Hugging Face's logo

chomeed
/
rewind_peg_insertion_v2

Robotics
LeRobot
Safetensors
rewind
Model card Files
xet
 Community
chomeed commited on
Commit
3a248e9
·
verified ·
1 Parent(s): 3c2bd63

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +90 -0
  3. model.safetensors +3 -0
  4. train_config.json +385 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: chomeed/peg_insertion_full_500_unprivileged_fixed_v2_lerobot_success_for_rewind
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: rewind
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - rewind
9
+ - lerobot
10
+ - robotics
11
+ ---
12
+
13
+ # Model Card for rewind
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+ _Model type not recognized — please update this template._
19
+
20
+
21
+ This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
22
+ See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
23
+
24
+ ---
25
+
26
+ ## How to Get Started with the Model
27
+
28
+ For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
29
+ Below is the short version on how to train and run inference/eval:
30
+
31
+ ### Train from scratch
32
+
33
+ ```bash
34
+ lerobot-train \
35
+ --dataset.repo_id=${HF_USER}/<dataset> \
36
+ --policy.type=act \
37
+ --output_dir=outputs/train/<desired_policy_repo_id> \
38
+ --job_name=lerobot_training \
39
+ --policy.device=cuda \
40
+ --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
41
+ --wandb.enable=true
42
+ ```
43
+
44
+ _Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
45
+
46
+ ### Evaluate the policy/run inference
47
+
48
+ ```bash
49
+ lerobot-record \
50
+ --robot.type=so100_follower \
51
+ --dataset.repo_id=<hf_user>/eval_<dataset> \
52
+ --policy.path=<hf_user>/<desired_policy_repo_id> \
53
+ --episodes=10
54
+ ```
55
+
56
+ Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
57
+
58
+ ---
59
+
60
+ ## Model Details
61
+
62
+ - **License:** apache-2.0
config.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "rewind",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.images.front_rgb": {
6
+ "type": "VISUAL",
7
+ "shape": [
8
+ 3,
9
+ 256,
10
+ 256
11
+ ]
12
+ },
13
+ "observation.images.wrist_rgb": {
14
+ "type": "VISUAL",
15
+ "shape": [
16
+ 3,
17
+ 256,
18
+ 256
19
+ ]
20
+ },
21
+ "observation.state": {
22
+ "type": "STATE",
23
+ "shape": [
24
+ 19
25
+ ]
26
+ },
27
+ "observation.dinov2.front_rgb": {
28
+ "type": "STATE",
29
+ "shape": [
30
+ 768
31
+ ]
32
+ },
33
+ "observation.dinov2.wrist_rgb": {
34
+ "type": "STATE",
35
+ "shape": [
36
+ 768
37
+ ]
38
+ }
39
+ },
40
+ "output_features": {
41
+ "action": {
42
+ "type": "ACTION",
43
+ "shape": [
44
+ 6
45
+ ]
46
+ }
47
+ },
48
+ "device": "cuda",
49
+ "use_amp": false,
50
+ "use_peft": false,
51
+ "push_to_hub": true,
52
+ "repo_id": "chomeed/rewind_peg_insertion_v2",
53
