Pamela153 commited on
Commit
d1a7c25
·
verified ·
1 Parent(s): 9fc703e

Upload folder using huggingface_hub

Browse files
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,385 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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/videos/atomic_split_train/AdjustToasterOvenTemperature/GraspObject/20250820/lerobot
65
+ - /root/videos/atomic_split_train/AdjustToasterOvenTemperature/TwistKnob/20250820/lerobot
66
+ - /root/videos/atomic_split_train/AdjustWaterTemperature/GraspObject/20250820/lerobot
67
+ - /root/videos/atomic_split_train/AdjustWaterTemperature/TwistKnob/20250820/lerobot
68
+ - /root/videos/atomic_split_train/CheesyBread/PickObject/20250714/lerobot
69
+ - /root/videos/atomic_split_train/CheesyBread/PlaceObject/20250714/lerobot
70
+ - /root/videos/atomic_split_train/CloseBlenderLid/PickObject/20250822/lerobot
71
+ - /root/videos/atomic_split_train/CloseBlenderLid/PlaceObject/20250822/lerobot
72
+ - /root/videos/atomic_split_train/CloseCabinet-DoubleDoor/ApproachToObject_0/20250819/lerobot
73
+ - /root/videos/atomic_split_train/CloseCabinet-DoubleDoor/ApproachToObject_1/20250819/lerobot
74
+ - /root/videos/atomic_split_train/CloseCabinet-DoubleDoor/CloseDoor_0/20250819/lerobot
75
+ - /root/videos/atomic_split_train/CloseCabinet-DoubleDoor/CloseDoor_1/20250819/lerobot
76
+ - /root/videos/atomic_split_train/CloseCabinet-SingleDoor/ApproachToObject/20250819/lerobot
77
+ - /root/videos/atomic_split_train/CloseCabinet-SingleDoor/CloseDoor/20250819/lerobot
78
+ - /root/videos/atomic_split_train/CloseCabinet/20250819/lerobot
79
+ - /root/videos/atomic_split_train/CloseDishwasher/ApproachToObject/20250820/lerobot
80
+ - /root/videos/atomic_split_train/CloseDishwasher/PushClose/20250820/lerobot
81
+ - /root/videos/atomic_split_train/CloseDrawer/ApproachToObject/20250819/lerobot
82
+ - /root/videos/atomic_split_train/CloseDrawer/PushClose/20250819/lerobot
83
+ - /root/videos/atomic_split_train/CloseElectricKettleLid/20250820/lerobot
84
+ - /root/videos/atomic_split_train/CloseFridge-DoubleDoor/ApproachToObject_0/20250819/lerobot
85
+ - /root/videos/atomic_split_train/CloseFridge-DoubleDoor/ApproachToObject_1/20250819/lerobot
86
+ - /root/videos/atomic_split_train/CloseFridge-DoubleDoor/CloseDoor_0/20250819/lerobot
87
+ - /root/videos/atomic_split_train/CloseFridge-DoubleDoor/CloseDoor_1/20250819/lerobot
88
+ - /root/videos/atomic_split_train/CloseFridge-SingleDoor/ApproachToObject/20250819/lerobot
89
+ - /root/videos/atomic_split_train/CloseFridge-SingleDoor/CloseDoor/20250819/lerobot
90
+ - /root/videos/atomic_split_train/CloseFridgeDrawer/ApproachToObject/20250821/lerobot
91
+ - /root/videos/atomic_split_train/CloseFridgeDrawer/PushClose/20250821/lerobot
92
+ - /root/videos/atomic_split_train/CloseMicrowave/ApproachToObject/20250819/lerobot
93
+ - /root/videos/atomic_split_train/CloseMicrowave/CloseDoor/20250819/lerobot
94
+ - /root/videos/atomic_split_train/CloseOven-split/ApproachToObject/20250820/lerobot
95
+ - /root/videos/atomic_split_train/CloseOven-split/PushClose/20250820/lerobot
96
+ - /root/videos/atomic_split_train/CloseOven/20250820/lerobot
97
+ - /root/videos/atomic_split_train/CloseStandMixerHead-split/ApproachToObject/20250820/lerobot
98
+ - /root/videos/atomic_split_train/CloseStandMixerHead-split/CloseDoor/20250820/lerobot
99
+ - /root/videos/atomic_split_train/CloseStandMixerHead/20250820/lerobot
100
+ - /root/videos/atomic_split_train/CloseToasterOvenDoor-split/ApproachToObject/20250820/lerobot
101
+ - /root/videos/atomic_split_train/CloseToasterOvenDoor-split/CloseDoor/20250820/lerobot
102
+ - /root/videos/atomic_split_train/CloseToasterOvenDoor/20250820/lerobot
103
+ - /root/videos/atomic_split_train/CoffeeServeMug/PickObject/20250819/lerobot
104
+ - /root/videos/atomic_split_train/CoffeeServeMug/PlaceObject/20250819/lerobot
105
+ - /root/videos/atomic_split_train/CoffeeSetupMug/PickObject/20250819/lerobot
106
+ - /root/videos/atomic_split_train/CoffeeSetupMug/PlaceObject/20250819/lerobot
107
+ - /root/videos/atomic_split_train/LowerHeat/GraspObject/20250805/lerobot
108
+ - /root/videos/atomic_split_train/LowerHeat/TwistKnob/20250805/lerobot
109
+ - /root/videos/atomic_split_train/MakeIcedCoffee/PickObject/20250801/lerobot
110
+ - /root/videos/atomic_split_train/MakeIcedCoffee/PlaceObject/20250801/lerobot
111
+ - /root/videos/atomic_split_train/OpenBlenderLid/PickObject/20250822/lerobot
112
+ - /root/videos/atomic_split_train/OpenBlenderLid/PlaceObject/20250822/lerobot
113
+ - /root/videos/atomic_split_train/OpenCabinet-DoubleDoor/GraspObject_0/20250819/lerobot
114
+ - /root/videos/atomic_split_train/OpenCabinet-DoubleDoor/GraspObject_1/20250819/lerobot
115
+ - /root/videos/atomic_split_train/OpenCabinet-DoubleDoor/OpenDoor_0/20250819/lerobot
116
+ - /root/videos/atomic_split_train/OpenCabinet-DoubleDoor/OpenDoor_1/20250819/lerobot
117
+ - /root/videos/atomic_split_train/OpenCabinet-SingleDoor/GraspObject/20250819/lerobot
118
+ - /root/videos/atomic_split_train/OpenCabinet-SingleDoor/OpenDoor/20250819/lerobot
119
+ - /root/videos/atomic_split_train/OpenDishwasher/GraspObject/20250820/lerobot
120
+ - /root/videos/atomic_split_train/OpenDishwasher/OpenDoor/20250820/lerobot
121
+ - /root/videos/atomic_split_train/OpenDrawer/GraspObject/20250819/lerobot
122
+ - /root/videos/atomic_split_train/OpenDrawer/PullOpen/20250819/lerobot
123
+ - /root/videos/atomic_split_train/OpenElectricKettleLid-split/ApproachToObject/20250820/lerobot
124
+ - /root/videos/atomic_split_train/OpenElectricKettleLid-split/PressButton/20250820/lerobot
125
