Pamela153 commited on
Commit
6eba116
·
verified ·
1 Parent(s): 9ad1cbe

Upload folder using huggingface_hub

Browse files
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,304 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d6
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Eagle-Block2A-2B-v2
6
+ backbone_model_type: eagle
7
+ model_revision: null
8
+ tune_top_llm_layers: 4
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: false
12
+ select_layer: 16
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ collator_overwrite_image_inputs: false
17
+ eagle_collator: true
18
+ backbone_trainable_params_fp32: true
19
+ image_crop_size: null
20
+ image_target_size: null
21
+ shortest_image_edge: 256
22
+ crop_fraction: 0.95
23
+ random_rotation_angle: null
24
+ color_jitter_params: null
25
+ use_albumentations_transforms: true
26
+ formalize_language: true
27
+ apply_sincos_state_encoding: false
28
+ use_relative_action: true
29
+ max_state_dim: 29
30
+ max_action_dim: 29
31
+ action_horizon: 16
32
+ hidden_size: 1024
33
+ input_embedding_dim: 1536
34
+ add_pos_embed: true
35
+ attn_dropout: 0.2
36
+ use_vlln: true
37
+ max_seq_len: 1024
38
+ use_alternate_vl_dit: true
39
+ attend_text_every_n_blocks: 2
40
+ diffusion_model_cfg:
41
+ positional_embeddings: null
42
+ num_layers: 32
43
+ num_attention_heads: 32
44
+ attention_head_dim: 48
45
+ norm_type: ada_norm
46
+ dropout: 0.2
47
+ final_dropout: true
48
+ output_dim: 1024
49
+ interleave_self_attention: true
50
+ num_inference_timesteps: 4
51
+ noise_beta_alpha: 1.5
52
+ noise_beta_beta: 1.0
53
+ noise_s: 0.999
54
+ num_timestep_buckets: 1000
55
+ tune_projector: true
56
+ tune_diffusion_model: true
57
+ tune_vlln: true
58
+ state_dropout_prob: 0.0
59
+ state_additive_noise_scale: 0.0
60
+ max_num_embodiments: 32
61
+ data:
62
+ datasets:
63
+ - dataset_paths:
64
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/AdjustToasterOvenTemperature/20250820/lerobot
65
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/AdjustWaterTemperature/20250820/lerobot
66
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CheesyBread/20250714/lerobot
67
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseBlenderLid/20250822/lerobot
68
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseCabinet/20250819/lerobot
69
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseDishwasher/20250820/lerobot
70
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseDrawer/20250819/lerobot
71
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseElectricKettleLid/20250820/lerobot
72
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseFridge/20250819/lerobot
73
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseFridgeDrawer/20250821/lerobot
74
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseMicrowave/20250819/lerobot
75
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseOven/20250820/lerobot
76
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseStandMixerHead/20250820/lerobot
77
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseToasterOvenDoor/20250820/lerobot
78
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CoffeeServeMug/20250819/lerobot
79
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CoffeeSetupMug/20250819/lerobot
80
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/LowerHeat/20250805/lerobot
81
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/MakeIcedCoffee/20250801/lerobot
82
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenBlenderLid/20250822/lerobot
83
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenCabinet/20250819/lerobot
84
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenDishwasher/20250820/lerobot
85
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenDrawer/20250819/lerobot
86
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenElectricKettleLid/20250820/lerobot
87
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenFridge/20250819/lerobot
88
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenFridgeDrawer/20250821/lerobot
89
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenMicrowave/20250819/lerobot
90
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenOven/20250820/lerobot
91
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenStandMixerHead/20250820/lerobot