+ "private": null,
54
+ "tags": null,
55
+ "license": null,
56
+ "pretrained_path": null,
57
+ "d_model": 512,
58
+ "vis_emb_dim": 768,
59
+ "text_emb_dim": 384,
60
+ "n_layers": 4,
61
+ "n_heads": 8,
62
+ "dropout": 0.1,
63
+ "num_cameras": 2,
64
+ "finetune_encoder": true,
65
+ "finetune_encoder_blocks": 2,
66
+ "encoder_ckpt": "facebook/dinov2-base",
67
+ "encoder_lr": 1e-05,
68
+ "encoder_camera_keys": [
69
+ "observation.images.front_rgb",
70
+ "observation.images.wrist_rgb"
71
+ ],
72
+ "optimizer_lr": 0.0001,
73
+ "optimizer_betas": [
74
+ 0.9,
75
+ 0.95
76
+ ],
77
+ "optimizer_eps": 1e-08,
78
+ "optimizer_weight_decay": 1e-10,
79
+ "optimizer_grad_clip_norm": 10.0,
80
+ "scheduler_warmup_steps": 1000,
81
+ "scheduler_decay_steps": 10000,
82
+ "scheduler_decay_lr": 2.5e-06,
83
+ "scheduler_peak_lr": 5e-05,
84
+ "normalization_mapping": {
85
+ "VISUAL": "IDENTITY",
86
+ "STATE": "MEAN_STD",
87
+ "LANGUAGE": "IDENTITY",
88
+ "REWARD": "IDENTITY"
89
+ }
90
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a7d379406a56d66f3e3b1f29c9a9b43aa37b04d535e9fca752b7adb4c9443dc
3
+ size 402321044
train_config.json ADDED
@@ -0,0 +1,385 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "chomeed/peg_insertion_full_500_unprivileged_fixed_v2_lerobot_success_for_rewind",
4
+ "root": null,
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": false,
8
+ "max_num_transforms": 3,
9
+ "random_order": false,
10
+ "tfs": {
11
+ "brightness": {
12
+ "weight": 1.0,
13
+ "type": "ColorJitter",
14
+ "kwargs": {
15
+ "brightness": [
16
+ 0.8,
17
+ 1.2
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
+ "saturation": {
32
+ "weight": 1.0,
33
+ "type": "ColorJitter",
34
+ "kwargs": {
35
+ "saturation": [
36
+ 0.5,
37
+ 1.5
38
+ ]
39
+ }
40
+ },
41
+ "hue": {
42
+ "weight": 1.0,
43
+ "type": "ColorJitter",
44
+ "kwargs": {
45
+ "hue": [
46
+ -0.05,
47
+ 0.05
48
+ ]
49
+ }
50
+ },
51
+ "sharpness": {
52
+ "weight": 1.0,
53
+ "type": "SharpnessJitter",
54
+ "kwargs": {
55
+ "sharpness": [
56
+ 0.5,
57
+ 1.5
58
+ ]
59
+ }
60
+ },
61
+ "affine": {
62
+ "weight": 1.0,
63
+ "type": "RandomAffine",
64
+ "kwargs": {
65
+ "degrees": [
66
+ -5.0,
67
+ 5.0
68
+ ],
69
+ "translate": [
70
+ 0.05,
71
+ 0.05
72
+ ]
73
+ }
74
+ }
75
+ }
76
+ },
77
+ "revision": null,
78
+ "use_imagenet_stats": true,
79
+ "video_backend": "torchcodec",
80
+ "streaming": false,
81
+ "cf_aug": false
82
+ },
83
+ "env": null,
84
+ "policy": {
85
+ "type": "rewind",
86
+ "n_obs_steps": 1,
87
+ "input_features": {
88
+ "observation.images.front_rgb": {
89
+ "type": "VISUAL",
90
+ "shape": [
91
+ 3,
92
+ 256,
93
+ 256
94
+ ]
95
+ },
96
+ "observation.images.wrist_rgb": {
97
+ "type": "VISUAL",
98
+ "shape": [
99
+ 3,
100
+ 256,
101
+ 256
102
+ ]
103
+ },
104
+ "observation.state": {
105
+ "type": "STATE",
106
+ "shape": [
