Hugging Face's logo

zonglin11
/
xvla-xlerobot-clean-table-v2

Robotics
LeRobot
Safetensors
xvla
Model card Files
xet
 Community
zonglin11 commited on
Commit
bdbd77d
·
verified ·
1 Parent(s): a6418ca

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +222 -0
  3. model.safetensors +3 -0
  4. train_config.json +366 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: zonglin11/xlerobot_clean_table
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: xvla
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - robotics
9
+ - xvla
10
+ - lerobot
11
+ ---
12
+
13
+ # Model Card for xvla
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 +1,222 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "xvla",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.images.image": {
6
+ "type": "VISUAL",
7
+ "shape": [
8
+ 3,
9
+ 256,
10
+ 256
11
+ ]
12
+ },
13
+ "observation.images.image2": {
14
+ "type": "VISUAL",
15
+ "shape": [
16
+ 3,
17
+ 256,
18
+ 256
19
+ ]
20
+ },
21
+ "observation.state": {
22
+ "type": "STATE",
23
+ "shape": [
24
+ 8
25
+ ]
26
+ },
27
+ "observation.images.image3": {
28
+ "type": "VISUAL",
29
+ "shape": [
30
+ 3,
31
+ 224,
32
+ 224
33
+ ]
34
+ }
35
+ },
36
+ "output_features": {
37
+ "action": {
38
+ "type": "ACTION",
39
+ "shape": [
40
+ 16
41
+ ]
42
+ }
43
+ },
44
+ "device": "cuda",
45
+ "use_amp": false,
46
+ "use_peft": false,
47
+ "push_to_hub": true,
48
+ "repo_id": "zonglin11/xvla-xlerobot-clean-table-v2",
49
+ "private": null,
50
+ "tags": null,
51
+ "license": null,
52
+ "pretrained_path": "lerobot/xvla-base",
53
+ "chunk_size": 30,
54
+ "n_action_steps": 30,
55
+ "dtype": "bfloat16",
56
+ "normalization_mapping": {
57
+ "STATE": "IDENTITY",
58
+ "ACTION": "MEAN_STD",
59
+ "VISUAL": "IDENTITY"
60
+ },
61
+ "florence_config": {
62
+ "model_type": "florence2",
63
+ "bos_token_id": 0,
64
+ "eos_token_id": 2,
65
+ "ignore_index": -100,
66
+ "pad_token_id": 1,
67
+ "projection_dim": 1024,
68
+ "text_config": {
69
+ "vocab_size": 51289,
70
+ "activation_dropout": 0.1,
71
+ "activation_function": "gelu",
72
+ "add_bias_logits": false,
73
+ "add_final_layer_norm": false,
74
+ "attention_dropout": 0.1,
75
+ "bos_token_id": 0,
76
+ "classif_dropout": 0.1,
77
+ "classifier_dropout": 0.0,
78
+ "d_model": 1024,
79
+ "decoder_attention_heads": 16,
80
+ "decoder_ffn_dim": 4096,
81
+ "decoder_layerdrop": 0.0,
82
+ "decoder_layers": 12,
83
+ "decoder_start_token_id": 2,
84
+ "dropout": 0.1,
85
+ "early_stopping": true,
86
+ "encoder_attention_heads": 16,
87
+ "encoder_ffn_dim": 4096,
88
+ "encoder_layerdrop": 0.0,
89
+ "encoder_layers": 12,
90
+ "eos_token_id": 2,
91
+ "forced_eos_token_id": 2,
92
+ "forced_bos_token_id": 0,
93
+ "gradient_checkpointing": false,
94
+ "init_std": 0.02,
95
+ "is_encoder_decoder": true,
96
+ "label2id": {
97
+ "LABEL_0": 0,
98
+ "LABEL_1": 1,
99
+ "LABEL_2": 2
100
+ },
101
+ "max_position_embeddings": 4096,
102
+ "no_repeat_ngram_size": 3,
103
+ "normalize_before": false,
104
+ "num_hidden_layers": 12,
105
+ "pad_token_id": 1,
106
+ "scale_embedding": false,
107
+ "num_beams": 3
108
+ },
109