92
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenToasterOvenDoor/20250820/lerobot
93
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PackDessert/20250806/lerobot
94
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCabinetToCounter/20250819/lerobot
95
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToBlender/20250822/lerobot
96
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToCabinet/20250819/lerobot
97
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToDrawer/20250821/lerobot
98
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToMicrowave/20250819/lerobot
99
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToOven/20250819/lerobot
100
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToSink/20250819/lerobot
101
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToStandMixer/20250820/lerobot
102
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToStove/20250819/lerobot
103
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToToasterOven/20250819/lerobot
104
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceDrawerToCounter/20250820/lerobot
105
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceFridgeDrawerToShelf/20250821/lerobot
106
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceFridgeShelfToDrawer/20250821/lerobot
107
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceMicrowaveToCounter/20250819/lerobot
108
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceSinkToCounter/20250819/lerobot
109
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceStoveToCounter/20250819/lerobot
110
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceToasterOvenToCounter/20250819/lerobot
111
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceToasterToCounter/20250819/lerobot
112
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PreheatOven/20250903/lerobot
113
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/SlideDishwasherRack/20250820/lerobot
114
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/SlideOvenRack/20250820/lerobot
115
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/SlideToasterOvenRack/20250820/lerobot
116
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/StartCoffeeMachine/20250819/lerobot
117
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOffMicrowave/20250819/lerobot
118
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOffSinkFaucet/20250819/lerobot
119
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOffStove/20250819/lerobot
120
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnBlender/20250822/lerobot
121
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnElectricKettle/20250820/lerobot
122
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnMicrowave/20250819/lerobot
123
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnSinkFaucet/20250819/lerobot
124
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnStove/20250819/lerobot
125
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnToaster/20250820/lerobot
126
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnToasterOven/20250820/lerobot
127
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnSinkSpout/20250820/lerobot
128
+ - /root/robocasa/datasets/v1.0_train/target/atomic/CloseBlenderLid/20250822/lerobot
129
+ - /root/robocasa/datasets/v1.0_train/target/atomic/CloseFridge/20250816/lerobot
130
+ - /root/robocasa/datasets/v1.0_train/target/atomic/CloseToasterOvenDoor/20250818/lerobot
131
+ - /root/robocasa/datasets/v1.0_train/target/atomic/CoffeeSetupMug/20250813/lerobot
132
+ - /root/robocasa/datasets/v1.0_train/target/atomic/OpenCabinet/20250813/lerobot
133
+ - /root/robocasa/datasets/v1.0_train/target/atomic/OpenDrawer/20250816/lerobot
134
+ - /root/robocasa/datasets/v1.0_train/target/atomic/OpenStandMixerHead/20250818/lerobot
135
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceCounterToCabinet/20250811/lerobot
136
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceCounterToStove/20250818/lerobot
137
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceDrawerToCounter/20250820/lerobot
138
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceSinkToCounter/20250813/lerobot
139
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceToasterToCounter/20250817/lerobot