+ - /root/videos/atomic_split_train/OpenElectricKettleLid/20250820/lerobot
126
+ - /root/videos/atomic_split_train/OpenFridge-DoubleDoor/GraspObject_0/20250819/lerobot
127
+ - /root/videos/atomic_split_train/OpenFridge-DoubleDoor/GraspObject_1/20250819/lerobot
128
+ - /root/videos/atomic_split_train/OpenFridge-DoubleDoor/OpenDoor_0/20250819/lerobot
129
+ - /root/videos/atomic_split_train/OpenFridge-DoubleDoor/OpenDoor_1/20250819/lerobot
130
+ - /root/videos/atomic_split_train/OpenFridge-SingleDoor/GraspObject/20250819/lerobot
131
+ - /root/videos/atomic_split_train/OpenFridge-SingleDoor/OpenDoor/20250819/lerobot
132
+ - /root/videos/atomic_split_train/OpenFridgeDrawer-split/GraspObject/20250821/lerobot
133
+ - /root/videos/atomic_split_train/OpenFridgeDrawer-split/PullOpen/20250821/lerobot
134
+ - /root/videos/atomic_split_train/OpenFridgeDrawer/20250821/lerobot
135
+ - /root/videos/atomic_split_train/OpenMicrowave/GraspObject/20250819/lerobot
136
+ - /root/videos/atomic_split_train/OpenMicrowave/OpenDoor/20250819/lerobot
137
+ - /root/videos/atomic_split_train/OpenOven/GraspObject/20250820/lerobot
138
+ - /root/videos/atomic_split_train/OpenOven/OpenDoor/20250820/lerobot
139
+ - /root/videos/atomic_split_train/OpenStandMixerHead-split/GraspObject/20250820/lerobot
140
+ - /root/videos/atomic_split_train/OpenStandMixerHead-split/OpenDoor/20250820/lerobot
141
+ - /root/videos/atomic_split_train/OpenStandMixerHead/20250820/lerobot
142
+ - /root/videos/atomic_split_train/OpenToasterOvenDoor/GraspObject/20250820/lerobot
143
+ - /root/videos/atomic_split_train/OpenToasterOvenDoor/OpenDoor/20250820/lerobot
144
+ - /root/videos/atomic_split_train/PackDessert/PickObject/20250806/lerobot
145
+ - /root/videos/atomic_split_train/PackDessert/PlaceObject/20250806/lerobot
146
+ - /root/videos/atomic_split_train/PickPlaceCabinetToCounter/PickObject/20250819/lerobot
147
+ - /root/videos/atomic_split_train/PickPlaceCabinetToCounter/PlaceObject/20250819/lerobot
148
+ - /root/videos/atomic_split_train/PickPlaceCounterToBlender/PickObject/20250822/lerobot
149
+ - /root/videos/atomic_split_train/PickPlaceCounterToBlender/PlaceObject/20250822/lerobot
150
+ - /root/videos/atomic_split_train/PickPlaceCounterToCabinet/PickObject/20250819/lerobot
151
+ - /root/videos/atomic_split_train/PickPlaceCounterToCabinet/PlaceObject/20250819/lerobot
152
+ - /root/videos/atomic_split_train/PickPlaceCounterToDrawer/PickObject/20250821/lerobot
153
+ - /root/videos/atomic_split_train/PickPlaceCounterToDrawer/PlaceObject/20250821/lerobot
154
+ - /root/videos/atomic_split_train/PickPlaceCounterToMicrowave/PickObject/20250819/lerobot
155
+ - /root/videos/atomic_split_train/PickPlaceCounterToMicrowave/PlaceObject/20250819/lerobot
156
+ - /root/videos/atomic_split_train/PickPlaceCounterToOven/PickObject/20250819/lerobot
157
+ - /root/videos/atomic_split_train/PickPlaceCounterToOven/PlaceObject/20250819/lerobot
158
+ - /root/videos/atomic_split_train/PickPlaceCounterToSink/PickObject/20250819/lerobot
159
+ - /root/videos/atomic_split_train/PickPlaceCounterToSink/PlaceObject/20250819/lerobot
160
+ - /root/videos/atomic_split_train/PickPlaceCounterToStandMixer/PickObject/20250820/lerobot
161
+ - /root/videos/atomic_split_train/PickPlaceCounterToStandMixer/PlaceObject/20250820/lerobot
162
+ - /root/videos/atomic_split_train/PickPlaceCounterToStove/PickObject/20250819/lerobot
163
+ - /root/videos/atomic_split_train/PickPlaceCounterToStove/PlaceObject/20250819/lerobot
164
+ - /root/videos/atomic_split_train/PickPlaceCounterToToasterOven/PickObject/20250819/lerobot
165
+ - /root/videos/atomic_split_train/PickPlaceCounterToToasterOven/PlaceObject/20250819/lerobot
166
+ - /root/videos/atomic_split_train/PickPlaceDrawerToCounter/PickObject/20250820/lerobot
167
+ - /root/videos/atomic_split_train/PickPlaceDrawerToCounter/PlaceObject/20250820/lerobot
168
+ - /root/videos/atomic_split_train/PickPlaceFridgeDrawerToShelf/PickObject/20250821/lerobot
169
+ - /root/videos/atomic_split_train/PickPlaceFridgeDrawerToShelf/PlaceObject/20250821/lerobot
170
+ - /root/videos/atomic_split_train/PickPlaceFridgeShelfToDrawer/PickObject/20250821/lerobot
171
+ - /root/videos/atomic_split_train/PickPlaceFridgeShelfToDrawer/PlaceObject/20250821/lerobot
172
+ - /root/videos/atomic_split_train/PickPlaceMicrowaveToCounter/PickObject/20250819/lerobot
173
+ - /root/videos/atomic_split_train/PickPlaceMicrowaveToCounter/PlaceObject/20250819/lerobot
174
+ - /root/videos/atomic_split_train/PickPlaceSinkToCounter/PickObject/20250819/lerobot
175
+ - /root/videos/atomic_split_train/PickPlaceSinkToCounter/PlaceObject/20250819/lerobot
176
+ - /root/videos/atomic_split_train/PickPlaceStoveToCounter/PickObject/20250819/lerobot
177
+ - /root/videos/atomic_split_train/PickPlaceStoveToCounter/PlaceObject/20250819/lerobot
178
+ - /root/videos/atomic_split_train/PickPlaceToasterOvenToCounter/PickObject/20250819/lerobot
179
+ - /root/videos/atomic_split_train/PickPlaceToasterOvenToCounter/PlaceObject/20250819/lerobot
180
+ - /root/videos/atomic_split_train/PickPlaceToasterToCounter/PickObject/20250819/lerobot
181
+ - /root/videos/atomic_split_train/PickPlaceToasterToCounter/PlaceObject/20250819/lerobot
182
+ - /root/videos/atomic_split_train/PreheatOven/GraspObject/20250903/lerobot
183
+ - /root/videos/atomic_split_train/PreheatOven/TwistKnob/20250903/lerobot
184
+ - /root/videos/atomic_split_train/SlideDishwasherRack-split/GraspObject/20250820/lerobot
185
+ - /root/videos/atomic_split_train/SlideDishwasherRack-split/SlideRack/20250820/lerobot
186
+ - /root/videos/atomic_split_train/SlideDishwasherRack/20250820/lerobot