+ "vision_config": {
110
+ "model_type": "davit",
111
+ "drop_path_rate": 0.1,
112
+ "patch_size": [
113
+ 7,
114
+ 3,
115
+ 3,
116
+ 3
117
+ ],
118
+ "patch_stride": [
119
+ 4,
120
+ 2,
121
+ 2,
122
+ 2
123
+ ],
124
+ "patch_padding": [
125
+ 3,
126
+ 1,
127
+ 1,
128
+ 1
129
+ ],
130
+ "patch_prenorm": [
131
+ false,
132
+ true,
133
+ true,
134
+ true
135
+ ],
136
+ "enable_checkpoint": false,
137
+ "dim_embed": [
138
+ 256,
139
+ 512,
140
+ 1024,
141
+ 2048
142
+ ],
143
+ "num_heads": [
144
+ 8,
145
+ 16,
146
+ 32,
147
+ 64
148
+ ],
149
+ "num_groups": [
150
+ 8,
151
+ 16,
152
+ 32,
153
+ 64
154
+ ],
155
+ "depths": [
156
+ 1,
157
+ 1,
158
+ 9,
159
+ 1
160
+ ],
161
+ "window_size": 12,
162
+ "projection_dim": 1024,
163
+ "visual_temporal_embedding": {
164
+ "type": "COSINE",
165
+ "max_temporal_embeddings": 100
166
+ },
167
+ "image_pos_embed": {
168
+ "type": "learned_abs_2d",
169
+ "max_pos_embeddings": 50
170
+ },
171
+ "image_feature_source": [
172
+ "spatial_avg_pool",
173
+ "temporal_avg_pool"
174
+ ]
175
+ },
176
+ "vocab_size": 51289,
177
+ "torch_dtype": "float32",
178
+ "is_encoder_decoder": true
179
+ },
180
+ "tokenizer_name": "facebook/bart-large",
181
+ "tokenizer_max_length": 1024,
182
+ "tokenizer_padding_side": "right",
183
+ "pad_language_to": "max_length",
184
+ "hidden_size": 1024,
185
+ "depth": 24,
186
+ "num_heads": 16,
187
+ "mlp_ratio": 4.0,
188
+ "num_domains": 30,
189
+ "len_soft_prompts": 32,
190
+ "dim_time": 32,
191
+ "max_len_seq": 512,
192
+ "use_hetero_proj": false,
193
+ "action_mode": "auto",
194
+ "num_denoising_steps": 10,
195
+ "use_proprio": true,
196
+ "max_state_dim": 20,
197
+ "max_action_dim": 20,
198
+ "domain_feature_key": null,
199
+ "resize_imgs_with_padding": [
200
+ 224,
201
+ 224
202
+ ],
203
+ "num_image_views": 3,
204
+ "empty_cameras": 0,
205
+ "freeze_vision_encoder": false,
206
+ "freeze_language_encoder": true,
207
+ "train_policy_transformer": true,
208
+ "train_soft_prompts": true,
209
+ "optimizer_lr": 0.0001,
210
+ "optimizer_betas": [
211
+ 0.9,
212
+ 0.95
213
+ ],
214
+ "optimizer_eps": 1e-08,
215
+ "optimizer_weight_decay": 0.0001,
216
+ "optimizer_grad_clip_norm": 10.0,
217
+ "optimizer_soft_prompt_lr_scale": 1.0,
218
+ "optimizer_soft_prompt_warmup_lr_scale": null,
219
+ "scheduler_warmup_steps": 1000,
220
+ "scheduler_decay_steps": 30000,
221
+ "scheduler_decay_lr": 2.5e-06
222
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9fa514befa94af1e6f71c6a548bb5a9cb2b9b1aaae76f4614491e1f97f57b131
3
+ size 1759596986
train_config.json ADDED
@@ -0,0 +1,366 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "zonglin11/xlerobot_clean_table",
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
+ "return_uint8": false,
81
+ "streaming": false
82
+ },
83
+ "env": null,
84
+ "policy": {
85
+ "type": "xvla",
86
+ "n_obs_steps": 1,
87
+ "input_features": {
88
+ "observation.images.image": {
89
+ "type": "VISUAL",
90
+ "shape": [
91
+ 3,
92
+ 256,
93
+ 256
94
+ ]
95
+ },
96
+ "observation.images.image2": {
97
+ "type": "VISUAL",
98
+ "shape": [
99
+ 3,
100
+ 256,
101
+ 256
102
+ ]
103
+ },