140
+ - /root/robocasa/datasets/v1.0_train/target/atomic/SlideDishwasherRack/20250820/lerobot
141
+ - /root/robocasa/datasets/v1.0_train/target/atomic/TurnOffStove/20250812/lerobot
142
+ - /root/robocasa/datasets/v1.0_train/target/atomic/TurnOnElectricKettle/20250817/lerobot
143
+ - /root/robocasa/datasets/v1.0_train/target/atomic/TurnOnMicrowave/20250813/lerobot
144
+ - /root/robocasa/datasets/v1.0_train/target/atomic/TurnOnSinkFaucet/20250812/lerobot
145
+ embodiment_tag: robocasa_panda_omron
146
+ mix_ratio: 1.0
147
+ dataset_type: physical_embodiment
148
+ val_dataset_path: null
149
+ modality_configs:
150
+ robocasa_panda_omron:
151
+ video:
152
+ delta_indices:
153
+ - 0
154
+ modality_keys:
155
+ - res256_image_side_0
156
+ - res256_image_side_1
157
+ - res256_image_wrist_0
158
+ sin_cos_embedding_keys: null
159
+ mean_std_embedding_keys: null
160
+ action_configs: null
161
+ state:
162
+ delta_indices:
163
+ - 0
164
+ modality_keys:
165
+ - end_effector_position_relative
166
+ - end_effector_rotation_relative
167
+ - gripper_qpos
168
+ - base_position
169
+ - base_rotation
170
+ sin_cos_embedding_keys: null
171
+ mean_std_embedding_keys: null
172
+ action_configs: null
173
+ action:
174
+ delta_indices:
175
+ - 0
176
+ - 1
177
+ - 2
178
+ - 3
179
+ - 4
180
+ - 5
181
+ - 6
182
+ - 7
183
+ - 8
184
+ - 9
185
+ - 10
186
+ - 11
187
+ - 12
188
+ - 13
189
+ - 14
190
+ - 15
191
+ modality_keys:
192
+ - end_effector_position
193
+ - end_effector_rotation
194
+ - gripper_close
195
+ - base_motion
196
+ - control_mode
197
+ sin_cos_embedding_keys: null
198
+ mean_std_embedding_keys: null
199
+ action_configs:
200
+ - rep: ABSOLUTE
201
+ type: NON_EEF
202
+ format: DEFAULT
203
+ state_key: null
204
+ - rep: ABSOLUTE
205
+ type: NON_EEF
206
+ format: DEFAULT
207
+ state_key: null
208
+ - rep: ABSOLUTE
209
+ type: NON_EEF
210
+ format: DEFAULT
211
+ state_key: null
212
+ - rep: ABSOLUTE
213
+ type: NON_EEF
214
+ format: DEFAULT
215
+ state_key: null
216
+ - rep: ABSOLUTE
217
+ type: NON_EEF
218
+ format: DEFAULT
219
+ state_key: null
220
+ language:
221
+ delta_indices:
222
+ - 0
223
+ modality_keys:
224
+ - annotation.human.task_description
225
+ sin_cos_embedding_keys: null
226
+ mean_std_embedding_keys: null
227
+ action_configs: null
228
+ download_cache: false
229
+ shard_size: 1024
230
+ episode_sampling_rate: 0.1
231
+ num_shards_per_epoch: 100000
232
+ override_pretraining_statistics: false
233
+ mode: single_turn
234
+ random_chop: 0.0
235
+ mock_dataset_mode: false
236
+ shuffle: true
237
+ seed: 42
238
+ multiprocessing_context: fork
239
+ allow_padding: false
240
+ subsample_ratio: 1.0
241
+ image_crop_size:
242
+ - 244
243
+ - 244
244
+ image_target_size:
245
+ - 224
246
+ - 224
247
+ video_backend: torchcodec
248
+ training:
249
+ output_dir: /root/outputs/gr00t-robocasa-v1
250
+ experiment_name: null
251
+ max_steps: 50000
252
+ global_batch_size: 64
253
+ batch_size: null
254
+ gradient_accumulation_steps: 1
255
+ learning_rate: 0.0001
256
+ lr_scheduler_type: cosine
257
+ weight_decay: 1.0e-05
258
+ warmup_ratio: 0.05
259
+ warmup_steps: 0
260
+ max_grad_norm: 1.0
261
+ optim: adamw_torch
262
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
263
+ tf32: true
264
+ fp16: false
265
+ bf16: true
266
+ eval_bf16: true
267
+ logging_steps: 10
268
+ save_steps: 2500
269
+ save_total_limit: 5
270
+ save_vl_model: false
271
+ upload_checkpoints: false
272
+ upload_every: 1000
273
+ upload_last_n_checkpoints: 5
274
+ max_concurrent_uploads: 2
275
+ eval_strategy: 'no'
276
+ eval_steps: 500
277
+ eval_set_split_ratio: 0.1
278
+ eval_batch_size: 2
279
+ save_best_eval_metric_name: ''
280
+ save_best_eval_metric_greater_is_better: true
281
+ deepspeed_stage: 2
282
+ gradient_checkpointing: false
283
+ transformers_trust_remote_code: true
284
+ transformers_local_files_only: false
285
+ transformers_cache_dir: null
286
+ transformers_access_token: null
287
+ use_ddp: false
288
+ ddp_bucket_cap_mb: 100
289
+ num_gpus: 8
290
+ dataloader_num_workers: 4
291
+ remove_unused_columns: false
292
+ use_wandb: false
293
+ wandb_project: finetune-gr00t-n1d6
294
+ enable_profiling: false
295
+ max_retries: 3
296
+ assert_loss_less_than: null
297