107
+ 19
108
+ ]
109
+ },
110
+ "observation.dinov2.front_rgb": {
111
+ "type": "STATE",
112
+ "shape": [
113
+ 768
114
+ ]
115
+ },
116
+ "observation.dinov2.wrist_rgb": {
117
+ "type": "STATE",
118
+ "shape": [
119
+ 768
120
+ ]
121
+ }
122
+ },
123
+ "output_features": {
124
+ "action": {
125
+ "type": "ACTION",
126
+ "shape": [
127
+ 6
128
+ ]
129
+ }
130
+ },
131
+ "device": "cuda",
132
+ "use_amp": false,
133
+ "use_peft": false,
134
+ "push_to_hub": true,
135
+ "repo_id": "chomeed/rewind_peg_insertion_v2",
136
+ "private": null,
137
+ "tags": null,
138
+ "license": null,
139
+ "pretrained_path": null,
140
+ "d_model": 512,
141
+ "vis_emb_dim": 768,
142
+ "text_emb_dim": 384,
143
+ "n_layers": 4,
144
+ "n_heads": 8,
145
+ "dropout": 0.1,
146
+ "num_cameras": 2,
147
+ "finetune_encoder": true,
148
+ "finetune_encoder_blocks": 2,
149
+ "encoder_ckpt": "facebook/dinov2-base",
150
+ "encoder_lr": 1e-05,
151
+ "encoder_camera_keys": [
152
+ "observation.images.front_rgb",
153
+ "observation.images.wrist_rgb"
154
+ ],
155
+ "optimizer_lr": 0.0001,
156
+ "optimizer_betas": [
157
+ 0.9,
158
+ 0.95
159
+ ],
160
+ "optimizer_eps": 1e-08,
161
+ "optimizer_weight_decay": 1e-10,
162
+ "optimizer_grad_clip_norm": 10.0,
163
+ "scheduler_warmup_steps": 1000,
164
+ "scheduler_decay_steps": 10000,
165
+ "scheduler_decay_lr": 2.5e-06,
166
+ "scheduler_peak_lr": 5e-05,
167
+ "normalization_mapping": {
168
+ "VISUAL": "IDENTITY",
169
+ "STATE": "MEAN_STD",
170
+ "LANGUAGE": "IDENTITY",
171
+ "REWARD": "IDENTITY"
172
+ }
173
+ },
174
+ "output_dir": "outputs/train/2026-04-23/04-01-03_rewind_peg_insertion_v2",
175
+ "job_name": "rewind_peg_insertion_v2",
176
+ "resume": false,
177
+ "seed": 1000,
178
+ "num_workers": 16,
179
+ "batch_size": 16,
180
+ "steps": 10000,
181
+ "eval_freq": 1000,
182
+ "log_freq": 100,
183
+ "tolerance_s": 0.0001,
184
+ "save_checkpoint": true,
185
+ "save_freq": 1000,
186
+ "use_policy_training_preset": true,
187
+ "optimizer": {
188
+ "type": "adamw",
189
+ "lr": 0.0001,
190
+ "weight_decay": 1e-10,
191
+ "grad_clip_norm": 10.0,
192
+ "betas": [
193
+ 0.9,
194
+ 0.95
195
+ ],
196
+ "eps": 1e-08
197
+ },
198
+ "scheduler": {
199
+ "type": "cosine_decay_with_warmup",
200
+ "num_warmup_steps": 1000,
201
+ "num_decay_steps": 10000,
202
+ "peak_lr": 5e-05,
203
+ "decay_lr": 2.5e-06
204
+ },
205
+ "val": {
206
+ "repo_id": "chomeed/peg_insertion_full_500_unprivileged_fixed_v2_lerobot_success_for_rewind",
207
+ "root": null,
208
+ "episodes": null,
209
+ "image_transforms": {
210
+ "enable": false,
211
+ "max_num_transforms": 3,
212
+ "random_order": false,
213
+ "tfs": {
214
+ "brightness": {
215
+ "weight": 1.0,
216
+ "type": "ColorJitter",
217
+ "kwargs": {
218
+ "brightness": [
219
+ 0.8,
220
+ 1.2
221
+ ]
222
+ }
223
+ },
224
+ "contrast": {
225
