243
+ "mean": [
244
+ 0.34521719813346863
245
+ ],
246
+ "std": [
247
+ 0.47949448227882385
248
+ ],
249
+ "q01": [
250
+ 0.0
251
+ ],
252
+ "q99": [
253
+ 1.0
254
+ ]
255
+ },
256
+ "base_motion": {
257
+ "min": [
258
+ 0.0,
259
+ 0.0,
260
+ 0.0,
261
+ 0.0
262
+ ],
263
+ "max": [
264
+ 0.0,
265
+ 0.0,
266
+ 0.0,
267
+ 0.0
268
+ ],
269
+ "mean": [
270
+ 0.0,
271
+ 0.0,
272
+ 0.0,
273
+ 0.0
274
+ ],
275
+ "std": [
276
+ 0.0,
277
+ 0.0,
278
+ 0.0,
279
+ 0.0
280
+ ],
281
+ "q01": [
282
+ 0.0,
283
+ 0.0,
284
+ 0.0,
285
+ 0.0
286
+ ],
287
+ "q99": [
288
+ 0.0,
289
+ 0.0,
290
+ 0.0,
291
+ 0.0
292
+ ]
293
+ },
294
+ "control_mode": {
295
+ "min": [
296
+ 0.0
297
+ ],
298
+ "max": [
299
+ 0.0
300
+ ],
301
+ "mean": [
302
+ 0.0
303
+ ],
304
+ "std": [
305
+ 0.0
306
+ ],
307
+ "q01": [
308
+ 0.0
309
+ ],
310
+ "q99": [
311
+ 0.0
312
+ ]
313
+ }
314
+ },
315
+ "relative_action": {}
316
+ }
317
+ }
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "Gr00tN1d7",
3
+ "model_dtype": "bfloat16",
4
+ "model_name": "nvidia/Cosmos-Reason2-2B",
5
+ "backbone_model_type": "qwen",
6
+ "model_revision": null,
7
+ "tune_top_llm_layers": 0,
8
+ "backbone_embedding_dim": 2048,
9
+ "tune_llm": false,
10
+ "tune_visual": false,
11
+ "select_layer": 16,
12
+ "reproject_vision": false,
13
+ "use_flash_attention": true,
14
+ "load_bf16": false,
15
+ "backbone_trainable_params_fp32": true,
16
+ "letter_box_transform": false,
17
+ "extra_augmentation_config": null,
18
+ "apply_sincos_state_encoding": false,
19
+ "use_percentiles": true,
20
+ "use_relative_action": false,
21
+ "max_state_dim": 132,
22
+ "max_action_dim": 132,
23
+ "action_horizon": 40,
24
+ "hidden_size": 1024,
25
+ "input_embedding_dim": 1536,
26
+ "state_history_length": 1,
27
+ "add_pos_embed": true,
28
+ "attn_dropout": 0.2,
29
+ "use_vlln": true,
30
+ "max_seq_len": 1024,
31
+ "use_alternate_vl_dit": true,
32
+ "attend_text_every_n_blocks": 2,
33
+ "diffusion_model_cfg": {
34
+ "attention_head_dim": 48,
35
+ "dropout": 0.2,
36
+ "final_dropout": true,
37
+ "interleave_self_attention": true,
38
+ "norm_type": "ada_norm",
39
+ "num_attention_heads": 32,
40
+ "num_layers": 32,
41
+ "output_dim": 1024,
42
+ "positional_embeddings": null
43
+ },
44
+ "num_inference_timesteps": 4,
45
+ "noise_beta_alpha": 1.5,
46
+ "noise_beta_beta": 1.0,
47
+ "noise_s": 0.999,
48
+ "num_timestep_buckets": 1000,
49
+ "tune_projector": true,
50
+ "tune_diffusion_model": true,
51
+ "tune_vlln": true,
52
+ "state_dropout_prob": 0.0,
53
+ "exclude_state": false,
54
+ "use_mean_std": false,
55
+ "max_num_embodiments": 32
56
+ }
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c45c85cffcddfcd05c963991f38b6a40df08153dc1126264313e806271e97431
3
+ size 4986650792
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ece06ce7e2c262e23a48c2a2ae23d22e522b4b49273ac28e0073ace3cae09286
3
+ size 4970800912
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6c46d8a0991692029711923a9cb2efbd3a33ca50ba22af68838d0a7f5023911
3
+ size 2636593584
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/processor_config.json ADDED
@@ -0,0 +1,1148 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "left_view",
1019
+ "right_view",
1020
+ "wrist_view"
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