+ add_rl_callback: false
298
+ enable_open_loop_eval: false
299
+ open_loop_eval_traj_ids:
300
+ - 0
301
+ open_loop_eval_steps_per_traj: 100
302
+ open_loop_eval_plot_indices: null
303
+ max_steps: 50000
304
+ save_steps: 2500
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,340 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ !!python/object:gr00t.configs.base_config.Config
2
+ data: !!python/object:gr00t.configs.data.data_config.DataConfig
3
+ allow_padding: false
4
+ datasets:
5
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
6
+ dataset_paths:
7
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/AdjustToasterOvenTemperature/20250820/lerobot
8
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/AdjustWaterTemperature/20250820/lerobot
9
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CheesyBread/20250714/lerobot
10
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseBlenderLid/20250822/lerobot
11
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseCabinet/20250819/lerobot
12
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseDishwasher/20250820/lerobot
13
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseDrawer/20250819/lerobot
14
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseElectricKettleLid/20250820/lerobot
15
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseFridge/20250819/lerobot
16
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseFridgeDrawer/20250821/lerobot
17
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseMicrowave/20250819/lerobot
18
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseOven/20250820/lerobot
19
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseStandMixerHead/20250820/lerobot
20
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CloseToasterOvenDoor/20250820/lerobot
21
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CoffeeServeMug/20250819/lerobot
22
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/CoffeeSetupMug/20250819/lerobot
23
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/LowerHeat/20250805/lerobot
24
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/MakeIcedCoffee/20250801/lerobot
25
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenBlenderLid/20250822/lerobot
26
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenCabinet/20250819/lerobot
27
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenDishwasher/20250820/lerobot
28
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenDrawer/20250819/lerobot
29
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenElectricKettleLid/20250820/lerobot
30
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenFridge/20250819/lerobot
31
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenFridgeDrawer/20250821/lerobot
32
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenMicrowave/20250819/lerobot
33
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenOven/20250820/lerobot
34
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenStandMixerHead/20250820/lerobot
35
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/OpenToasterOvenDoor/20250820/lerobot
36
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PackDessert/20250806/lerobot
37
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCabinetToCounter/20250819/lerobot
38
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToBlender/20250822/lerobot
39
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToCabinet/20250819/lerobot
40
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToDrawer/20250821/lerobot
41
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToMicrowave/20250819/lerobot
42
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToOven/20250819/lerobot
43
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToSink/20250819/lerobot
44
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToStandMixer/20250820/lerobot
45
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToStove/20250819/lerobot
46
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceCounterToToasterOven/20250819/lerobot
47
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceDrawerToCounter/20250820/lerobot
48
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceFridgeDrawerToShelf/20250821/lerobot
49