+ "weight": 1.0,
226
+ "type": "ColorJitter",
227
+ "kwargs": {
228
+ "contrast": [
229
+ 0.8,
230
+ 1.2
231
+ ]
232
+ }
233
+ },
234
+ "saturation": {
235
+ "weight": 1.0,
236
+ "type": "ColorJitter",
237
+ "kwargs": {
238
+ "saturation": [
239
+ 0.5,
240
+ 1.5
241
+ ]
242
+ }
243
+ },
244
+ "hue": {
245
+ "weight": 1.0,
246
+ "type": "ColorJitter",
247
+ "kwargs": {
248
+ "hue": [
249
+ -0.05,
250
+ 0.05
251
+ ]
252
+ }
253
+ },
254
+ "sharpness": {
255
+ "weight": 1.0,
256
+ "type": "SharpnessJitter",
257
+ "kwargs": {
258
+ "sharpness": [
259
+ 0.5,
260
+ 1.5
261
+ ]
262
+ }
263
+ },
264
+ "affine": {
265
+ "weight": 1.0,
266
+ "type": "RandomAffine",
267
+ "kwargs": {
268
+ "degrees": [
269
+ -5.0,
270
+ 5.0
271
+ ],
272
+ "translate": [
273
+ 0.05,
274
+ 0.05
275
+ ]
276
+ }
277
+ }
278
+ }
279
+ },
280
+ "revision": null,
281
+ "use_imagenet_stats": true,
282
+ "video_backend": "torchcodec",
283
+ "streaming": false
284
+ },
285
+ "test": {
286
+ "repo_id": [
287
+ "chomeed/aloha_static_coffee_single",
288
+ "jellyho/aloha_handover_box_single"
289
+ ],
290
+ "root": null,
291
+ "episodes": null,
292
+ "image_transforms": {
293
+ "enable": false,
294
+ "max_num_transforms": 3,
295
+ "random_order": false,
296
+ "tfs": {
297
+ "brightness": {
298
+ "weight": 1.0,
299
+ "type": "ColorJitter",
300
+ "kwargs": {
301
+ "brightness": [
302
+ 0.8,
303
+ 1.2
304
+ ]
305
+ }
306
+ },
307
+ "contrast": {
308
+ "weight": 1.0,
309
+ "type": "ColorJitter",
310
+ "kwargs": {
311
+ "contrast": [
312
+ 0.8,
313
+ 1.2
314
+ ]
315
+ }
316
+ },
317
+ "saturation": {
318
+ "weight": 1.0,
319
+ "type": "ColorJitter",
320
+ "kwargs": {
321
+ "saturation": [
322
+ 0.5,
323
+ 1.5
324
+ ]
325
+ }
326
+ },
327
+ "hue": {
328
+ "weight": 1.0,
329
+ "type": "ColorJitter",
330
+ "kwargs": {
331
+ "hue": [
332
+ -0.05,
333
+ 0.05
334
+ ]
335
+ }
336
+ },
337
+ "sharpness": {
338
+ "weight": 1.0,
339
+ "type": "SharpnessJitter",
340
+ "kwargs": {
341
+ "sharpness": [
342
+ 0.5,
343
+ 1.5
344
+ ]
345
+ }
346
+ },
347
+ "affine": {
348
+ "weight": 1.0,
349
+ "type": "RandomAffine",
350
+ "kwargs": {
351
+ "degrees": [
352
+ -5.0,
353
+ 5.0
354
+ ],
355
+ "translate": [
356
+ 0.05,
357
+ 0.05
358
+ ]
359
+ }
360
+ }
361
+ }
362
+ },
363
+ "revision": null,
364
+ "use_imagenet_stats": true,
365
+ "video_backend": "torchcodec",
366
+ "streaming": false
367
+ },
368
+ "wandb": {
369
+ "enable": true,
370
+ "disable_artifact": true,
371
+ "project": "long_horizon_reward_modeling",
372
+ "entity": null,
373
+ "notes": null,
374
+ "run_id": "c7ho7ell",
375
+ "mode": null
376
+ },
377
+ "peft": null,
378
+ "use_rabc": false,
379
+ "rabc_progress_path": null,
380
+ "rabc_kappa": 0.01,
381
+ "rabc_epsilon": 1e-06,
382
+ "rabc_head_mode": "sparse",
383
+ "rename_map": {},
384
+ "checkpoint_path": null
385
+ }