Hugging Face's logo

crislmfroes
/
xvla-boris-sim

Robotics
LeRobot
Safetensors
xvla
Model card Files
xet
Community
crislmfroes commited on
Commit
eae2b71
·
verified ·
1 Parent(s): 92d302e

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +221 -0
  3. model.safetensors +3 -0
  4. train_config.json +355 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: crislmfroes/boris-sim
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: xvla
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - robotics
9
+ - lerobot
10
+ - xvla
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 +1,221 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ 14
25
+ ]
26
+ },
27
+ "observation.images.empty_camera_0": {
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
+ 7
41
+ ]
42
+ }
43
+ },
44
+ "device": "cuda",
45
+ "use_amp": false,
46
+ "push_to_hub": true,
47
+ "repo_id": "crislmfroes/xvla-boris-sim",
48
+ "private": null,
49
+ "tags": null,
50
+ "license": null,
51
+ "pretrained_path": "lerobot/xvla-base",
52
+ "chunk_size": 30,
53
+ "n_action_steps": 30,
54
+ "dtype": "bfloat16",
55
+ "normalization_mapping": {
56
+ "STATE": "IDENTITY",
57
+ "ACTION": "MEAN_STD",
58
+ "VISUAL": "IDENTITY"
59
+ },
60
+ "florence_config": {
61
+ "model_type": "florence2",
62
+ "bos_token_id": 0,
63
+ "eos_token_id": 2,
64
+ "ignore_index": -100,
65
+ "pad_token_id": 1,
66
+ "projection_dim": 1024,
67
+ "text_config": {
68
+ "vocab_size": 51289,
69
+ "activation_dropout": 0.1,
70
+ "activation_function": "gelu",
71
+ "add_bias_logits": false,
72
+ "add_final_layer_norm": false,
73
+ "attention_dropout": 0.1,
74
+ "bos_token_id": 0,
75
+ "classif_dropout": 0.1,
76
+ "classifier_dropout": 0.0,
77
+ "d_model": 1024,
78
+ "decoder_attention_heads": 16,
79
+ "decoder_ffn_dim": 4096,
80
+ "decoder_layerdrop": 0.0,
81
+ "decoder_layers": 12,
82
+ "decoder_start_token_id": 2,
83
+ "dropout": 0.1,
84
+ "early_stopping": true,
85
+ "encoder_attention_heads": 16,
86
+ "encoder_ffn_dim": 4096,
87
+ "encoder_layerdrop": 0.0,
88
+ "encoder_layers": 12,
89
+ "eos_token_id": 2,
90
+ "forced_eos_token_id": 2,
91
+ "forced_bos_token_id": 0,
92
+ "gradient_checkpointing": false,
93
+ "init_std": 0.02,
94
+ "is_encoder_decoder": true,
95
+ "label2id": {
96
+ "LABEL_0": 0,
97
+ "LABEL_1": 1,
98
+ "LABEL_2": 2
99
+ },
100
+ "max_position_embeddings": 4096,
101
+ "no_repeat_ngram_size": 3,
102
+ "normalize_before": false,
103
+ "num_hidden_layers": 12,
104
+ "pad_token_id": 1,
105
+ "scale_embedding": false,
106
+ "num_beams": 3
107
+ },
108
+ "vision_config": {
109
+ "model_type": "davit",
110
+ "drop_path_rate": 0.1,
111
+ "patch_size": [
112
+ 7,
113
+ 3,
114
+ 3,
115
+ 3
116
+ ],
117
+ "patch_stride": [
118
+ 4,
119
+ 2,
120
+ 2,
121
+ 2
122
+ ],
123
+ "patch_padding": [
124
+ 3,
125
+ 1,
126
+ 1,
127
+ 1
128
+ ],
129
+ "patch_prenorm": [
130
+ false,