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceFridgeShelfToDrawer/20250821/lerobot
50
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceMicrowaveToCounter/20250819/lerobot
51
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceSinkToCounter/20250819/lerobot
52
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceStoveToCounter/20250819/lerobot
53
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceToasterOvenToCounter/20250819/lerobot
54
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PickPlaceToasterToCounter/20250819/lerobot
55
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/PreheatOven/20250903/lerobot
56
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/SlideDishwasherRack/20250820/lerobot
57
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/SlideOvenRack/20250820/lerobot
58
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/SlideToasterOvenRack/20250820/lerobot
59
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/StartCoffeeMachine/20250819/lerobot
60
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOffMicrowave/20250819/lerobot
61
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOffSinkFaucet/20250819/lerobot
62
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOffStove/20250819/lerobot
63
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnBlender/20250822/lerobot
64
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnElectricKettle/20250820/lerobot
65
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnMicrowave/20250819/lerobot
66
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnSinkFaucet/20250819/lerobot
67
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnStove/20250819/lerobot
68
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnToaster/20250820/lerobot
69
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnOnToasterOven/20250820/lerobot
70
+ - /root/robocasa/datasets/v1.0_train/pretrain/atomic/TurnSinkSpout/20250820/lerobot
71
+ - /root/robocasa/datasets/v1.0_train/target/atomic/CloseBlenderLid/20250822/lerobot
72
+ - /root/robocasa/datasets/v1.0_train/target/atomic/CloseFridge/20250816/lerobot
73
+ - /root/robocasa/datasets/v1.0_train/target/atomic/CloseToasterOvenDoor/20250818/lerobot
74
+ - /root/robocasa/datasets/v1.0_train/target/atomic/CoffeeSetupMug/20250813/lerobot
75
+ - /root/robocasa/datasets/v1.0_train/target/atomic/OpenCabinet/20250813/lerobot
76
+ - /root/robocasa/datasets/v1.0_train/target/atomic/OpenDrawer/20250816/lerobot
77
+ - /root/robocasa/datasets/v1.0_train/target/atomic/OpenStandMixerHead/20250818/lerobot
78
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceCounterToCabinet/20250811/lerobot
79
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceCounterToStove/20250818/lerobot
80
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceDrawerToCounter/20250820/lerobot
81
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceSinkToCounter/20250813/lerobot
82
+ - /root/robocasa/datasets/v1.0_train/target/atomic/PickPlaceToasterToCounter/20250817/lerobot
83
+ - /root/robocasa/datasets/v1.0_train/target/atomic/SlideDishwasherRack/20250820/lerobot
84
+ - /root/robocasa/datasets/v1.0_train/target/atomic/TurnOffStove/20250812/lerobot
85
+ - /root/robocasa/datasets/v1.0_train/target/atomic/TurnOnElectricKettle/20250817/lerobot
86
+ - /root/robocasa/datasets/v1.0_train/target/atomic/TurnOnMicrowave/20250813/lerobot
87
+ - /root/robocasa/datasets/v1.0_train/target/atomic/TurnOnSinkFaucet/20250812/lerobot
88
+ dataset_type: physical_embodiment
89
+ embodiment_tag: robocasa_panda_omron
90
+ mix_ratio: 1.0
91
+ val_dataset_path: null
92
+ download_cache: false
93
+ episode_sampling_rate: 0.1
94
+ image_crop_size:
95
+ - 244
96
+ - 244
97
+ image_target_size:
98
+ - 224
99
+ - 224
100
+ mock_dataset_mode: false
101
+ modality_configs:
102
+ robocasa_panda_omron:
103
+ action: !!python/object:gr00t.data.types.ModalityConfig
104
+ action_configs:
105
+ - !!python/object:gr00t.data.types.ActionConfig
106
+ format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
107
+ - default
108
+ rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
109
+ - absolute
110
+ state_key: null
111
+ type: &id003 !!python/object/apply:gr00t.data.types.ActionType
112
+ - non_eef
113
+ - !!python/object:gr00t.data.types.ActionConfig
