Hugging Face's logo

binhng
/
FOCA_aloha_140k_scale40

Safetensors
Model card Files
xet
 Community
binhng commited on
Commit
68f5e5f
·
verified ·
1 Parent(s): 95f745e

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. pretrained_model/config.json +99 -0
  2. pretrained_model/model.safetensors +3 -0
  3. pretrained_model/train_config.json +329 -0
pretrained_model/config.json ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "pi0",
3
+ "n_obs_steps": 1,
4
+ "normalization_mapping": {
5
+ "VISUAL": "IDENTITY",
6
+ "STATE": "MEAN_STD",
7
+ "ACTION": "MEAN_STD"
8
+ },
9
+ "input_features": {
10
+ "observation.state": {
11
+ "type": "STATE",
12
+ "shape": [
13
+ 14
14
+ ]
15
+ },
16
+ "observation.images.color.high": {
17
+ "type": "VISUAL",
18
+ "shape": [
19
+ 3,
20
+ 480,
21
+ 640
22
+ ]
23
+ },
24
+ "observation.images.color.wrist_right": {
25
+ "type": "VISUAL",
26
+ "shape": [
27
+ 3,
28
+ 480,
29
+ 640
30
+ ]
31
+ },
32
+ "observation.images.color.wrist_left": {
33
+ "type": "VISUAL",
34
+ "shape": [
35
+ 3,
36
+ 480,
37
+ 640
38
+ ]
39
+ }
40
+ },
41
+ "output_features": {
42
+ "action": {
43
+ "type": "ACTION",
44
+ "shape": [
45
+ 14
46
+ ]
47
+ }
48
+ },
49
+ "device": "cuda",
50
+ "use_amp": false,
51
+ "gradient_accumulation_steps": 2,
52
+ "chunk_size": 50,
53
+ "n_action_steps": 50,
54
+ "visual_prompt": false,
55
+ "use_both_origin": false,
56
+ "zero_init": false,
57
+ "coordinate_text_prod": false,
58
+ "coordinate_posEmbed_prod": false,
59
+ "posEmbed_to_preTrained": false,
60
+ "num_obs_layers": 30,
61
+ "last_obs_layers": false,
62
+ "add_posEmbed": false,
63
+ "not_used_coord": false,
64
+ "use_depth": false,
65
+ "depth_model_type": "frozen_siglip",
66
+ "use_track": true,
67
+ "use_detect_object": false,
68
+ "track_mask": false,
69
+ "use_slot_att": "none",
70
+ "future_obj": "all",
71
+ "original_pi0": false,
72
+ "max_state_dim": 32,
73
+ "max_action_dim": 32,
74
+ "resize_imgs_with_padding": [
75
+ 224,
76
+ 224
77
+ ],
78
+ "empty_cameras": 0,
79
+ "adapt_to_pi_aloha": false,
80
+ "use_delta_joint_actions_aloha": false,
81
+ "tokenizer_max_length": 48,
82
+ "proj_width": 1024,
83
+ "num_steps": 10,
84
+ "use_cache": true,
85
+ "attention_implementation": "eager",
86
+ "freeze_vision_encoder": true,
87
+ "train_expert_only": false,
88
+ "train_state_proj": true,
89
+ "optimizer_lr": 2.5e-06,
90
+ "optimizer_betas": [
91
+ 0.9,
92
+ 0.95
93
+ ],
94
+ "optimizer_eps": 1e-08,
95
+ "optimizer_weight_decay": 1e-10,
96
+ "scheduler_warmup_steps": 1000,
97
+ "scheduler_decay_steps": 240000,
98
+ "scheduler_decay_lr": 2.5e-06
99
+ }
pretrained_model/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f78169d30d29acb632cf1dbdb1ae388e1fc8d50a868075640d40d15cab5a3809
3
+ size 7791670388
pretrained_model/train_config.json ADDED
@@ -0,0 +1,329 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "ducido/calvin_task_D_D_scale_100_lerobo_format",
4
+ "root": "/mnt/data/sftp/data/vla_intern/workspace/data/realworld/aloha_3_tasks_large_40",
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": true,
8
+ "max_num_transforms": 3,
9
+ "random_order": false,
10
+ "image_tfs": {
11