187
+ - /root/videos/atomic_split_train/SlideOvenRack-split/GraspObject/20250820/lerobot
188
+ - /root/videos/atomic_split_train/SlideOvenRack-split/SlideRack/20250820/lerobot
189
+ - /root/videos/atomic_split_train/SlideOvenRack/20250820/lerobot
190
+ - /root/videos/atomic_split_train/SlideToasterOvenRack-split/GraspObject/20250820/lerobot
191
+ - /root/videos/atomic_split_train/SlideToasterOvenRack-split/SlideRack/20250820/lerobot
192
+ - /root/videos/atomic_split_train/SlideToasterOvenRack/20250820/lerobot
193
+ - /root/videos/atomic_split_train/StartCoffeeMachine-split/ApproachToObject/20250819/lerobot
194
+ - /root/videos/atomic_split_train/StartCoffeeMachine-split/PressButton/20250819/lerobot
195
+ - /root/videos/atomic_split_train/StartCoffeeMachine/20250819/lerobot
196
+ - /root/videos/atomic_split_train/TurnOffMicrowave-split/ApproachToObject/20250819/lerobot
197
+ - /root/videos/atomic_split_train/TurnOffMicrowave-split/PressButton/20250819/lerobot
198
+ - /root/videos/atomic_split_train/TurnOffMicrowave/20250819/lerobot
199
+ - /root/videos/atomic_split_train/TurnOffSinkFaucet-split/GraspObject/20250819/lerobot
200
+ - /root/videos/atomic_split_train/TurnOffSinkFaucet-split/TwistKnob/20250819/lerobot
201
+ - /root/videos/atomic_split_train/TurnOffSinkFaucet/20250819/lerobot
202
+ - /root/videos/atomic_split_train/TurnOffStove/GraspObject/20250819/lerobot
203
+ - /root/videos/atomic_split_train/TurnOffStove/TwistKnob/20250819/lerobot
204
+ - /root/videos/atomic_split_train/TurnOnBlender-split/ApproachToObject/20250822/lerobot
205
+ - /root/videos/atomic_split_train/TurnOnBlender-split/PressButton/20250822/lerobot
206
+ - /root/videos/atomic_split_train/TurnOnBlender/20250822/lerobot
207
+ - /root/videos/atomic_split_train/TurnOnElectricKettle-split/ApproachToObject/20250820/lerobot
208
+ - /root/videos/atomic_split_train/TurnOnElectricKettle-split/PressButton/20250820/lerobot
209
+ - /root/videos/atomic_split_train/TurnOnElectricKettle/20250820/lerobot
210
+ - /root/videos/atomic_split_train/TurnOnMicrowave-split/ApproachToObject/20250819/lerobot
211
+ - /root/videos/atomic_split_train/TurnOnMicrowave-split/PressButton/20250819/lerobot
212
+ - /root/videos/atomic_split_train/TurnOnMicrowave/20250819/lerobot
213
+ - /root/videos/atomic_split_train/TurnOnSinkFaucet-split/GraspObject/20250819/lerobot
214
+ - /root/videos/atomic_split_train/TurnOnSinkFaucet-split/TwistKnob/20250819/lerobot
215
+ - /root/videos/atomic_split_train/TurnOnSinkFaucet/20250819/lerobot
216
+ - /root/videos/atomic_split_train/TurnOnStove/GraspObject/20250819/lerobot
217
+ - /root/videos/atomic_split_train/TurnOnStove/TwistKnob/20250819/lerobot
218
+ - /root/videos/atomic_split_train/TurnOnToaster-split/ApproachToObject/20250820/lerobot
219
+ - /root/videos/atomic_split_train/TurnOnToaster-split/PressButton/20250820/lerobot
220
+ - /root/videos/atomic_split_train/TurnOnToaster/20250820/lerobot
221
+ - /root/videos/atomic_split_train/TurnOnToasterOven/GraspObject/20250820/lerobot
222
+ - /root/videos/atomic_split_train/TurnOnToasterOven/TwistKnob/20250820/lerobot
223
+ - /root/videos/atomic_split_train/TurnSinkSpout/20250820/lerobot
224
+ - /root/videos/atomic_split_train/TurnSinkSpout/GraspObject/20250820/lerobot
225
+ - /root/videos/atomic_split_train/TurnSinkSpout/TurnSinkSpout/20250820/lerobot
226
+ embodiment_tag: robocasa_panda_omron
227
+ mix_ratio: 1.0
228
+ dataset_type: physical_embodiment
229
+ val_dataset_path: null
230
+ modality_configs:
231
+ robocasa_panda_omron:
232
+ video:
233
+ delta_indices:
234
+ - 0
235
+ modality_keys:
236
+ - res256_image_side_0
237
+ - res256_image_side_1
238
+ - res256_image_wrist_0
239
+ sin_cos_embedding_keys: null
240
+ mean_std_embedding_keys: null
241
+ action_configs: null
242
+ state:
243
+ delta_indices:
244
+ - 0
245
+ modality_keys:
246
+ - end_effector_position_relative
247
+ - end_effector_rotation_relative
248
+ - gripper_qpos
249
+ - base_position
250
+ - base_rotation
251
+ sin_cos_embedding_keys: null
252
+ mean_std_embedding_keys: null
253
+ action_configs: null
254
+ action:
255
+ delta_indices:
256
+ - 0
257
+ - 1
258
+ - 2
259
+ - 3
260
+ - 4
261
+ - 5
262
+ - 6
263
+ - 7
264
+ - 8
265
+ - 9
266
+ - 10
267
+ - 11
268
+ - 12
269
+ - 13
270
+ - 14
271
+ - 15
272
+ modality_keys:
273
+ - end_effector_position
274
+ - end_effector_rotation
275
+ - gripper_close
276
+ - base_motion
277
+ - control_mode
278
+ sin_cos_embedding_keys: null
279
+ mean_std_embedding_keys: null
280
+ action_configs:
281
+ - rep: ABSOLUTE
282
+ type: NON_EEF
283
+ format: DEFAULT
284
+ state_key: null
285
+ - rep: ABSOLUTE
286
+ type: NON_EEF
287
+ format: DEFAULT
288
+ state_key: null
289
+ - rep: ABSOLUTE
290
+ type: NON_EEF
291
+ format: DEFAULT
292
+ state_key: null
293
+ - rep: ABSOLUTE
294
+ type: NON_EEF
295
+ format: DEFAULT
296
+ state_key: null
297
+ - rep: ABSOLUTE
298
+ type: NON_EEF
299
+ format: DEFAULT
300
+ state_key: null
301
+ language:
302
+ delta_indices:
303
+ - 0
304
+ modality_keys:
305
+ - annotation.human.task_description
306
+ sin_cos_embedding_keys: null
307
+ mean_std_embedding_keys: null
308
+ action_configs: null
309
+ download_cache: false
310
+ shard_size: 1024
311
+ episode_sampling_rate: 0.1
312
+ num_shards_per_epoch: 100000
313
+ override_pretraining_statistics: false
314
+ mode: single_turn
315
+ random_chop: 0.0
316
+ mock_dataset_mode: false
317
+ shuffle: true
318
+ seed: 42
319
+ multiprocessing_context: fork
320
+ allow_padding: false
321
+ subsample_ratio: 1.0
322
+ image_crop_size:
323
+ - 244
324
+ - 244
325
+ image_target_size:
326
+ - 224
327
+ - 224
328
+ video_backend: torchcodec
329
+ training:
330
+ output_dir: /root/outputs/gr00t-all-tasks
331
+ experiment_name: null
332
+ max_steps: 50000
333
+ global_batch_size: 64
334
+ batch_size: null
335
+ gradient_accumulation_steps: 1
336
+ learning_rate: 0.0001
337
+ lr_scheduler_type: cosine
338
+ weight_decay: 1.0e-05
339
+ warmup_ratio: 0.05
340
+ warmup_steps: 0
341
+ max_grad_norm: 1.0
342
+ optim: adamw_torch
343
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
344
+ tf32: true
345
+ fp16: false
346
+ bf16: true
347
+ eval_bf16: true
348
+ logging_steps: 10
349
+ save_steps: 2500
350
+ save_total_limit: 5
351
+ save_vl_model: false
352
+ upload_checkpoints: false
353
+ upload_every: 1000
354
+ upload_last_n_checkpoints: 5
355
+ max_concurrent_uploads: 2
356
+ eval_strategy: 'no'
357
+ eval_steps: 500
358
+ eval_set_split_ratio: 0.1
359
+ eval_batch_size: 2
360
+ save_best_eval_metric_name: ''
361
+ save_best_eval_metric_greater_is_better: true
362
+ deepspeed_stage: 2
363
+ gradient_checkpointing: false
364
+ transformers_trust_remote_code: true
365
+ transformers_local_files_only: false
366
+ transformers_cache_dir: null
367
+ transformers_access_token: null
368
+ use_ddp: false
369
+ ddp_bucket_cap_mb: 100
370
+ num_gpus: 8
371
+ dataloader_num_workers: 4
372
+ remove_unused_columns: false
373
+ use_wandb: false
374
+ wandb_project: finetune-gr00t-n1d6
375
+ enable_profiling: false
376
+ max_retries: 3
377
+ assert_loss_less_than: null
378
+ add_rl_callback: false
379
+ enable_open_loop_eval: false
380
+ open_loop_eval_traj_ids:
381
+ - 0
382
+ open_loop_eval_steps_per_traj: 100
383
+ open_loop_eval_plot_indices: null
384
+ max_steps: 50000
385
+ save_steps: 2500
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,421 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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/videos/atomic_split_train/AdjustToasterOvenTemperature/GraspObject/20250820/lerobot
8
+ - /root/videos/atomic_split_train/AdjustToasterOvenTemperature/TwistKnob/20250820/lerobot
9
+ - /root/videos/atomic_split_train/AdjustWaterTemperature/GraspObject/20250820/lerobot
10
+ - /root/videos/atomic_split_train/AdjustWaterTemperature/TwistKnob/20250820/lerobot
11
+ - /root/videos/atomic_split_train/CheesyBread/PickObject/20250714/lerobot
12
+ - /root/videos/atomic_split_train/CheesyBread/PlaceObject/20250714/lerobot
13
+ - /root/videos/atomic_split_train/CloseBlenderLid/PickObject/20250822/lerobot
14
+ - /root/videos/atomic_split_train/CloseBlenderLid/PlaceObject/20250822/lerobot
15
+ - /root/videos/atomic_split_train/CloseCabinet-DoubleDoor/ApproachToObject_0/20250819/lerobot
16
+ - /root/videos/atomic_split_train/CloseCabinet-DoubleDoor/ApproachToObject_1/20250819/lerobot
17
+ - /root/videos/atomic_split_train/CloseCabinet-DoubleDoor/CloseDoor_0/20250819/lerobot
18
+ - /root/videos/atomic_split_train/CloseCabinet-DoubleDoor/CloseDoor_1/20250819/lerobot
19
+ - /root/videos/atomic_split_train/CloseCabinet-SingleDoor/ApproachToObject/20250819/lerobot
20
+ - /root/videos/atomic_split_train/CloseCabinet-SingleDoor/CloseDoor/20250819/lerobot
21
+ - /root/videos/atomic_split_train/CloseCabinet/20250819/lerobot
22
+ - /root/videos/atomic_split_train/CloseDishwasher/ApproachToObject/20250820/lerobot
23
+ - /root/videos/atomic_split_train/CloseDishwasher/PushClose/20250820/lerobot
24
+ - /root/videos/atomic_split_train/CloseDrawer/ApproachToObject/20250819/lerobot
25
+ - /root/videos/atomic_split_train/CloseDrawer/PushClose/20250819/lerobot
26
+ - /root/videos/atomic_split_train/CloseElectricKettleLid/20250820/lerobot
27
+ - /root/videos/atomic_split_train/CloseFridge-DoubleDoor/ApproachToObject_0/20250819/lerobot
28
+ - /root/videos/atomic_split_train/CloseFridge-DoubleDoor/ApproachToObject_1/20250819/lerobot
29
+ - /root/videos/atomic_split_train/CloseFridge-DoubleDoor/CloseDoor_0/20250819/lerobot
30
+ - /root/videos/atomic_split_train/CloseFridge-DoubleDoor/CloseDoor_1/20250819/lerobot
31
+ - /root/videos/atomic_split_train/CloseFridge-SingleDoor/ApproachToObject/20250819/lerobot
32
+ - /root/videos/atomic_split_train/CloseFridge-SingleDoor/CloseDoor/20250819/lerobot
33
+ - /root/videos/atomic_split_train/CloseFridgeDrawer/ApproachToObject/20250821/lerobot
34
+ - /root/videos/atomic_split_train/CloseFridgeDrawer/PushClose/20250821/lerobot
35
+ - /root/videos/atomic_split_train/CloseMicrowave/ApproachToObject/20250819/lerobot
36
+ - /root/videos/atomic_split_train/CloseMicrowave/CloseDoor/20250819/lerobot
37
+ - /root/videos/atomic_split_train/CloseOven-split/ApproachToObject/20250820/lerobot
38
+ - /root/videos/atomic_split_train/CloseOven-split/PushClose/20250820/lerobot
39
+ - /root/videos/atomic_split_train/CloseOven/20250820/lerobot
40
+ - /root/videos/atomic_split_train/CloseStandMixerHead-split/ApproachToObject/20250820/lerobot
41
+ - /root/videos/atomic_split_train/CloseStandMixerHead-split/CloseDoor/20250820/lerobot
42
+ - /root/videos/atomic_split_train/CloseStandMixerHead/20250820/lerobot
43
+ - /root/videos/atomic_split_train/CloseToasterOvenDoor-split/ApproachToObject/20250820/lerobot
44
+ - /root/videos/atomic_split_train/CloseToasterOvenDoor-split/CloseDoor/20250820/lerobot
45
+ - /root/videos/atomic_split_train/CloseToasterOvenDoor/20250820/lerobot
46
+ - /root/videos/atomic_split_train/CoffeeServeMug/PickObject/20250819/lerobot
47
+ - /root/videos/atomic_split_train/CoffeeServeMug/PlaceObject/20250819/lerobot
48
+ - /root/videos/atomic_split_train/CoffeeSetupMug/PickObject/20250819/lerobot
49
+ - /root/videos/atomic_split_train/CoffeeSetupMug/PlaceObject/20250819/lerobot
50
+ - /root/videos/atomic_split_train/LowerHeat/GraspObject/20250805/lerobot
51
+ - /root/videos/atomic_split_train/LowerHeat/TwistKnob/20250805/lerobot
52
+ - /root/videos/atomic_split_train/MakeIcedCoffee/PickObject/20250801/lerobot
53
+ - /root/videos/atomic_split_train/MakeIcedCoffee/PlaceObject/20250801/lerobot
54
+ - /root/videos/atomic_split_train/OpenBlenderLid/PickObject/20250822/lerobot
55
+ - /root/videos/atomic_split_train/OpenBlenderLid/PlaceObject/20250822/lerobot
56
+ - /root/videos/atomic_split_train/OpenCabinet-DoubleDoor/GraspObject_0/20250819/lerobot
57
+ - /root/videos/atomic_split_train/OpenCabinet-DoubleDoor/GraspObject_1/20250819/lerobot
58
+ - /root/videos/atomic_split_train/OpenCabinet-DoubleDoor/OpenDoor_0/20250819/lerobot
59
+ - /root/videos/atomic_split_train/OpenCabinet-DoubleDoor/OpenDoor_1/20250819/lerobot
60
+ - /root/videos/atomic_split_train/OpenCabinet-SingleDoor/GraspObject/20250819/lerobot
61
+ - /root/videos/atomic_split_train/OpenCabinet-SingleDoor/OpenDoor/20250819/lerobot
62
+ - /root/videos/atomic_split_train/OpenDishwasher/GraspObject/20250820/lerobot
63
+ - /root/videos/atomic_split_train/OpenDishwasher/OpenDoor/20250820/lerobot
64
+ - /root/videos/atomic_split_train/OpenDrawer/GraspObject/20250819/lerobot
65
+ - /root/videos/atomic_split_train/OpenDrawer/PullOpen/20250819/lerobot
66
+ - /root/videos/atomic_split_train/OpenElectricKettleLid-split/ApproachToObject/20250820/lerobot
67
+ - /root/videos/atomic_split_train/OpenElectricKettleLid-split/PressButton/20250820/lerobot
68
+ - /root/videos/atomic_split_train/OpenElectricKettleLid/20250820/lerobot
69
+ - /root/videos/atomic_split_train/OpenFridge-DoubleDoor/GraspObject_0/20250819/lerobot
70
+ - /root/videos/atomic_split_train/OpenFridge-DoubleDoor/GraspObject_1/20250819/lerobot
71
+ - /root/videos/atomic_split_train/OpenFridge-DoubleDoor/OpenDoor_0/20250819/lerobot
72
+ - /root/videos/atomic_split_train/OpenFridge-DoubleDoor/OpenDoor_1/20250819/lerobot
73
+ - /root/videos/atomic_split_train/OpenFridge-SingleDoor/GraspObject/20250819/lerobot
74
+ - /root/videos/atomic_split_train/OpenFridge-SingleDoor/OpenDoor/20250819/lerobot
75
+ - /root/videos/atomic_split_train/OpenFridgeDrawer-split/GraspObject/20250821/lerobot
76
+ - /root/videos/atomic_split_train/OpenFridgeDrawer-split/PullOpen/20250821/lerobot
77
+ - /root/videos/atomic_split_train/OpenFridgeDrawer/20250821/lerobot
78
+ - /root/videos/atomic_split_train/OpenMicrowave/GraspObject/20250819/lerobot
79
+ - /root/videos/atomic_split_train/OpenMicrowave/OpenDoor/20250819/lerobot
80
+ - /root/videos/atomic_split_train/OpenOven/GraspObject/20250820/lerobot
81
+ - /root/videos/atomic_split_train/OpenOven/OpenDoor/20250820/lerobot
82
+ - /root/videos/atomic_split_train/OpenStandMixerHead-split/GraspObject/20250820/lerobot
83
+ - /root/videos/atomic_split_train/OpenStandMixerHead-split/OpenDoor/20250820/lerobot
84
+ - /root/videos/atomic_split_train/OpenStandMixerHead/20250820/lerobot
85
+ - /root/videos/atomic_split_train/OpenToasterOvenDoor/GraspObject/20250820/lerobot
86
+ - /root/videos/atomic_split_train/OpenToasterOvenDoor/OpenDoor/20250820/lerobot
87
+ - /root/videos/atomic_split_train/PackDessert/PickObject/20250806/lerobot
88
+ - /root/videos/atomic_split_train/PackDessert/PlaceObject/20250806/lerobot
89
+ - /root/videos/atomic_split_train/PickPlaceCabinetToCounter/PickObject/20250819/lerobot
90
+ - /root/videos/atomic_split_train/PickPlaceCabinetToCounter/PlaceObject/20250819/lerobot
91
+ - /root/videos/atomic_split_train/PickPlaceCounterToBlender/PickObject/20250822/lerobot
92
+ - /root/videos/atomic_split_train/PickPlaceCounterToBlender/PlaceObject/20250822/lerobot
93
+ - /root/videos/atomic_split_train/PickPlaceCounterToCabinet/PickObject/20250819/lerobot
94
+ - /root/videos/atomic_split_train/PickPlaceCounterToCabinet/PlaceObject/20250819/lerobot
95
+ - /root/videos/atomic_split_train/PickPlaceCounterToDrawer/PickObject/20250821/lerobot
96
+ - /root/videos/atomic_split_train/PickPlaceCounterToDrawer/PlaceObject/20250821/lerobot
97
+ - /root/videos/atomic_split_train/PickPlaceCounterToMicrowave/PickObject/20250819/lerobot
98
+ - /root/videos/atomic_split_train/PickPlaceCounterToMicrowave/PlaceObject/20250819/lerobot
99
+ - /root/videos/atomic_split_train/PickPlaceCounterToOven/PickObject/20250819/lerobot
100
+ - /root/videos/atomic_split_train/PickPlaceCounterToOven/PlaceObject/20250819/lerobot
101
+ - /root/videos/atomic_split_train/PickPlaceCounterToSink/PickObject/20250819/lerobot
102
+ - /root/videos/atomic_split_train/PickPlaceCounterToSink/PlaceObject/20250819/lerobot
103
+ - /root/videos/atomic_split_train/PickPlaceCounterToStandMixer/PickObject/20250820/lerobot
104
+ - /root/videos/atomic_split_train/PickPlaceCounterToStandMixer/PlaceObject/20250820/lerobot
105
+ - /root/videos/atomic_split_train/PickPlaceCounterToStove/PickObject/20250819/lerobot
106
+ - /root/videos/atomic_split_train/PickPlaceCounterToStove/PlaceObject/20250819/lerobot
107
+ - /root/videos/atomic_split_train/PickPlaceCounterToToasterOven/PickObject/20250819/lerobot
108
+ - /root/videos/atomic_split_train/PickPlaceCounterToToasterOven/PlaceObject/20250819/lerobot
109
+ - /root/videos/atomic_split_train/PickPlaceDrawerToCounter/PickObject/20250820/lerobot
110
+ - /root/videos/atomic_split_train/PickPlaceDrawerToCounter/PlaceObject/20250820/lerobot
111
+ - /root/videos/atomic_split_train/PickPlaceFridgeDrawerToShelf/PickObject/20250821/lerobot
112
+ - /root/videos/atomic_split_train/PickPlaceFridgeDrawerToShelf/PlaceObject/20250821/lerobot
113