+ "end_effector_position_relative",
1032
+ "end_effector_rotation_relative",
1033
+ "gripper_qpos",
1034
+ "base_position",
1035
+ "base_rotation"
1036
+ ],
1037
+ "sin_cos_embedding_keys": null,
1038
+ "mean_std_embedding_keys": null,
1039
+ "action_configs": null
1040
+ },
1041
+ "action": {
1042
+ "delta_indices": [
1043
+ 0,
1044
+ 1,
1045
+ 2,
1046
+ 3,
1047
+ 4,
1048
+ 5,
1049
+ 6,
1050
+ 7,
1051
+ 8,
1052
+ 9,
1053
+ 10,
1054
+ 11,
1055
+ 12,
1056
+ 13,
1057
+ 14,
1058
+ 15
1059
+ ],
1060
+ "modality_keys": [
1061
+ "end_effector_position",
1062
+ "end_effector_rotation",
1063
+ "gripper_close",
1064
+ "base_motion",
1065
+ "control_mode"
1066
+ ],
1067
+ "sin_cos_embedding_keys": null,
1068
+ "mean_std_embedding_keys": null,
1069
+ "action_configs": [
1070
+ {
1071
+ "rep": "DELTA",
1072
+ "type": "EEF",
1073
+ "format": "DEFAULT",
1074
+ "state_key": null
1075
+ },
1076
+ {
1077
+ "rep": "DELTA",
1078
+ "type": "EEF",
1079
+ "format": "DEFAULT",
1080
+ "state_key": null
1081
+ },
1082
+ {
1083
+ "rep": "ABSOLUTE",
1084
+ "type": "NON_EEF",
1085
+ "format": "DEFAULT",
1086
+ "state_key": null
1087
+ },
1088
+ {
1089
+ "rep": "DELTA",
1090
+ "type": "NON_EEF",
1091
+ "format": "DEFAULT",
1092
+ "state_key": null
1093
+ },
1094
+ {
1095
+ "rep": "ABSOLUTE",
1096
+ "type": "NON_EEF",
1097
+ "format": "DEFAULT",
1098
+ "state_key": null
1099
+ }
1100
+ ]
1101
+ },
1102
+ "language": {
1103
+ "delta_indices": [
1104
+ 0
1105
+ ],
1106
+ "modality_keys": [
1107
+ "annotation.human.action.task_description"
1108
+ ],
1109
+ "sin_cos_embedding_keys": null,
1110
+ "mean_std_embedding_keys": null,
1111
+ "action_configs": null
1112
+ }
1113
+ }
1114
+ },
1115
+ "image_crop_size": [
1116
+ 230,
1117
+ 230
1118
+ ],
1119
+ "image_target_size": [
1120
+ 256,
1121
+ 256
1122
+ ],
1123
+ "use_albumentations": true,
1124
+ "random_rotation_angle": 0,
1125
+ "color_jitter_params": {
1126
+ "brightness": 0.3,
1127
+ "contrast": 0.4,
1128
+ "saturation": 0.5,
1129
+ "hue": 0.08
1130
+ },
1131
+ "shortest_image_edge": 256,
1132
+ "crop_fraction": 0.95,
1133
+ "letter_box_transform": false,
1134
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1135
+ "model_type": "qwen",
1136
+ "formalize_language": true,
1137
+ "max_state_dim": 132,
1138
+ "max_action_dim": 132,
1139
+ "max_action_horizon": 40,
1140
+ "use_percentiles": true,
1141
+ "use_mean_std": false,
1142
+ "clip_outliers": true,
1143
+ "apply_sincos_state_encoding": false,
1144
+ "use_relative_action": true,
1145
+ "exclude_state": false,
1146
+ "state_dropout_prob": 0.0
1147
+ }
1148
+ }
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:312d73e50be6077a42db8a6ae3a82ff1446c73daf256bebff0a2308ab56ab5fe
3
+ size 14645
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:35adbb86eec8dc428b8196503e24b287361532643b2026cd76063e7dda8fef26
3
+ size 1465
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef3d6fae6c7666bdeb3b8dbe756e1cf6cd3bcb5d8d367f20d3cd2f60be35cd09
3
+ size 12052813
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5322fec277869ed49dafd36ba33c44be0e4b5d5b1e46b86bacc16ddbdbec2727
3
+ size 6097
objective/CSN_MarginHuber_N1d7_Full_BS64_60K/checkpoint-60000/wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d7", "run_id": "CSN_MarginHuber_N1d7_Full_BS64_60K"}