+ "hue": {
12
+ "weight": 1.0,
13
+ "type": "ColorJitter",
14
+ "kwargs": {
15
+ "hue": [
16
+ -0.05,
17
+ 0.05
18
+ ]
19
+ }
20
+ },
21
+ "contrast": {
22
+ "weight": 1.0,
23
+ "type": "ColorJitter",
24
+ "kwargs": {
25
+ "contrast": [
26
+ 0.8,
27
+ 1.2
28
+ ]
29
+ }
30
+ },
31
+ "sharpness": {
32
+ "weight": 1.0,
33
+ "type": "SharpnessJitter",
34
+ "kwargs": {
35
+ "sharpness": [
36
+ 0.5,
37
+ 1.5
38
+ ]
39
+ }
40
+ },
41
+ "brightness": {
42
+ "weight": 1.0,
43
+ "type": "ColorJitter",
44
+ "kwargs": {
45
+ "brightness": [
46
+ 0.8,
47
+ 1.2
48
+ ]
49
+ }
50
+ },
51
+ "saturation": {
52
+ "weight": 1.0,
53
+ "type": "ColorJitter",
54
+ "kwargs": {
55
+ "saturation": [
56
+ 0.5,
57
+ 1.5
58
+ ]
59
+ }
60
+ },
61
+ "crop_resize": {
62
+ "weight": 1.0,
63
+ "type": "RandomResizedCrop",
64
+ "kwargs": {
65
+ "size": [
66
+ 256,
67
+ 256
68
+ ],
69
+ "ratio": [
70
+ 1,
71
+ 1
72
+ ],
73
+ "scale": [
74
+ 0.9,
75
+ 0.95
76
+ ]
77
+ }
78
+ },
79
+ "rotate": {
80
+ "weight": 1.0,
81
+ "type": "RandomRotate",
82
+ "kwargs": {
83
+ "degrees": [
84
+ -5,
85
+ 5
86
+ ]
87
+ }
88
+ }
89
+ },
90
+ "wrist_tfs": {
91
+ "hue": {
92
+ "weight": 1.0,
93
+ "type": "ColorJitter",
94
+ "kwargs": {
95
+ "hue": [
96
+ -0.05,
97
+ 0.05
98
+ ]
99
+ }
100
+ },
101
+ "contrast": {
102
+ "weight": 1.0,
103
+ "type": "ColorJitter",
104
+ "kwargs": {
105
+ "contrast": [
106
+ 0.8,
107
+ 1.2
108
+ ]
109
+ }
110
+ },
111
+ "crop_resize": {
112
+ "weight": 1.0,
113
+ "type": "RandomResizedCrop",
114
+ "kwargs": {
115
+ "size": [
116
+ 256,
117
+ 256
118
+ ],
119
+ "ratio": [
120
+ 1,
121
+ 1
122
+ ],
123
+ "scale": [
124
+ 0.9,
125
+ 0.95
126
+ ]
127
+ }
128
+ },
129
+ "sharpness": {
130
+ "weight": 1.0,
131
+ "type": "SharpnessJitter",
132
+ "kwargs": {
133
+ "sharpness": [
134
+ 0.5,
135
+ 1.5
136
+ ]
137
+ }
138
+ },
139
+ "brightness": {
140
+ "weight": 1.0,
141
+ "type": "ColorJitter",
142
+ "kwargs": {
143
+ "brightness": [
144
+ 0.8,
145
+ 1.2
146
+ ]
147
+ }
148
+ },
149
+ "saturation": {
150
+ "weight": 1.0,
151
+ "type": "ColorJitter",
152
+ "kwargs": {
153
+ "saturation": [
154
+ 0.5,
155
+ 1.5
156
+ ]
157
+ }
158
+ }
159
+ }
160
+ },
161
+ "revision": null,
162
+ "use_imagenet_stats": true,
163
+ "video_backend": "torchcodec",
164
+ "vqa_data_path": null
165
+ },
166
+ "env": null,
167
+ "policy": {
168
+ "type": "pi0",
169
+ "n_obs_steps": 1,
170
+ "normalization_mapping": {
171
+ "VISUAL": "IDENTITY",
172
+ "STATE": "MEAN_STD",
173
+ "ACTION": "MEAN_STD"
174
+ },
175
+ "input_features": {
176
+ "observation.state": {
177
+ "type": "STATE",
178
+ "shape": [
179
+ 14
180
+ ]
181
+ },
182
+ "observation.images.color.high": {
183
+ "type": "VISUAL",
184
+ "shape": [
185
+ 3,
186
+ 480,
187
+ 640
188
+ ]
189
+ },
190
+ "observation.images.color.wrist_right": {
191
+ "type": "VISUAL",
192
+ "shape": [
193
+ 3,
194
+ 480,
195
+ 640
196
+ ]
197
+ },
198
+ "observation.images.color.wrist_left": {
199
+ "type": "VISUAL",
200
+ "shape": [
201
+ 3,
202
+ 480,
203
+ 640
204
+ ]
205
+ }
206
+ },
207
+ "output_features": {
208
+ "action": {
209
+ "type": "ACTION",
210
+ "shape": [
211
+ 14
212
+ ]
213
+ }
214
+ },
215
+ "device": "cuda",
