Hugging Face's logo

schmevlin
/
xvla_test

Robotics
LeRobot
Safetensors
xvla
Model card Files
xet
 Community
schmevlin commited on
Commit
deffe6f
·
verified ·
1 Parent(s): f119953

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +196 -0
  3. model.safetensors +3 -0
  4. train_config.json +334 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: schmevlin/plain_background
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: xvla
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - lerobot
9
+ - xvla
10
+ - robotics
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 +1,196 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "xvla",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.images.camera1": {
6
+ "type": "VISUAL",
7
+ "shape": [
8
+ 3,
9
+ 640,
10
+ 480
11
+ ]
12
+ },
13
+ "observation.images.camera2": {
14
+ "type": "VISUAL",
15
+ "shape": [
16
+ 3,
17
+ 640,
18
+ 480
19
+ ]
20
+ },
21
+ "observation.images.camera3": {
22
+ "type": "VISUAL",
23
+ "shape": [
24
+ 3,
25
+ 640,
26
+ 480
27
+ ]
28
+ },
29
+ "observation.state": {
30
+ "type": "STATE",
31
+ "shape": [
32
+ 8
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
+ "use_peft": false,
47
+ "push_to_hub": true,
48
+ "repo_id": "schmevlin/xvla_test",
49
+ "private": null,
50
+ "tags": null,
51
+ "license": null,
52
+ "pretrained_path": "outputs/train/xvla_test/checkpoints/last/pretrained_model",
53
+ "chunk_size": 30,
54
+ "n_action_steps": 30,
55
+ "dtype": "bfloat16",
56
+ "normalization_mapping": {
57
+ "STATE": "IDENTITY",
58
+ "ACTION": "IDENTITY",
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
+ "attention_dropout": 0.1,
73
+ "d_model": 1024,
74
+ "decoder_attention_heads": 16,
75
+ "decoder_layers": 12,
76
+ "encoder_attention_heads": 16,
77
+ "encoder_layers": 12,
78
+ "dropout": 0.1,
79
+ "max_position_embeddings": 4096,
80
+ "num_hidden_layers": 12,
81
+ "num_beams": 3
82
+ },
83
+ "vision_config": {
84
+ "model_type": "davit",
85
+ "drop_path_rate": 0.1,
86
+ "patch_size": [
87
+ 7,
88
+ 3,
89
+ 3,
90
+ 3
91
+ ],
92
+ "patch_stride": [
93
+ 4,
94
+ 2,
95
+ 2,
96
+ 2
97
+ ],
98
+ "patch_padding": [
99
+ 3,
100
+ 1,
101
+ 1,
102
+ 1
103
+ ],
104
+ "patch_prenorm": [
105
+ false,
106
+ true,
107
+ true,
108
+ true
109
+ ],
110
+ "enable_checkpoint": false,
111
+ "dim_embed": [
112
+ 256,
113
+ 512,
114
+ 1024,
115
+ 2048
116
+ ],
117
+ "num_heads": [
118
+ 8,
119
+ 16,
120
+ 32,
121
+ 64
122
+ ],
123
+ "num_groups": [
124
+ 8,
125
+ 16,
126
+ 32,
127
+ 64
128
+ ],
129
+ "depths": [
130
+ 1,
131
+ 1,
132
+ 9,
133
+ 1
134
+ ],
135
+ "window_size": 12,
136
+ "projection_dim": 1024,
137
+ "visual_temporal_embedding": {
138
+ "type": "COSINE",
139
+ "max_temporal_embeddings": 100
140
+ },
141
+ "image_pos_embed": {
142
+ "type": "learned_abs_2d",
143
+ "max_pos_embeddings": 50
144
+ },
145
+ "image_feature_source": [
146
+ "spatial_avg_pool",
147
+ "temporal_avg_pool"
148
+ ]
149
+ },
150
+ "vocab_size": 51289,
151
+ "torch_dtype": "float32",
152
+ "is_encoder_decoder": true
153
+ },
154
+ "tokenizer_name": "facebook/bart-large",
155
+ "tokenizer_max_length": 50,
156
+ "tokenizer_padding_side": "right",
157
+ "pad_language_to": "max_length",
158
+ "hidden_size": 1024,
159
+ "depth": 24,
160
+ "num_heads": 16,
161
+ "mlp_ratio": 4.0,
162
+ "num_domains": 30,