104
+ "observation.state": {
105
+ "type": "STATE",
106
+ "shape": [
107
+ 8
108
+ ]
109
+ },
110
+ "observation.images.image3": {
111
+ "type": "VISUAL",
112
+ "shape": [
113
+ 3,
114
+ 224,
115
+ 224
116
+ ]
117
+ }
118
+ },
119
+ "output_features": {
120
+ "action": {
121
+ "type": "ACTION",
122
+ "shape": [
123
+ 16
124
+ ]
125
+ }
126
+ },
127
+ "device": "cuda",
128
+ "use_amp": false,
129
+ "use_peft": false,
130
+ "push_to_hub": true,
131
+ "repo_id": "zonglin11/xvla-xlerobot-clean-table-v2",
132
+ "private": null,
133
+ "tags": null,
134
+ "license": null,
135
+ "pretrained_path": "lerobot/xvla-base",
136
+ "chunk_size": 30,
137
+ "n_action_steps": 30,
138
+ "dtype": "bfloat16",
139
+ "normalization_mapping": {
140
+ "STATE": "IDENTITY",
141
+ "ACTION": "MEAN_STD",
142
+ "VISUAL": "IDENTITY"
143
+ },
144
+ "florence_config": {
145
+ "model_type": "florence2",
146
+ "bos_token_id": 0,
147
+ "eos_token_id": 2,
148
+ "ignore_index": -100,
149
+ "pad_token_id": 1,
150
+ "projection_dim": 1024,
151
+ "text_config": {
152
+ "vocab_size": 51289,
153
+ "activation_dropout": 0.1,
154
+ "activation_function": "gelu",
155
+ "add_bias_logits": false,
156
+ "add_final_layer_norm": false,
157
+ "attention_dropout": 0.1,
158
+ "bos_token_id": 0,
159
+ "classif_dropout": 0.1,
160
+ "classifier_dropout": 0.0,
161
+ "d_model": 1024,
162
+ "decoder_attention_heads": 16,
163
+ "decoder_ffn_dim": 4096,
164
+ "decoder_layerdrop": 0.0,
165
+ "decoder_layers": 12,
166
+ "decoder_start_token_id": 2,
167
+ "dropout": 0.1,
168
+ "early_stopping": true,
169
+ "encoder_attention_heads": 16,
170
+ "encoder_ffn_dim": 4096,
171
+ "encoder_layerdrop": 0.0,
172
+ "encoder_layers": 12,
173
+ "eos_token_id": 2,
174
+ "forced_eos_token_id": 2,
175
+ "forced_bos_token_id": 0,
176
+ "gradient_checkpointing": false,
177
+ "init_std": 0.02,
178
+ "is_encoder_decoder": true,
179
+ "label2id": {
180
+ "LABEL_0": 0,
181
+ "LABEL_1": 1,
182
+ "LABEL_2": 2
183
+ },
184
+ "max_position_embeddings": 4096,
185
+ "no_repeat_ngram_size": 3,
186
+ "normalize_before": false,
187
+ "num_hidden_layers": 12,
188
+ "pad_token_id": 1,
189
+ "scale_embedding": false,
190
+ "num_beams": 3
191
+ },
192
+ "vision_config": {
193
+ "model_type": "davit",
194
+ "drop_path_rate": 0.1,
195
+ "patch_size": [
196
+ 7,
197
+ 3,
198
+ 3,
199
+ 3
200
+ ],
201
+ "patch_stride": [
202
+ 4,
203
+ 2,
204
+ 2,
205
+ 2
206
+ ],
207
+ "patch_padding": [
208
+ 3,
209
+ 1,
210
+ 1,
211
+ 1
212
+ ],
213
+ "patch_prenorm": [
214
+ false,
215
+ true,
216
+ true,
217
+ true
218
+ ],
219
+ "enable_checkpoint": false,
220
+ "dim_embed": [
221
+ 256,
222
+ 512,
223
+ 1024,
224
+ 2048
225
+ ],
226
+ "num_heads": [
227
+ 8,
228
+ 16,
229
+ 32,
230
+ 64
231
+ ],
232
+ "num_groups": [
233
+ 8,
234
+ 16,
235
+ 32,
236
+ 64
237
+ ],
238
+ "depths": [
239
+ 1,
240
+ 1,
241
+ 9,
242
+ 1
243
+ ],
244
+ "window_size": 12,
245
+ "projection_dim": 1024,
246
+ "visual_temporal_embedding": {
247
+ "type": "COSINE",
248
+ "max_temporal_embeddings": 100
249
+ },
250
+ "image_pos_embed": {
251
+ "type": "learned_abs_2d",
252
