Hugging Face's logo

pmoller
/
foca_40_40k

Safetensors
Model card Files
xet
 Community
pmoller commited on
Commit
81ac704
·
verified ·
1 Parent(s): da2215c

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +81 -0
  2. model.safetensors +3 -0
  3. train_config.json +274 -0
config.json ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "front_image": {
11
+ "type": "VISUAL",
12
+ "shape": [
13
+ 3,
14
+ 480,
15
+ 640
16
+ ]
17
+ },
18
+ "right_image": {
19
+ "type": "VISUAL",
20
+ "shape": [
21
+ 3,
22
+ 480,
23
+ 640
24
+ ]
25
+ },
26
+ "wrist_image": {
27
+ "type": "VISUAL",
28
+ "shape": [
29
+ 3,
30
+ 480,
31
+ 640
32
+ ]
33
+ },
34
+ "state": {
35
+ "type": "STATE",
36
+ "shape": [
37
+ 7
38
+ ]
39
+ }
40
+ },
41
+ "output_features": {
42
+ "actions": {
43
+ "type": "ACTION",
44
+ "shape": [
45
+ 7
46
+ ]
47
+ }
48
+ },
49
+ "device": "cuda",
50
+ "use_amp": false,
51
+ "gradient_accumulation_steps": 6,
52
+ "chunk_size": 50,
53
+ "n_action_steps": 50,
54
+ "max_state_dim": 32,
55
+ "max_action_dim": 32,
56
+ "resize_imgs_with_padding": [
57
+ 224,
58
+ 224
59
+ ],
60
+ "empty_cameras": 0,
61
+ "adapt_to_pi_aloha": false,
62
+ "use_delta_joint_actions_aloha": false,
63
+ "tokenizer_max_length": 48,
64
+ "proj_width": 1024,
65
+ "num_steps": 10,
66
+ "use_cache": true,
67
+ "attention_implementation": "eager",
68
+ "freeze_vision_encoder": true,
69
+ "train_expert_only": false,
70
+ "train_state_proj": true,
71
+ "optimizer_lr": 0.0001,
72
+ "optimizer_betas": [
73
+ 0.9,
74
+ 0.95
75
+ ],
76
+ "optimizer_eps": 1e-08,
77
+ "optimizer_weight_decay": 1e-10,
78
+ "scheduler_warmup_steps": 1000,
79
+ "scheduler_decay_steps": 300000,
80
+ "scheduler_decay_lr": 2.5e-06
81
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76943d141982450ab59befe40355e6b8833a92dcf2a1a38cffb65a7d58c250f3
3
+ size 7536022552
train_config.json ADDED
@@ -0,0 +1,274 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": ".",
4
+ "root": "/dtu/p1/plund/FOCA/data/ur5_lerobot_foca_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
+ "sharpness": {
112
+ "weight": 1.0,
113
+ "type": "SharpnessJitter",
114
+ "kwargs": {
115
+ "sharpness": [
116
+ 0.5,
117
+ 1.5
118
+ ]
119
+ }
120
+ },
121
+ "brightness": {
122
+ "weight": 1.0,
123
+ "type": "ColorJitter",
124
+ "kwargs": {
125
+ "brightness": [
126
+ 0.8,
127
+ 1.2
128
+ ]
129
+ }
130
+ },
131
+ "saturation": {
132
+ "weight": 1.0,
133
+ "type": "ColorJitter",
134
+ "kwargs": {
135
+ "saturation": [
136
+ 0.5,
137
+ 1.5
138
+ ]
139
+ }
140
+ }
141
+ }
142
+ },
143
+ "revision": null,
144
+ "use_imagenet_stats": true,
145
+ "video_backend": "torchcodec",
146
+ "vqa_data_path": null
147
+ },
148
+ "env": null,
149
+ "policy": {
150
+ "type": "pi0",
151