163
+ "len_soft_prompts": 32,
164
+ "dim_time": 32,
165
+ "max_len_seq": 512,
166
+ "use_hetero_proj": false,
167
+ "action_mode": "ee6d",
168
+ "num_denoising_steps": 10,
169
+ "use_proprio": true,
170
+ "max_state_dim": 20,
171
+ "max_action_dim": 20,
172
+ "domain_feature_key": null,
173
+ "resize_imgs_with_padding": [
174
+ 224,
175
+ 224
176
+ ],
177
+ "num_image_views": 3,
178
+ "empty_cameras": 0,
179
+ "freeze_vision_encoder": false,
180
+ "freeze_language_encoder": false,
181
+ "train_policy_transformer": true,
182
+ "train_soft_prompts": true,
183
+ "optimizer_lr": 0.0001,
184
+ "optimizer_betas": [
185
+ 0.9,
186
+ 0.95
187
+ ],
188
+ "optimizer_eps": 1e-08,
189
+ "optimizer_weight_decay": 0.0001,
190
+ "optimizer_grad_clip_norm": 10.0,
191
+ "optimizer_soft_prompt_lr_scale": 1.0,
192
+ "optimizer_soft_prompt_warmup_lr_scale": null,
193
+ "scheduler_warmup_steps": 1000,
194
+ "scheduler_decay_steps": 30000,
195
+ "scheduler_decay_lr": 2.5e-06
196
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4be5dc47b30953d21056fb37639c69c4fc01ac695adc3f427414e8e5d3d79e87
3
+ size 1759596986
train_config.json ADDED
@@ -0,0 +1,334 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "schmevlin/plain_background",
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.camera1": {
88
+ "type": "VISUAL",
89
+ "shape": [
90
+ 3,
91
+ 640,
92
+ 480
93
+ ]
94
+ },
95
+ "observation.images.camera2": {
96
+ "type": "VISUAL",
97
+ "shape": [
98
+ 3,
99
+ 640,
100
+ 480
101
+ ]
102
+ },
103
+ "observation.images.camera3": {
104
+ "type": "VISUAL",
105
+ "shape": [
106
+ 3,
107
+ 640,
108
+ 480
109
+ ]
110
+ },
111
+ "observation.state": {
112
+ "type": "STATE",
113
+ "shape": [
114
+ 8
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
+ "use_peft": false,
129
+ "push_to_hub": true,
130
+ "repo_id": "schmevlin/xvla_test",
131
+ "private": null,
132
+ "tags": null,
133
+ "license": null,
134
+ "pretrained_path": "outputs/train/xvla_test/checkpoints/last/pretrained_model",
135
+ "chunk_size": 30,
136
+ "n_action_steps": 30,
137
+ "dtype": "bfloat16",
138
+ "normalization_mapping": {
139
+ "STATE": "IDENTITY",
140
+ "ACTION": "IDENTITY",
141
+ "VISUAL": "IDENTITY"
142
+ },
143
+ "florence_config": {
144
+ "model_type": "florence2",
145
+ "bos_token_id": 0,
146
+ "eos_token_id": 2,
147
+ "ignore_index": -100,
148
+ "pad_token_id": 1,
149
+ "projection_dim": 1024,
150
+ "text_config": {
151
+ "vocab_size": 51289,
152
+ "activation_dropout": 0.1,
153
+ "activation_function": "gelu",
154
+ "attention_dropout": 0.1,
155
+ "d_model": 1024,
156
+ "decoder_attention_heads": 16,
157
+ "decoder_layers": 12,
158
+ "encoder_attention_heads": 16,
159
+ "encoder_layers": 12,
160
+ "dropout": 0.1,
161
+ "max_position_embeddings": 4096,
162
+ "num_hidden_layers": 12,
163
+ "num_beams": 3
164
+ },
165
+ "vision_config": {
166
+ "model_type": "davit",
167
+ "drop_path_rate": 0.1,
168
+ "patch_size": [
169
+ 7,
170
+ 3,
171
+ 3,
172
+ 3
173
+ ],
174
+ "patch_stride": [
175
+ 4,
176
+ 2,
177
+ 2,
178
+ 2
179
+ ],
180
+ "patch_padding": [
181
+ 3,
182
+ 1,
183
+ 1,
184
+ 1
185
+ ],
186
+ "patch_prenorm": [
187
+ false,
188
+ true,
189
+ true,
190
+ true
191
+ ],
192
+ "enable_checkpoint": false,
193
+ "dim_embed": [
194
+ 256,
195
+ 512,
196
+ 1024,
197
+ 2048
198
+ ],
199
+ "num_heads": [