216
+ "use_amp": false,
217
+ "gradient_accumulation_steps": 2,
218
+ "chunk_size": 50,
219
+ "n_action_steps": 50,
220
+ "visual_prompt": false,
221
+ "use_both_origin": false,
222
+ "zero_init": false,
223
+ "coordinate_text_prod": false,
224
+ "coordinate_posEmbed_prod": false,
225
+ "posEmbed_to_preTrained": false,
226
+ "num_obs_layers": 30,
227
+ "last_obs_layers": false,
228
+ "add_posEmbed": false,
229
+ "not_used_coord": false,
230
+ "use_depth": false,
231
+ "depth_model_type": "frozen_siglip",
232
+ "use_track": true,
233
+ "use_detect_object": false,
234
+ "track_mask": false,
235
+ "use_slot_att": "none",
236
+ "future_obj": "all",
237
+ "original_pi0": false,
238
+ "max_state_dim": 32,
239
+ "max_action_dim": 32,
240
+ "resize_imgs_with_padding": [
241
+ 224,
242
+ 224
243
+ ],
244
+ "empty_cameras": 0,
245
+ "adapt_to_pi_aloha": false,
246
+ "use_delta_joint_actions_aloha": false,
247
+ "tokenizer_max_length": 48,
248
+ "proj_width": 1024,
249
+ "num_steps": 10,
250
+ "use_cache": true,
251
+ "attention_implementation": "eager",
252
+ "freeze_vision_encoder": true,
253
+ "train_expert_only": false,
254
+ "train_state_proj": true,
255
+ "optimizer_lr": 2.5e-06,
256
+ "optimizer_betas": [
257
+ 0.9,
258
+ 0.95
259
+ ],
260
+ "optimizer_eps": 1e-08,
261
+ "optimizer_weight_decay": 1e-10,
262
+ "scheduler_warmup_steps": 1000,
263
+ "scheduler_decay_steps": 240000,
264
+ "scheduler_decay_lr": 2.5e-06
265
+ },
266
+ "output_dir": "outputs/train/2026-01-24/19-17-07_FOCA_pi0_aloha_40_v1_continue",
267
+ "job_name": "FOCA_pi0_aloha_40_v1_continue",
268
+ "visual_prompt": false,
269
+ "use_both_origin": false,
270
+ "zero_init": false,
271
+ "coordinate_text_prod": false,
272
+ "coordinate_posEmbed_prod": false,
273
+ "posEmbed_to_preTrained": false,
274
+ "num_obs_layers": 30,
275
+ "last_obs_layers": false,
276
+ "add_posEmbed": false,
277
+ "not_used_coord": false,
278
+ "train_expert_only": false,
279
+ "use_depth": false,
280
+ "use_track": true,
281
+ "use_detect_object": false,
282
+ "track_mask": false,
283
+ "use_slot_att": "none",
284
+ "future_obj": "all",
285
+ "depth_model_type": "frozen_siglip",
286
+ "original_pi0": false,
287
+ "resume": false,
288
+ "seed": 42,
289
+ "num_workers": 8,
290
+ "batch_size": 12,
291
+ "steps": 50000,
292
+ "eval_freq": 20000,
293
+ "log_freq": 200,
294
+ "save_checkpoint": true,
295
+ "save_freq": 10000,
296
+ "use_policy_training_preset": true,
297
+ "optimizer": {
298
+ "type": "adamw",
299
+ "lr": 2.5e-06,
300
+ "weight_decay": 1e-10,
301
+ "grad_clip_norm": 10.0,
302
+ "betas": [
303
+ 0.9,
304
+ 0.95
305
+ ],
306
+ "eps": 1e-08
307
+ },
308
+ "scheduler": {
309
+ "type": "cosine_decay_with_warmup",
310
+ "num_warmup_steps": 1000,
311
+ "num_decay_steps": 240000,
312
+ "peak_lr": 2.5e-06,
313
+ "decay_lr": 2.5e-06
314
+ },
315
+ "eval": {
316
+ "n_episodes": 50,
317
+ "batch_size": 50,
318
+ "use_async_envs": false
319
+ },
320
+ "wandb": {
321
+ "enable": true,
322
+ "disable_artifact": true,
323
+ "project": "pi0_lerobot",
324
+ "entity": "Robotics_VLA",
325
+ "notes": null,
326
+ "run_id": null,
327
+ "mode": "online"
328
+ }
329
+ }