+ "max_pos_embeddings": 50
253
+ },
254
+ "image_feature_source": [
255
+ "spatial_avg_pool",
256
+ "temporal_avg_pool"
257
+ ]
258
+ },
259
+ "vocab_size": 51289,
260
+ "torch_dtype": "float32",
261
+ "is_encoder_decoder": true
262
+ },
263
+ "tokenizer_name": "facebook/bart-large",
264
+ "tokenizer_max_length": 1024,
265
+ "tokenizer_padding_side": "right",
266
+ "pad_language_to": "max_length",
267
+ "hidden_size": 1024,
268
+ "depth": 24,
269
+ "num_heads": 16,
270
+ "mlp_ratio": 4.0,
271
+ "num_domains": 30,
272
+ "len_soft_prompts": 32,
273
+ "dim_time": 32,
274
+ "max_len_seq": 512,
275
+ "use_hetero_proj": false,
276
+ "action_mode": "auto",
277
+ "num_denoising_steps": 10,
278
+ "use_proprio": true,
279
+ "max_state_dim": 20,
280
+ "max_action_dim": 20,
281
+ "domain_feature_key": null,
282
+ "resize_imgs_with_padding": [
283
+ 224,
284
+ 224
285
+ ],
286
+ "num_image_views": 3,
287
+ "empty_cameras": 0,
288
+ "freeze_vision_encoder": false,
289
+ "freeze_language_encoder": true,
290
+ "train_policy_transformer": true,
291
+ "train_soft_prompts": true,
292
+ "optimizer_lr": 0.0001,
293
+ "optimizer_betas": [
294
+ 0.9,
295
+ 0.95
296
+ ],
297
+ "optimizer_eps": 1e-08,
298
+ "optimizer_weight_decay": 0.0001,
299
+ "optimizer_grad_clip_norm": 10.0,
300
+ "optimizer_soft_prompt_lr_scale": 1.0,
301
+ "optimizer_soft_prompt_warmup_lr_scale": null,
302
+ "scheduler_warmup_steps": 1000,
303
+ "scheduler_decay_steps": 30000,
304
+ "scheduler_decay_lr": 2.5e-06
305
+ },
306
+ "reward_model": null,
307
+ "output_dir": "outputs/xvla_xlerobot_v2",
308
+ "job_name": "xvla_xlerobot_cloth_fold",
309
+ "resume": false,
310
+ "seed": 1000,
311
+ "cudnn_deterministic": false,
312
+ "num_workers": 4,
313
+ "batch_size": 4,
314
+ "prefetch_factor": 4,
315
+ "persistent_workers": true,
316
+ "steps": 30000,
317
+ "eval_freq": 20000,
318
+ "log_freq": 200,
319
+ "tolerance_s": 0.0001,
320
+ "save_checkpoint": true,
321
+ "save_freq": 20000,
322
+ "use_policy_training_preset": true,
323
+ "optimizer": {
324
+ "type": "xvla-adamw",
325
+ "lr": 0.0001,
326
+ "weight_decay": 0.0001,
327
+ "grad_clip_norm": 10.0,
328
+ "betas": [
329
+ 0.9,
330
+ 0.95
331
+ ],
332
+ "eps": 1e-08,
333
+ "soft_prompt_lr_scale": 1.0,
334
+ "soft_prompt_warmup_lr_scale": null
335
+ },
336
+ "scheduler": {
337
+ "type": "cosine_decay_with_warmup",
338
+ "num_warmup_steps": 1000,
339
+ "num_decay_steps": 30000,
340
+ "peak_lr": 0.0001,
341
+ "decay_lr": 2.5e-06
342
+ },
343
+ "eval": {
344
+ "n_episodes": 50,
345
+ "batch_size": 11,
346
+ "use_async_envs": true
347
+ },
348
+ "wandb": {
349
+ "enable": false,
350
+ "disable_artifact": false,
351
+ "project": "lerobot",
352
+ "entity": null,
353
+ "notes": null,
354
+ "run_id": null,
355
+ "mode": null,
356
+ "add_tags": true
357
+ },
358
+ "peft": null,
359
+ "sample_weighting": null,
360
+ "rename_map": {
361
+ "observation.images.top": "observation.images.image",
362
+ "observation.images.left_wrist": "observation.images.image2",
363
+ "observation.images.right_wrist": "observation.images.image3"
364
+ },
365
+ "checkpoint_path": null
366
+ }