114
+ format: *id001
115
+ rep: *id002
116
+ state_key: null
117
+ type: *id003
118
+ - !!python/object:gr00t.data.types.ActionConfig
119
+ format: *id001
120
+ rep: *id002
121
+ state_key: null
122
+ type: *id003
123
+ - !!python/object:gr00t.data.types.ActionConfig
124
+ format: *id001
125
+ rep: *id002
126
+ state_key: null
127
+ type: *id003
128
+ - !!python/object:gr00t.data.types.ActionConfig
129
+ format: *id001
130
+ rep: *id002
131
+ state_key: null
132
+ type: *id003
133
+ delta_indices:
134
+ - 0
135
+ - 1
136
+ - 2
137
+ - 3
138
+ - 4
139
+ - 5
140
+ - 6
141
+ - 7
142
+ - 8
143
+ - 9
144
+ - 10
145
+ - 11
146
+ - 12
147
+ - 13
148
+ - 14
149
+ - 15
150
+ mean_std_embedding_keys: null
151
+ modality_keys:
152
+ - end_effector_position
153
+ - end_effector_rotation
154
+ - gripper_close
155
+ - base_motion
156
+ - control_mode
157
+ sin_cos_embedding_keys: null
158
+ language: !!python/object:gr00t.data.types.ModalityConfig
159
+ action_configs: null
160
+ delta_indices:
161
+ - 0
162
+ mean_std_embedding_keys: null
163
+ modality_keys:
164
+ - annotation.human.task_description
165
+ sin_cos_embedding_keys: null
166
+ state: !!python/object:gr00t.data.types.ModalityConfig
167
+ action_configs: null
168
+ delta_indices:
169
+ - 0
170
+ mean_std_embedding_keys: null
171
+ modality_keys:
172
+ - end_effector_position_relative
173
+ - end_effector_rotation_relative
174
+ - gripper_qpos
175
+ - base_position
176
+ - base_rotation
177
+ sin_cos_embedding_keys: null
178
+ video: !!python/object:gr00t.data.types.ModalityConfig
179
+ action_configs: null
180
+ delta_indices:
181
+ - 0
182
+ mean_std_embedding_keys: null
183
+ modality_keys:
184
+ - res256_image_side_0
185
+ - res256_image_side_1
186
+ - res256_image_wrist_0
187
+ sin_cos_embedding_keys: null
188
+ mode: single_turn
189
+ multiprocessing_context: fork
190
+ num_shards_per_epoch: 100000
191
+ override_pretraining_statistics: false
192
+ random_chop: 0.0
193
+ seed: 42
194
+ shard_size: 1024
195
+ shuffle: true
196
+ subsample_ratio: 1.0
197
+ video_backend: torchcodec
198
+ load_config_path: null
199
+ model: !!python/object:gr00t.configs.model.gr00t_n1d6.Gr00tN1d6Config
200
+ _attn_implementation_autoset: false
201
+ _attn_implementation_internal: null
202
+ _commit_hash: null
203
+ _name_or_path: ''
204
+ add_cross_attention: false
205
+ architectures: null
206
+ backbone_model_type: eagle
207
+ backbone_trainable_params_fp32: true
208
+ bad_words_ids: null
209
+ begin_suppress_tokens: null
210
+ bos_token_id: null
211
+ chunk_size_feed_forward: 0
212
+ color_jitter_params: null
213
+ cross_attention_hidden_size: null
214
+ decoder_start_token_id: null
215
+ diffusion_model_cfg:
216
+ attention_head_dim: 48
217
+ dropout: 0.2
218
+ final_dropout: true
219
+ interleave_self_attention: true
220
+ norm_type: ada_norm
221
+ num_attention_heads: 32
222
+ num_layers: 32
223
+ output_dim: 1024
224
+ positional_embeddings: null
225
+ diversity_penalty: 0.0
226
+ do_sample: false
227
+ eagle_collator: true
228
+ early_stopping: false
229
+ encoder_no_repeat_ngram_size: 0
230
+ eos_token_id: null
231
+ exponential_decay_length_penalty: null
232
+ finetuning_task: null
233
+ forced_bos_token_id: null
234
+ forced_eos_token_id: null
235
+ id2label:
236
+ 0: LABEL_0
237
+ 1: LABEL_1
238
+ is_decoder: false
239
+ is_encoder_decoder: false
240
+ label2id:
241
+ LABEL_0: 0
242
+ LABEL_1: 1
243
+ length_penalty: 1.0
244
+ load_bf16: false
245
+ max_length: 20
246
+ min_length: 0
247
+ model_name: nvidia/Eagle-Block2A-2B-v2
248
+ no_repeat_ngram_size: 0
249
+ num_beam_groups: 1
250
+ num_beams: 1
251
+ num_return_sequences: 1
252
+ output_attentions: false
253
+ output_hidden_states: false
254
+ output_scores: false
255
+ pad_token_id: null
256
+ prefix: null
257
+ problem_type: null
258
+ pruned_heads: {}
259
+ random_rotation_angle: null
260
+ remove_invalid_values: false
261
+ repetition_penalty: 1.0
262
+ reproject_vision: false
263
+ return_dict: true
264
+ return_dict_in_generate: false
265
+ sep_token_id: null
266
+ state_dropout_prob: 0.0
267
+ suppress_tokens: null
268
+ task_specific_params: null
269
+ temperature: 1.0
270
+ tf_legacy_loss: false
271
+ tie_encoder_decoder: false
272
+ tie_word_embeddings: true
273