+ - /root/videos/atomic_split_train/PickPlaceFridgeShelfToDrawer/PickObject/20250821/lerobot
114
+ - /root/videos/atomic_split_train/PickPlaceFridgeShelfToDrawer/PlaceObject/20250821/lerobot
115
+ - /root/videos/atomic_split_train/PickPlaceMicrowaveToCounter/PickObject/20250819/lerobot
116
+ - /root/videos/atomic_split_train/PickPlaceMicrowaveToCounter/PlaceObject/20250819/lerobot
117
+ - /root/videos/atomic_split_train/PickPlaceSinkToCounter/PickObject/20250819/lerobot
118
+ - /root/videos/atomic_split_train/PickPlaceSinkToCounter/PlaceObject/20250819/lerobot
119
+ - /root/videos/atomic_split_train/PickPlaceStoveToCounter/PickObject/20250819/lerobot
120
+ - /root/videos/atomic_split_train/PickPlaceStoveToCounter/PlaceObject/20250819/lerobot
121
+ - /root/videos/atomic_split_train/PickPlaceToasterOvenToCounter/PickObject/20250819/lerobot
122
+ - /root/videos/atomic_split_train/PickPlaceToasterOvenToCounter/PlaceObject/20250819/lerobot
123
+ - /root/videos/atomic_split_train/PickPlaceToasterToCounter/PickObject/20250819/lerobot
124
+ - /root/videos/atomic_split_train/PickPlaceToasterToCounter/PlaceObject/20250819/lerobot
125
+ - /root/videos/atomic_split_train/PreheatOven/GraspObject/20250903/lerobot
126
+ - /root/videos/atomic_split_train/PreheatOven/TwistKnob/20250903/lerobot
127
+ - /root/videos/atomic_split_train/SlideDishwasherRack-split/GraspObject/20250820/lerobot
128
+ - /root/videos/atomic_split_train/SlideDishwasherRack-split/SlideRack/20250820/lerobot
129
+ - /root/videos/atomic_split_train/SlideDishwasherRack/20250820/lerobot
130
+ - /root/videos/atomic_split_train/SlideOvenRack-split/GraspObject/20250820/lerobot
131
+ - /root/videos/atomic_split_train/SlideOvenRack-split/SlideRack/20250820/lerobot
132
+ - /root/videos/atomic_split_train/SlideOvenRack/20250820/lerobot
133
+ - /root/videos/atomic_split_train/SlideToasterOvenRack-split/GraspObject/20250820/lerobot
134
+ - /root/videos/atomic_split_train/SlideToasterOvenRack-split/SlideRack/20250820/lerobot
135
+ - /root/videos/atomic_split_train/SlideToasterOvenRack/20250820/lerobot
136
+ - /root/videos/atomic_split_train/StartCoffeeMachine-split/ApproachToObject/20250819/lerobot
137
+ - /root/videos/atomic_split_train/StartCoffeeMachine-split/PressButton/20250819/lerobot
138
+ - /root/videos/atomic_split_train/StartCoffeeMachine/20250819/lerobot
139
+ - /root/videos/atomic_split_train/TurnOffMicrowave-split/ApproachToObject/20250819/lerobot
140
+ - /root/videos/atomic_split_train/TurnOffMicrowave-split/PressButton/20250819/lerobot
141
+ - /root/videos/atomic_split_train/TurnOffMicrowave/20250819/lerobot
142
+ - /root/videos/atomic_split_train/TurnOffSinkFaucet-split/GraspObject/20250819/lerobot
143
+ - /root/videos/atomic_split_train/TurnOffSinkFaucet-split/TwistKnob/20250819/lerobot
144
+ - /root/videos/atomic_split_train/TurnOffSinkFaucet/20250819/lerobot
145
+ - /root/videos/atomic_split_train/TurnOffStove/GraspObject/20250819/lerobot
146
+ - /root/videos/atomic_split_train/TurnOffStove/TwistKnob/20250819/lerobot
147
+ - /root/videos/atomic_split_train/TurnOnBlender-split/ApproachToObject/20250822/lerobot
148
+ - /root/videos/atomic_split_train/TurnOnBlender-split/PressButton/20250822/lerobot
149
+ - /root/videos/atomic_split_train/TurnOnBlender/20250822/lerobot
150
+ - /root/videos/atomic_split_train/TurnOnElectricKettle-split/ApproachToObject/20250820/lerobot
151
+ - /root/videos/atomic_split_train/TurnOnElectricKettle-split/PressButton/20250820/lerobot
152
+ - /root/videos/atomic_split_train/TurnOnElectricKettle/20250820/lerobot
153
+ - /root/videos/atomic_split_train/TurnOnMicrowave-split/ApproachToObject/20250819/lerobot
154
+ - /root/videos/atomic_split_train/TurnOnMicrowave-split/PressButton/20250819/lerobot
155
+ - /root/videos/atomic_split_train/TurnOnMicrowave/20250819/lerobot
156
+ - /root/videos/atomic_split_train/TurnOnSinkFaucet-split/GraspObject/20250819/lerobot
157
+ - /root/videos/atomic_split_train/TurnOnSinkFaucet-split/TwistKnob/20250819/lerobot
158
+ - /root/videos/atomic_split_train/TurnOnSinkFaucet/20250819/lerobot
159
+ - /root/videos/atomic_split_train/TurnOnStove/GraspObject/20250819/lerobot
160
+ - /root/videos/atomic_split_train/TurnOnStove/TwistKnob/20250819/lerobot
161
+ - /root/videos/atomic_split_train/TurnOnToaster-split/ApproachToObject/20250820/lerobot
162
+ - /root/videos/atomic_split_train/TurnOnToaster-split/PressButton/20250820/lerobot
163
+ - /root/videos/atomic_split_train/TurnOnToaster/20250820/lerobot
164
+ - /root/videos/atomic_split_train/TurnOnToasterOven/GraspObject/20250820/lerobot
165
+ - /root/videos/atomic_split_train/TurnOnToasterOven/TwistKnob/20250820/lerobot
166
+ - /root/videos/atomic_split_train/TurnSinkSpout/20250820/lerobot
167
+ - /root/videos/atomic_split_train/TurnSinkSpout/GraspObject/20250820/lerobot
168
+ - /root/videos/atomic_split_train/TurnSinkSpout/TurnSinkSpout/20250820/lerobot
169
+ dataset_type: physical_embodiment
170
+ embodiment_tag: robocasa_panda_omron
171
+ mix_ratio: 1.0
172
+ val_dataset_path: null
173
+ download_cache: false
174
+ episode_sampling_rate: 0.1
175
+ image_crop_size:
176
+ - 244
177
+ - 244
178
+ image_target_size:
179
+ - 224
180
+ - 224
181
+ mock_dataset_mode: false
182
+ modality_configs:
183
+ robocasa_panda_omron:
184
+ action: !!python/object:gr00t.data.types.ModalityConfig
185
+ action_configs:
186
+ - !!python/object:gr00t.data.types.ActionConfig
187
+ format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
188
+ - default
189
+ rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
190
+ - absolute
191
+ state_key: null
192
+ type: &id003 !!python/object/apply:gr00t.data.types.ActionType
193
+ - non_eef
194
+ - !!python/object:gr00t.data.types.ActionConfig
195
+ format: *id001
196
+ rep: *id002
197
+ state_key: null
198
+ type: *id003
199
+ - !!python/object:gr00t.data.types.ActionConfig
200
+ format: *id001
201
+ rep: *id002
202
+ state_key: null
203
+ type: *id003
204
+ - !!python/object:gr00t.data.types.ActionConfig
205
+ format: *id001
206
+ rep: *id002
207
+ state_key: null
208
+ type: *id003
209
+ - !!python/object:gr00t.data.types.ActionConfig
210
+ format: *id001
211
+ rep: *id002
212
+ state_key: null
213
+ type: *id003
214
+ delta_indices:
215
+ - 0
216
+ - 1
217
+ - 2
218
+ - 3
219
+ - 4
220
+ - 5
221
+ - 6
222
+ - 7
223
+ - 8
224
+ - 9
225
+ - 10
226
+ - 11
227
+ - 12
228
+ - 13
229
+ - 14
230
+ - 15
231
+ mean_std_embedding_keys: null
232
+ modality_keys:
233
+ - end_effector_position
234
+ - end_effector_rotation
235
+ - gripper_close
236
+ - base_motion
237
+ - control_mode
238
+ sin_cos_embedding_keys: null
239
+ language: !!python/object:gr00t.data.types.ModalityConfig
240
+ action_configs: null
241
+ delta_indices:
242
+ - 0
243
+ mean_std_embedding_keys: null
244
+ modality_keys:
245
+ - annotation.human.task_description
246
+ sin_cos_embedding_keys: null
247
+ state: !!python/object:gr00t.data.types.ModalityConfig
248
+ action_configs: null
249
+ delta_indices:
250
+ - 0
251
+ mean_std_embedding_keys: null
252
+ modality_keys:
253
+ - end_effector_position_relative
254
+ - end_effector_rotation_relative
255
+ - gripper_qpos
256
+ - base_position
257
+ - base_rotation
258
+ sin_cos_embedding_keys: null
259
+ video: !!python/object:gr00t.data.types.ModalityConfig
260
+ action_configs: null
261
+ delta_indices:
262
+ - 0
263
+ mean_std_embedding_keys: null
264
+ modality_keys:
265
+ - res256_image_side_0
266
+ - res256_image_side_1
267
+ - res256_image_wrist_0
268
+ sin_cos_embedding_keys: null
269
+ mode: single_turn
270
+ multiprocessing_context: fork
271
+ num_shards_per_epoch: 100000
272
+ override_pretraining_statistics: false
273
+ random_chop: 0.0
274
+ seed: 42
275
+ shard_size: 1024
276
+ shuffle: true
277
+ subsample_ratio: 1.0
278
+ video_backend: torchcodec
279
+ load_config_path: null
280
+ model: !!python/object:gr00t.configs.model.gr00t_n1d6.Gr00tN1d6Config
281
+ _attn_implementation_autoset: false
282
+ _attn_implementation_internal: null
283
+ _commit_hash: null
284
+ _name_or_path: ''
285
+ add_cross_attention: false
286
+ architectures: null
287
+ backbone_model_type: eagle
288
+ backbone_trainable_params_fp32: true
289
+ bad_words_ids: null
290
+ begin_suppress_tokens: null
291
+ bos_token_id: null
292
+ chunk_size_feed_forward: 0
293
+ color_jitter_params: null
294
+ cross_attention_hidden_size: null
295
+ decoder_start_token_id: null
296
+ diffusion_model_cfg:
297
+ attention_head_dim: 48
298
+ dropout: 0.2
299
+ final_dropout: true
300
+ interleave_self_attention: true
301
+ norm_type: ada_norm
302
+ num_attention_heads: 32
303
+ num_layers: 32
304
+ output_dim: 1024
305
+ positional_embeddings: null
306
+ diversity_penalty: 0.0
307
+ do_sample: false
308
+ eagle_collator: true
309
+ early_stopping: false
310
+ encoder_no_repeat_ngram_size: 0
311
+ eos_token_id: null
312
+ exponential_decay_length_penalty: null
313
+ finetuning_task: null
314
+ forced_bos_token_id: null
315
+ forced_eos_token_id: null
316
+ id2label:
317
+ 0: LABEL_0
318
+ 1: LABEL_1
319
+ is_decoder: false
320
+ is_encoder_decoder: false
321
+ label2id:
322
+ LABEL_0: 0
323
+ LABEL_1: 1
324
+ length_penalty: 1.0
325
+ load_bf16: false
326
+ max_length: 20
327
+ min_length: 0
328
+ model_name: nvidia/Eagle-Block2A-2B-v2
329
+ no_repeat_ngram_size: 0
330
+ num_beam_groups: 1
331
+ num_beams: 1
332
+ num_return_sequences: 1
333
+ output_attentions: false
334
+ output_hidden_states: false
335
+ output_scores: false
336
+ pad_token_id: null
337
+ prefix: null
338
+ problem_type: null
339
+ pruned_heads: {}
340
+ random_rotation_angle: null
341
+ remove_invalid_values: false
342
+ repetition_penalty: 1.0
343
+ reproject_vision: false
344
+ return_dict: true
345
+ return_dict_in_generate: false
346
+ sep_token_id: null
347
+ state_dropout_prob: 0.0
348
+ suppress_tokens: null
349
+ task_specific_params: null
350
+ temperature: 1.0
351
+ tf_legacy_loss: false
352
+ tie_encoder_decoder: false
353
+ tie_word_embeddings: true
354
+ tokenizer_class: null
355
+ top_k: 50
356
+ top_p: 1.0
357
+ torch_dtype: null
358
+ torchscript: false
359
+ transformers_version: null
360
+ tune_diffusion_model: true
361
+ tune_llm: false
362
+ tune_projector: true
363
+ tune_visual: false
364
+ typical_p: 1.0
365
+ use_bfloat16: false
366
+ use_relative_action: true
367
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
368
+ add_rl_callback: false
369
+ assert_loss_less_than: null
370
+ batch_size: null
371
+ bf16: true
372
+ dataloader_num_workers: 4
373
+ ddp_bucket_cap_mb: 100
374
+ deepspeed_stage: 2
375
+ enable_open_loop_eval: false
376
+ enable_profiling: false
377
+ eval_batch_size: 2
378
+ eval_bf16: true
379
+ eval_set_split_ratio: 0.1
380
+ eval_steps: 500
381
+ eval_strategy: 'no'
382
+ experiment_name: null
383
+ fp16: false
384
+ global_batch_size: 64
385
+ gradient_accumulation_steps: 1
386
+ gradient_checkpointing: false
387
+ learning_rate: 0.0001
388
+ logging_steps: 10
389
+ lr_scheduler_type: cosine
390
+ max_concurrent_uploads: 2
391
+ max_grad_norm: 1.0
392
+ max_retries: 3
393
+ max_steps: 50000
394
+ num_gpus: 8
395
+ open_loop_eval_plot_indices: null
396
+ open_loop_eval_steps_per_traj: 100
397
+ open_loop_eval_traj_ids:
398
+ - 0
399
+ optim: adamw_torch
400
+ output_dir: /root/outputs/gr00t-all-tasks
401
+ remove_unused_columns: false
402
+ save_best_eval_metric_greater_is_better: true
403
+ save_best_eval_metric_name: ''
404
+ save_steps: 2500
405
+ save_total_limit: 5
406
+ save_vl_model: false
407
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
408
+ tf32: true
409
+ transformers_access_token: null
410
+ transformers_cache_dir: null
411
+ transformers_local_files_only: false
412
+ transformers_trust_remote_code: true
413
+ upload_checkpoints: false
414
+ upload_every: 1000
415
+ upload_last_n_checkpoints: 5
416
+ use_ddp: false
417
+ use_wandb: false
418
+ wandb_project: finetune-gr00t-n1d6
419
+ warmup_ratio: 0.05
420
+ warmup_steps: 0
421
+ 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
+ -4.821934223175049,
7
+ -6.890198230743408,
8
+ 0.6994727253913879
9
+ ],
10
+ "max": [
11
+ 7.312802314758301,
12
+ 0.4876587688922882,
13
+ 0.7223060727119446
14
+ ],
15
+ "mean": [
16
+ 2.632137834062483,
17
+ -2.1012502467089176,
18
+ 0.7004640659262085
19
+ ],
20
+ "std": [
21
+ 1.6703172900404137,
22
+ 1.3527759463804145,
23
+ 0.0011049435086240903
24
+ ],
25
+ "q01": [
26
+ -4.570115985870362,
27
+ -6.890125727653503,
28
+ 0.6998056852817536
29
+ ],
30
+ "q99": [
31
+ 7.312699794769287,
32
+ 0.48759907245635986,
33
+ 0.7159942960739136
34
+ ]
35
+ },
36
+ "end_effector_rotation_relative": {
37
+ "min": [
38
+ 0.0,
39
+ 0.0,
40
+ -1.0,
41
+ 0.0
42
+ ],
43
+ "max": [
44
+ 0.0,
45
+ 0.0,
46
+ 1.0,
47
+ 1.0
48
+ ],
49
+ "mean": [
50
+ 0.0,
51
+ 0.0,
52
+ 0.24719667310306095,
53
+ 0.604356052193361
54
+ ],
55
+ "std": [
56
+ 0.0,
57
+ 0.0,
58
+ 0.6612804897677296,
59
+ 0.36922738359963947
60
+ ],
61
+ "q01": [
62
+ 0.0,
63
+ 0.0,
64
+ -1.0,
65
+ 1.4936076922822391e-07
66
+ ],
67
+ "q99": [
68
+ 0.0,
69
+ 0.0,
70
+ 1.0,
71
+ 1.0
72
+ ]
73
+ },
74
+ "gripper_qpos": {
75
+ "min": [
76
+ -0.5828549861907959,
77
+ -0.8486236333847046
78
+ ],
79
+ "max": [
80
+ 0.8984705209732056,
81
+ 0.8276846408843994
82
+ ],
83
+ "mean": [
84
+ 0.2756862304733528,
85
+ -0.01928894071868549
86
+ ],
87
+ "std": [
88
+ 0.16439036890222622,
89
+ 0.2507434896484107
90
+ ],
91
+ "q01": [
92
+ -0.38679012656211853,
93
+ -0.737857882976532
94
+ ],
95
+ "q99": [
96
+ 0.8634055067216393,
97
+ 0.7653709030151367
98
+ ]
99
+ },
100
+ "base_position": {
101
+ "min": [
102
+ -0.3446122705936432,
103
+ -0.9999930262565613,
104
+ -0.9999884366989136
105
+ ],
106
+ "max": [
107
+ 1.0538847019960584,
108
+ 0.9999998807907104,
109
+ 0.9997354745864868
110
+ ],
111
+ "mean": [
112
+ 0.4386128584770705,
113
+ -0.236187959678786,
114
+ 0.03440304100469987
115
+ ],
116
+ "std": [
117
+ 0.2138587124612812,
118
+ 0.8254625564899771,
119
+ 0.30061384229178
120
+ ],
121
+ "q01": [
122
+ -0.2664716064929962,
123
+ -0.999696980714798,
124
+ -0.9779104268550873
125
+ ],
126
+ "q99": [
127
+ 0.9607213485847528,
128
+ 0.9996132826805115,
129
+ 0.983684875965118
130
+ ]
131
+ },
132
+ "base_rotation": {
133
+ "min": [
134
+ -0.9975369572639465,
135
+ 6.754255821306288e-08,
136
+ -0.011436971835792065,
137
+ -0.06307756155729294
138
+ ],
139
+ "max": [
140
+ 0.9900747537612915,
141
+ 0.9735639095306396,
142
+ 0.05586916867422667,
143
+ 0.010916369967162609
144
+ ],
145
+ "mean": [
146
+ -0.07227329122419625,
147
+ 0.14411132909092314,
148
+ 0.03140565182425891,
149
+ -0.0316361944376031
150
+ ],
151
+ "std": [
152
+ 0.34065961750646406,
153
+ 0.17097897402192125,
154
+ 0.012491187867591168,
155
+ 0.012261243048427888
156
+ ],
157
+ "q01": [
158
+ -0.9171609520912171,
159
+ 0.00014284539400250652,
160
+ -0.0034570247586816548,
161
+ -0.05656256314543609
162
+ ],
163
+ "q99": [
164
+ 0.9476015937328336,
165
+ 0.9460982489585876,
166
+ 0.05580083291027281,
167
+ 0.0037876180186867726
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
+ 0.003215061227666985,
185
+ 0.0015200642241156692,
186
+ -0.0011421955591327765
187
+ ],
188
+ "std": [
189
+ 0.09016256759823495,
190
+ 0.08451339450696209,
191
+ 0.06564718888397388
192
+ ],
193
+ "q01": [
194
+ -1.0,
195
+ -1.0,
196
+ -0.9657142758369446
197
+ ],
198
+ "q99": [
199
+ 1.0,
200
+ 1.0,
201
+ 1.0
202
+ ]
203
+ },
204
+ "end_effector_rotation": {
205
+ "min": [
206
+ 0.0,
207
+ -1.0,
208
+ -1.0
209
+ ],
210
+ "max": [
211
+ 0.0,
212
+ 1.0,
213
+ 1.0
214
+ ],
215
+ "mean": [
216
+ 0.0,
217
+ -0.9463277607073037,
218
+ -0.00632101623049963
219
+ ],
220
+ "std": [
221
+ 0.0,
222
+ 0.3232245701712277,
223
+ 0.4491370524676613
224
+ ],
225
+ "q01": [
226
+ 0.0,
227
+ -1.0,
228
+ -1.0
229
+ ],
230
+ "q99": [
231
+ 0.0,
232
+ 1.0,
233
+ 1.0
234
+ ]
235
+ },
236
+ "gripper_close": {
237
+ "min": [
238
+ -1.0
239
+ ],
240
+ "max": [
241
+ 1.0
242
+ ],
243
+ "mean": [
244
+ -0.0057538288612615255
245
+ ],
246
+ "std": [
247
+ 0.4104968917775479
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
+ -1.0
262
+ ],
263
+ "max": [
264
+ 1.0,
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "mean": [
270
+ -0.06451660020185172,
271
+ 0.007555278668096627,
272
+ -0.023032587824433987,
273
+ 0.0011521431231020069
274
+ ],
275
+ "std": [
276
+ 0.4041732707681895,
277
+ 0.11050048169521205,
278
+ 0.1305818519167956,
279
+ 0.12145431652616569
280
+ ],
281
+ "q01": [
282
+ -1.0,
283
+ -0.8114285469055176,
284
+ -1.0,
285
+ -1.0
286
+ ],
287
+ "q99": [
288
+ 1.0,
289
+ 1.0,
290
+ 1.0,
291
+ 1.0
292
+ ]
293
+ },
294
+ "control_mode": {
295
+ "min": [
296
+ -1.0
297
+ ],
298
+ "max": [
299
+ 1.0
300
+ ],
301
+ "mean": [
302
+ -0.264296489609035
303
+ ],
304
+ "std": [
305
+ 0.9644378227331523
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