200
+ 8,
201
+ 16,
202
+ 32,
203
+ 64
204
+ ],
205
+ "num_groups": [
206
+ 8,
207
+ 16,
208
+ 32,
209
+ 64
210
+ ],
211
+ "depths": [
212
+ 1,
213
+ 1,
214
+ 9,
215
+ 1
216
+ ],
217
+ "window_size": 12,
218
+ "projection_dim": 1024,
219
+ "visual_temporal_embedding": {
220
+ "type": "COSINE",
221
+ "max_temporal_embeddings": 100
222
+ },
223
+ "image_pos_embed": {
224
+ "type": "learned_abs_2d",
225
+ "max_pos_embeddings": 50
226
+ },
227
+ "image_feature_source": [
228
+ "spatial_avg_pool",
229
+ "temporal_avg_pool"
230
+ ]
231
+ },
232
+ "vocab_size": 51289,
233
+ "torch_dtype": "float32",
234
+ "is_encoder_decoder": true
235
+ },
236
+ "tokenizer_name": "facebook/bart-large",
237
+ "tokenizer_max_length": 50,
238
+ "tokenizer_padding_side": "right",
239
+ "pad_language_to": "max_length",
240
+ "hidden_size": 1024,
241
+ "depth": 24,
242
+ "num_heads": 16,
243
+ "mlp_ratio": 4.0,
244
+ "num_domains": 30,
245
+ "len_soft_prompts": 32,
246
+ "dim_time": 32,
247
+ "max_len_seq": 512,
248
+ "use_hetero_proj": false,
249
+ "action_mode": "ee6d",
250
+ "num_denoising_steps": 10,
251
+ "use_proprio": true,
252
+ "max_state_dim": 20,
253
+ "max_action_dim": 20,
254
+ "domain_feature_key": null,
255
+ "resize_imgs_with_padding": [
256
+ 224,
257
+ 224
258
+ ],
259
+ "num_image_views": 3,
260
+ "empty_cameras": 0,
261
+ "freeze_vision_encoder": false,
262
+ "freeze_language_encoder": false,
263
+ "train_policy_transformer": true,
264
+ "train_soft_prompts": true,
265
+ "optimizer_lr": 0.0001,
266
+ "optimizer_betas": [
267
+ 0.9,
268
+ 0.95
269
+ ],
270
+ "optimizer_eps": 1e-08,
271
+ "optimizer_weight_decay": 0.0001,
272
+ "optimizer_grad_clip_norm": 10.0,
273
+ "optimizer_soft_prompt_lr_scale": 1.0,
274
+ "optimizer_soft_prompt_warmup_lr_scale": null,
275
+ "scheduler_warmup_steps": 1000,
276
+ "scheduler_decay_steps": 30000,
277
+ "scheduler_decay_lr": 2.5e-06
278
+ },
279
+ "output_dir": "outputs/train/xvla_test",
280
+ "job_name": "xvla_training",
281
+ "resume": true,
282
+ "seed": 1000,
283
+ "num_workers": 4,
284
+ "batch_size": 4,
285
+ "steps": 100000,
286
+ "eval_freq": 20000,
287
+ "log_freq": 200,
288
+ "tolerance_s": 0.0001,
289
+ "save_checkpoint": true,
290
+ "save_freq": 20000,
291
+ "use_policy_training_preset": true,
292
+ "optimizer": {
293
+ "type": "xvla-adamw",
294
+ "lr": 0.0001,
295
+ "weight_decay": 0.0001,
296
+ "grad_clip_norm": 10.0,
297
+ "betas": [
298
+ 0.9,
299
+ 0.95
300
+ ],
301
+ "eps": 1e-08,
302
+ "soft_prompt_lr_scale": 1.0,
303
+ "soft_prompt_warmup_lr_scale": null
304
+ },
305
+ "scheduler": {
306
+ "type": "cosine_decay_with_warmup",
307
+ "num_warmup_steps": 1000,
308
+ "num_decay_steps": 30000,
309
+ "peak_lr": 0.0001,
310
+ "decay_lr": 2.5e-06
311
+ },
312
+ "eval": {
313
+ "n_episodes": 50,
314
+ "batch_size": 50,
315
+ "use_async_envs": false
316
+ },
317
+ "wandb": {
318
+ "enable": false,
319
+ "disable_artifact": false,
320
+ "project": "lerobot",
321
+ "entity": null,
322
+ "notes": null,
323
+ "run_id": null,
324
+ "mode": null
325
+ },
326
+ "peft": null,
327
+ "use_rabc": false,
328
+ "rabc_progress_path": null,
329
+ "rabc_kappa": 0.01,
330
+ "rabc_epsilon": 1e-06,
331
+ "rabc_head_mode": "sparse",
332
+ "rename_map": {},
333
+ "checkpoint_path": "outputs/train/xvla_test/checkpoints/last"
334
+ }