+ tokenizer_class: null
274
+ top_k: 50
275
+ top_p: 1.0
276
+ torch_dtype: null
277
+ torchscript: false
278
+ transformers_version: null
279
+ tune_diffusion_model: true
280
+ tune_llm: false
281
+ tune_projector: true
282
+ tune_visual: false
283
+ typical_p: 1.0
284
+ use_bfloat16: false
285
+ use_relative_action: true
286
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
287
+ add_rl_callback: false
288
+ assert_loss_less_than: null
289
+ batch_size: null
290
+ bf16: true
291
+ dataloader_num_workers: 4
292
+ ddp_bucket_cap_mb: 100
293
+ deepspeed_stage: 2
294
+ enable_open_loop_eval: false
295
+ enable_profiling: false
296
+ eval_batch_size: 2
297
+ eval_bf16: true
298
+ eval_set_split_ratio: 0.1
299
+ eval_steps: 500
300
+ eval_strategy: 'no'
301
+ experiment_name: null
302
+ fp16: false
303
+ global_batch_size: 64
304
+ gradient_accumulation_steps: 1
305
+ gradient_checkpointing: false
306
+ learning_rate: 0.0001
307
+ logging_steps: 10
308
+ lr_scheduler_type: cosine
309
+ max_concurrent_uploads: 2
310
+ max_grad_norm: 1.0
311
+ max_retries: 3
312
+ max_steps: 50000
313
+ num_gpus: 8
314
+ open_loop_eval_plot_indices: null
315
+ open_loop_eval_steps_per_traj: 100
316
+ open_loop_eval_traj_ids:
317
+ - 0
318
+ optim: adamw_torch
319
+ output_dir: /root/outputs/gr00t-robocasa-v1
320
+ remove_unused_columns: false
321
+ save_best_eval_metric_greater_is_better: true
322
+ save_best_eval_metric_name: ''
323
+ save_steps: 2500
324
+ save_total_limit: 5
325
+ save_vl_model: false
326
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
327
+ tf32: true
328
+ transformers_access_token: null
329
+ transformers_cache_dir: null
330
+ transformers_local_files_only: false
331
+ transformers_trust_remote_code: true
332
+ upload_checkpoints: false
333
+ upload_every: 1000
334
+ upload_last_n_checkpoints: 5
335
+ use_ddp: false
336
+ use_wandb: false
337
+ wandb_project: finetune-gr00t-n1d6
338
+ warmup_ratio: 0.05
339
+ warmup_steps: 0
340
+ weight_decay: 1.0e-05
experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,317 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "robocasa_panda_omron": {
3
+ "state": {
4
+ "end_effector_position_relative": {
5
+ "min": [
6
+ -0.6169384717941284,
7
+ -0.8689756989479065,
8
+ -0.3446122705936432
9
+ ],
10
+ "max": [
11
+ 0.9091058969497681,
12
+ 0.8569459915161133,
13
+ 1.0843316316604614
14
+ ],
15
+ "mean": [
16
+ 0.2651550115464023,
17
+ -0.02962475071638439,
18
+ 0.45191954730417777
19
+ ],
20
+ "std": [
21
+ 0.16571417800338054,
22
+ 0.254853038171651,
23
+ 0.21899660506818075
24
+ ],
25
+ "q01": [
26
+ -0.3761294186115265,
27
+ -0.7233268618583679,
28
+ -0.23106376826763153
29
+ ],
30
+ "q99": [
31
+ 0.8378565311431885,
32
+ 0.7546205520629883,
33
+ 0.9249030947685242
34
+ ]
35
+ },
36
+ "end_effector_rotation_relative": {
37
+ "min": [
38
+ -0.9999971985816956,
39
+ -0.9999884366989136,
40
+ -0.9975369572639465,
41
+ 6.754255821306288e-08
42
+ ],
43
+ "max": [
44
+ 0.9999998807907104,
45
+ 0.9999621510505676,
46
+ 0.9900747537612915,
47
+ 0.9735639095306396
48
+ ],
49
+ "mean": [
50
+ -0.2449173631670233,
51
+ 0.033624006536664273,
52
+ -0.0721874569598496,
53
+ 0.16001812256396578
54
+ ],
55
+ "std": [
56
+ 0.8057003219674179,
57
+ 0.3257059498633488,
58
+ 0.34975302043047835,
59
+ 0.17463791053257643
60
+ ],
61
+ "q01": [
62
+ -0.9995729327201843,
63
+ -0.9585485458374023,
64
+ -0.8866488337516785,
65
+ 0.0003100793983321637
66
+ ],
67
+ "q99": [
68
+ 0.9993855357170105,
69
+ 0.9764737486839294,
70
+ 0.8878775835037231,
71
+ 0.8871033191680908
72
+ ]
73
+ },
74
+ "gripper_qpos": {
75
+ "min": [
76
+ -0.029329104349017143,
77
+ -0.07029139250516891
78
+ ],
79
+ "max": [
80
+ 0.06876722723245621,
81
+ 0.026710085570812225
82
+ ],
83
+ "mean": [
84
+ 0.03142353526923579,
85
+ -0.031758853815757805
86
+ ],
87
+ "std": [
88
+ 0.012816575843103204,
89
+ 0.012470656645251686
90
+ ],
91
+ "q01": [
92
+ -0.0028626956045627594,
93
+ -0.04449209198355675
94
+ ],
95
+ "q99": [
96
+ 0.04488873854279518,
97
+ 0.0015663582598790526
98
+ ]
99
+ },
100
+ "base_position": {
101
+ "min": [
102
+ -4.821934223175049,
103
+ -6.890198230743408,
104
+ 0.6994456052780151
105
+ ],
106
+ "max": [
107
+ 7.778976917266846,
108
+ 0.4876587688922882,