131
+ true,
132
+ true,
133
+ true
134
+ ],
135
+ "enable_checkpoint": false,
136
+ "dim_embed": [
137
+ 256,
138
+ 512,
139
+ 1024,
140
+ 2048
141
+ ],
142
+ "num_heads": [
143
+ 8,
144
+ 16,
145
+ 32,
146
+ 64
147
+ ],
148
+ "num_groups": [
149
+ 8,
150
+ 16,
151
+ 32,
152
+ 64
153
+ ],
154
+ "depths": [
155
+ 1,
156
+ 1,
157
+ 9,
158
+ 1
159
+ ],
160
+ "window_size": 12,
161
+ "projection_dim": 1024,
162
+ "visual_temporal_embedding": {
163
+ "type": "COSINE",
164
+ "max_temporal_embeddings": 100
165
+ },
166
+ "image_pos_embed": {
167
+ "type": "learned_abs_2d",
168
+ "max_pos_embeddings": 50
169
+ },
170
+ "image_feature_source": [
171
+ "spatial_avg_pool",
172
+ "temporal_avg_pool"
173
+ ]
174
+ },
175
+ "vocab_size": 51289,
176
+ "torch_dtype": "float32",
177
+ "is_encoder_decoder": true
178
+ },
179
+ "tokenizer_name": "facebook/bart-large",
180
+ "tokenizer_max_length": 1024,
181
+ "tokenizer_padding_side": "right",
182
+ "pad_language_to": "max_length",
183
+ "hidden_size": 1024,
184
+ "depth": 24,
185
+ "num_heads": 16,
186
+ "mlp_ratio": 4.0,
187
+ "num_domains": 30,
188
+ "len_soft_prompts": 32,
189
+ "dim_time": 32,
190
+ "max_len_seq": 512,
191
+ "use_hetero_proj": false,
192
+ "action_mode": "auto",
193
+ "num_denoising_steps": 10,
194
+ "use_proprio": true,
195
+ "max_state_dim": 20,
196
+ "max_action_dim": 20,
197
+ "domain_feature_key": null,
198
+ "resize_imgs_with_padding": [
199
+ 224,
200
+ 224
201
+ ],
202
+ "num_image_views": 4,
203
+ "empty_cameras": 1,
204
+ "freeze_vision_encoder": false,
205
+ "freeze_language_encoder": false,
206
+ "train_policy_transformer": true,
207
+ "train_soft_prompts": true,
208
+ "optimizer_lr": 0.0001,
209
+ "optimizer_betas": [
210
+ 0.9,
211
+ 0.95
212
+ ],
213
+ "optimizer_eps": 1e-08,
214
+ "optimizer_weight_decay": 0.0001,
215
+ "optimizer_grad_clip_norm": 10.0,
216
+ "optimizer_soft_prompt_lr_scale": 1.0,
217
+ "optimizer_soft_prompt_warmup_lr_scale": null,
218
+ "scheduler_warmup_steps": 1000,
219
+ "scheduler_decay_steps": 30000,
220
+ "scheduler_decay_lr": 2.5e-06
221
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f95dda1021260ec0310fe9a4c50d84e9b910d77648c0b22046d97282fae586be
3
+ size 1759596986
train_config.json ADDED
@@ -0,0 +1,355 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "crislmfroes/boris-sim",
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
+ },
82
+ "env": null,
83
+ "policy": {
84
+ "type": "xvla",
85
+ "n_obs_steps": 1,
86
+ "input_features": {
87
+ "observation.images.image": {
88
+ "type": "VISUAL",
89
+ "shape": [
90
+ 3,
91
+ 256,
92
+ 256
93
+ ]
94
+ },
95
+ "observation.images.image2": {
96
+ "type": "VISUAL",
97
+ "shape": [
98
+ 3,
99
+ 256,
100
+ 256
101
+ ]
102
+ },
103
+ "observation.state": {
104
+ "type": "STATE",
105
+ "shape": [
106
+ 14
107
+ ]
108
+ },
109
+ "observation.images.empty_camera_0": {
110
+ "type": "VISUAL",
111
+ "shape": [
112
+ 3,
113
+ 224,
114
+ 224
115
+ ]
116
+ }
117
+ },
118
+ "output_features": {
119
+ "action": {
120
+ "type": "ACTION",
121
+ "shape": [
122
+ 7
123
+ ]
124
+ }
125
+ },
126
+ "device": "cuda",
127
+ "use_amp": false,
128
+ "push_to_hub": true,
129
+ "repo_id": "crislmfroes/xvla-boris-sim",
130
+ "private": null,
131
+ "tags": null,
132
+ "license": null,
133
+ "pretrained_path": "lerobot/xvla-base",
134
+ "chunk_size": 30,
135
+ "n_action_steps": 30,
136
+ "dtype": "bfloat16",
137
+ "normalization_mapping": {
138
+ "STATE": "IDENTITY",
139
+ "ACTION": "MEAN_STD",
140
+ "VISUAL": "IDENTITY"
141
+ },
142
+ "florence_config": {
143
+ "model_type": "florence2",
144
+ "bos_token_id": 0,
145
+ "eos_token_id": 2,
146
+ "ignore_index": -100,
147
+ "pad_token_id": 1,
148
+ "projection_dim": 1024,
149
+ "text_config": {
150
+ "vocab_size": 51289,
151
+ "activation_dropout": 0.1,
152
+ "activation_function": "gelu",
153
+ "add_bias_logits": false,
154
+ "add_final_layer_norm": false,
155
+ "attention_dropout": 0.1,
156
+ "bos_token_id": 0,
157
+ "classif_dropout": 0.1,
158
+ "classifier_dropout": 0.0,
159
+ "d_model": 1024,
160
+ "decoder_attention_heads": 16,
161
+ "decoder_ffn_dim": 4096,
162
+ "decoder_layerdrop": 0.0,
163
+ "decoder_layers": 12,
164
+ "decoder_start_token_id": 2,
165
+ "dropout": 0.1,
166
+ "early_stopping": true,
167
+ "encoder_attention_heads": 16,
168
+ "encoder_ffn_dim": 4096,
169
+ "encoder_layerdrop": 0.0,
170
+ "encoder_layers": 12,
171
+ "eos_token_id": 2,
172
+ "forced_eos_token_id": 2,
173
+ "forced_bos_token_id": 0,
174
+ "gradient_checkpointing": false,
175
+ "init_std": 0.02,
176
+ "is_encoder_decoder": true,
177
+ "label2id": {
178
+ "LABEL_0": 0,
179
+ "LABEL_1": 1,
180
+ "LABEL_2": 2
181
+ },
182
+ "max_position_embeddings": 4096,
183
+ "no_repeat_ngram_size": 3,
184
+ "normalize_before": false,
185
+ "num_hidden_layers": 12,
186
+ "pad_token_id": 1,
187
+ "scale_embedding": false,
188
+ "num_beams": 3
189
+ },
190
+ "vision_config": {
191
+ "model_type": "davit",
192
+ "drop_path_rate": 0.1,
193
+ "patch_size": [
194
+ 7,
195
+ 3,
196
+ 3,
197
+ 3
198
+ ],
199
+ "patch_stride": [
200
+ 4,
201
+ 2,
202
+ 2,
203
+ 2
204
+ ],
205
+ "patch_padding": [
206
+ 3,
207
+ 1,
208
+ 1,
209
+ 1
210
+ ],
211
+ "patch_prenorm": [
212
+ false,
213
+ true,
214
+ true,
215
+ true
216
+ ],
217
+ "enable_checkpoint": false,
218
+ "dim_embed": [
219
+ 256,
220
+ 512,
221
+ 1024,
222
+ 2048
223
+ ],
224
+ "num_heads": [
225
+ 8,
226
+ 16,
227
+ 32,
228
+ 64
229
+ ],
230
+ "num_groups": [
231
+ 8,
232
+ 16,
233
+ 32,
234
+ 64
235
+ ],
236
+ "depths": [
237
+ 1,
238
+ 1,
239
+ 9,
240
+ 1
241
+ ],
242
+ "window_size": 12,
243
+ "projection_dim": 1024,
244
+ "visual_temporal_embedding": {
245
+ "type": "COSINE",
246
+ "max_temporal_embeddings": 100
247
+ },
248
+ "image_pos_embed": {
249
+ "type": "learned_abs_2d",
250
+ "max_pos_embeddings": 50
251
+ },
252
+ "image_feature_source": [
253
+ "spatial_avg_pool",
254
+ "temporal_avg_pool"
255
+ ]
256
+ },
257
+ "vocab_size": 51289,
258
+ "torch_dtype": "float32",
259
+ "is_encoder_decoder": true
260
+ },
261
+ "tokenizer_name": "facebook/bart-large",
262
+ "tokenizer_max_length": 1024,
263
+ "tokenizer_padding_side": "right",
264
+ "pad_language_to": "max_length",
265
+ "hidden_size": 1024,
266
+ "depth": 24,
267
+ "num_heads": 16,
268
+ "mlp_ratio": 4.0,
269
+ "num_domains": 30,
270
+ "len_soft_prompts": 32,
271
+ "dim_time": 32,
272
+ "max_len_seq": 512,
273
+ "use_hetero_proj": false,
274
+ "action_mode": "auto",
275
+ "num_denoising_steps": 10,
276
+ "use_proprio": true,
277
+ "max_state_dim": 20,
278
+ "max_action_dim": 20,
279
+ "domain_feature_key": null,
280
+ "resize_imgs_with_padding": [
281
+ 224,
282
+ 224
283
+ ],
284
+ "num_image_views": 4,
285
+ "empty_cameras": 1,
286
+ "freeze_vision_encoder": false,
287
+ "freeze_language_encoder": false,
288
+ "train_policy_transformer": true,
289
+ "train_soft_prompts": true,
290
+ "optimizer_lr": 0.0001,
291
+ "optimizer_betas": [
292
+ 0.9,
293
+ 0.95
294
+ ],
295
+ "optimizer_eps": 1e-08,
296
+ "optimizer_weight_decay": 0.0001,
297
+ "optimizer_grad_clip_norm": 10.0,
298
+ "optimizer_soft_prompt_lr_scale": 1.0,
299
+ "optimizer_soft_prompt_warmup_lr_scale": null,
300
+ "scheduler_warmup_steps": 1000,
301
+ "scheduler_decay_steps": 30000,
302
+ "scheduler_decay_lr": 2.5e-06
303
+ },
304
+ "output_dir": "outputs/xvla_training",
305
+ "job_name": "xvla_training",
306
+ "resume": false,
307
+ "seed": 1000,
308
+ "num_workers": 4,
309
+ "batch_size": 4,
310
+ "steps": 20000,
311
+ "eval_freq": 20000,
312
+ "log_freq": 200,
313
+ "save_checkpoint": true,
314
+ "save_freq": 20000,
315
+ "use_policy_training_preset": true,
316
+ "optimizer": {
317
+ "type": "xvla-adamw",
318
+ "lr": 0.0001,
319
+ "weight_decay": 0.0001,
320
+ "grad_clip_norm": 10.0,
321
+ "betas": [
322
+ 0.9,
323
+ 0.95
324
+ ],
325
+ "eps": 1e-08,
326
+ "soft_prompt_lr_scale": 1.0,
327
+ "soft_prompt_warmup_lr_scale": null
328
+ },
329
+ "scheduler": {
330
+ "type": "cosine_decay_with_warmup",
331
+ "num_warmup_steps": 1000,
332
+ "num_decay_steps": 30000,
333
+ "peak_lr": 0.0001,
334
+ "decay_lr": 2.5e-06
335
+ },
336
+ "eval": {
337
+ "n_episodes": 50,
338
+ "batch_size": 50,
339
+ "use_async_envs": false
340
+ },
341
+ "wandb": {
342
+ "enable": false,
343
+ "disable_artifact": false,
344
+ "project": "lerobot",
345
+ "entity": null,
346
+ "notes": null,
347
+ "run_id": null,
348
+ "mode": null
349
+ },
350
+ "checkpoint_path": null,
351
+ "rename_map": {
352
+ "observation.images.top": "observation.images.image",
353
+ "observation.images.wrist": "observation.images.image2"
354
+ }
355
+ }