+ "n_obs_steps": 1,
152
+ "normalization_mapping": {
153
+ "VISUAL": "IDENTITY",
154
+ "STATE": "MEAN_STD",
155
+ "ACTION": "MEAN_STD"
156
+ },
157
+ "input_features": {
158
+ "front_image": {
159
+ "type": "VISUAL",
160
+ "shape": [
161
+ 3,
162
+ 480,
163
+ 640
164
+ ]
165
+ },
166
+ "right_image": {
167
+ "type": "VISUAL",
168
+ "shape": [
169
+ 3,
170
+ 480,
171
+ 640
172
+ ]
173
+ },
174
+ "wrist_image": {
175
+ "type": "VISUAL",
176
+ "shape": [
177
+ 3,
178
+ 480,
179
+ 640
180
+ ]
181
+ },
182
+ "state": {
183
+ "type": "STATE",
184
+ "shape": [
185
+ 7
186
+ ]
187
+ }
188
+ },
189
+ "output_features": {
190
+ "actions": {
191
+ "type": "ACTION",
192
+ "shape": [
193
+ 7
194
+ ]
195
+ }
196
+ },
197
+ "device": "cuda",
198
+ "use_amp": false,
199
+ "gradient_accumulation_steps": 6,
200
+ "chunk_size": 50,
201
+ "n_action_steps": 50,
202
+ "max_state_dim": 32,
203
+ "max_action_dim": 32,
204
+ "resize_imgs_with_padding": [
205
+ 224,
206
+ 224
207
+ ],
208
+ "empty_cameras": 0,
209
+ "adapt_to_pi_aloha": false,
210
+ "use_delta_joint_actions_aloha": false,
211
+ "tokenizer_max_length": 48,
212
+ "proj_width": 1024,
213
+ "num_steps": 10,
214
+ "use_cache": true,
215
+ "attention_implementation": "eager",
216
+ "freeze_vision_encoder": true,
217
+ "train_expert_only": false,
218
+ "train_state_proj": true,
219
+ "optimizer_lr": 0.0001,
220
+ "optimizer_betas": [
221
+ 0.9,
222
+ 0.95
223
+ ],
224
+ "optimizer_eps": 1e-08,
225
+ "optimizer_weight_decay": 1e-10,
226
+ "scheduler_warmup_steps": 1000,
227
+ "scheduler_decay_steps": 300000,
228
+ "scheduler_decay_lr": 2.5e-06
229
+ },
230
+ "output_dir": "/dtu/p1/plund/FOCA/outputs/train/2026-01-27/10-47-40_pi0_base_ur5e_hpc_40",
231
+ "job_name": "pi0_base_ur5e_hpc_40",
232
+ "resume": false,
233
+ "seed": 42,
234
+ "num_workers": 8,
235
+ "batch_size": 2,
236
+ "steps": 100000,
237
+ "eval_freq": 20000,
238
+ "log_freq": 10,
239
+ "save_checkpoint": true,
240
+ "save_freq": 10000,
241
+ "use_policy_training_preset": true,
242
+ "optimizer": {
243
+ "type": "adamw",
244
+ "lr": 0.0001,
245
+ "weight_decay": 1e-10,
246
+ "grad_clip_norm": 10.0,
247
+ "betas": [
248
+ 0.9,
249
+ 0.95
250
+ ],
251
+ "eps": 1e-08
252
+ },
253
+ "scheduler": {
254
+ "type": "cosine_decay_with_warmup",
255
+ "num_warmup_steps": 1000,
256
+ "num_decay_steps": 300000,
257
+ "peak_lr": 0.0001,
258
+ "decay_lr": 2.5e-06
259
+ },
260
+ "eval": {
261
+ "n_episodes": 50,
262
+ "batch_size": 50,
263
+ "use_async_envs": false
264
+ },
265
+ "wandb": {
266
+ "enable": true,
267
+ "disable_artifact": true,
268
+ "project": "pi0_lerobot",
269
+ "entity": "plund-danmarks-tekniske-universitet-dtu",
270
+ "notes": null,
271
+ "run_id": null,
272
+ "mode": "online"
273
+ }
274
+ }