109
+ 0.7259154915809631
110
+ ],
111
+ "mean": [
112
+ 2.583225218328043,
113
+ -1.7625188175481725,
114
+ 0.7007490723468245
115
+ ],
116
+ "std": [
117
+ 1.5287747127625415,
118
+ 1.139524533368529,
119
+ 0.0013712100676446942
120
+ ],
121
+ "q01": [
122
+ -4.570116996765137,
123
+ -6.843641757965088,
124
+ 0.6998441815376282
125
+ ],
126
+ "q99": [
127
+ 7.292758464813232,
128
+ 0.48759591579437256,
129
+ 0.714806854724884
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
+ 0.26989995834630387,
149
+ 0.621002213509319
150
+ ],
151
+ "std": [
152
+ 0.0,
153
+ 0.0,
154
+ 0.6453511712345816,
155
+ 0.35400514622404494
156
+ ],
157
+ "q01": [
158
+ 0.0,
159
+ 0.0,
160
+ -1.0,
161
+ 1.9421531760599464e-07
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
+ -0.0003490836577513803,
185
+ -0.014827997382293595,
186
+ -0.06616271975548409
187
+ ],
188
+ "std": [
189
+ 0.42983297822841915,
190
+ 0.41306588501905284,
191
+ 0.38557634504813654
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
+ 0.007756964983150791,
217
+ -0.02445770047077635,
218
+ -0.0014353462994126467
219
+ ],
220
+ "std": [
221
+ 0.11873000006390404,
222
+ 0.13205572265565643,
223
+ 0.12890852648938061
224
+ ],
225
+ "q01": [
226
+ -0.6857143044471741,
227
+ -0.9028571248054504,
228
+ -1.0
229
+ ],
230
+ "q99": [
231
+ 0.9028571248054504,
232
+ 0.8342857360839844,
233
+ 0.8718582391738892
234
+ ]
235
+ },
236
+ "gripper_close": {
237
+ "min": [
238
+ -1.0
239
+ ],
240
+ "max": [
241
+ 1.0
242
+ ],
243
+ "mean": [
244
+ -0.34592562958645734
245
+ ],
246
+ "std": [
247
+ 0.9382915564972965
248
+ ],
249
+ "q01": [
250
+ -1.0
251
+ ],
252
+ "q99": [
253
+ 1.0
254
+ ]
255
+ },
256
+ "base_motion": {
257
+ "min": [
258
+ -1.0,
259
+ -1.0,
260
+ -1.0,
261
+ 0.0
262
+ ],
263
+ "max": [
264
+ 1.0,
265
+ 1.0,
266
+ 1.0,
267
+ 0.0
268
+ ],
269
+ "mean": [
270
+ 0.001149358291412807,
271
+ 0.0002466484662750117,
272
+ -0.00040534232871083475,
273
+ 0.0
274
+ ],
275
+ "std": [
276
+ 0.06498312870886347,
277
+ 0.0645472655438852,
278
+ 0.04779180108146778,
279
+ 0.0
280
+ ],
281
+ "q01": [
282
+ -1.0,
283
+ -0.9285714030265808,
284
+ -0.9257143139839172,
285
+ 0.0
286
+ ],
287
+ "q99": [
288
+ 1.0,
289
+ 1.0,
290
+ 0.9885714054107666,
291
+ 0.0
292
+ ]
293
+ },
294
+ "control_mode": {
295
+ "min": [
296
+ -1.0
297
+ ],
298
+ "max": [
299
+ 1.0
300
+ ],
301
+ "mean": [
302
+ -0.966280820556342
303
+ ],
304
+ "std": [
305
+ 0.25748093445938525
306
+ ],
307
+ "q01": [
308
+ -1.0
309
+ ],
310
+ "q99": [
311
+ 1.0
312
+ ]
313
+ }
314
+ },
315
+ "relative_action": {}
316
+ }
317
+ }
experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "Gr00tN1d6",
3
+ "model_dtype": "bfloat16",
4
+ "model_name": "nvidia/Eagle-Block2A-2B-v2",
5
+ "backbone_model_type": "eagle",
6
+ "model_revision": null,
7
+ "tune_top_llm_layers": 4,
8
+ "backbone_embedding_dim": 2048,
9
+ "tune_llm": false,
10
+ "tune_visual": false,
11
+ "select_layer": 16,
12
+ "reproject_vision": false,
13
+ "use_flash_attention": true,
14
+ "load_bf16": true,
15
+ "collator_overwrite_image_inputs": false,
16
+ "eagle_collator": true,
17
+ "backbone_trainable_params_fp32": true,
18
+ "apply_sincos_state_encoding": true,
19
+ "use_relative_action": true,
20
+ "max_state_dim": 128,
21
+ "max_action_dim": 128,
22
+ "action_horizon": 50,
23
+ "hidden_size": 1024,
24
+ "input_embedding_dim": 1536,
25
+ "add_pos_embed": true,
26
+ "attn_dropout": 0.2,
27
+ "use_vlln": true,
28
+ "max_seq_len": 1024,
29
+ "use_alternate_vl_dit": true,
30
+ "attend_text_every_n_blocks": 2,
31
+ "diffusion_model_cfg": {
32
+ "attention_head_dim": 48,
33
+ "dropout": 0.2,
34
+ "final_dropout": true,
35
+ "interleave_self_attention": true,
36
+ "norm_type": "ada_norm",
37
+ "num_attention_heads": 32,
38
+ "num_layers": 32,
39
+ "output_dim": 1024,
40
+ "positional_embeddings": null
41
+ },
42
+ "num_inference_timesteps": 4,
43
+ "noise_beta_alpha": 1.5,
44
+ "noise_beta_beta": 1.0,
45
+ "noise_s": 0.999,
46
+ "num_timestep_buckets": 1000,
47
+ "tune_projector": true,
48
+ "tune_diffusion_model": true,
49
+ "tune_vlln": true,
50
+ "state_dropout_prob": 0.0,
51
+ "state_additive_noise_scale": 0.0,
52
+ "max_num_embodiments": 32
53
+ }
experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff