AriRyo commited on
Commit
0c0b98a
·
verified ·
1 Parent(s): cd1003e

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 +378 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: AriRyo/blockpick_gray_03
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: xvla
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - lerobot
9
+ - robotics
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 +1,222 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ 6
41
+ ]
42
+ }
43
+ },
44
+ "device": "cuda",
45
+ "use_amp": false,
46
+ "use_peft": false,
47
+ "push_to_hub": true,
48
+ "repo_id": "AriRyo/blockpick_gray_xvla_12",
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": false,
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 sha256:0b11fe72fc380031ac617d33a79715d236467c938f779469b583ab30204f7856
3
+ size 1759596986
train_config.json ADDED
@@ -0,0 +1,378 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "AriRyo/blockpick_gray_03",
4
+ "root": null,
5
+ "episodes": [
6
+ 0,
7
+ 1,
8
+ 2,
9
+ 3,
10
+ 4,
11
+ 5,
12
+ 6,
13
+ 7,
14
+ 8,
15
+ 9,
16
+ 10,
17
+ 11
18
+ ],
19
+ "image_transforms": {
20
+ "enable": false,
21
+ "max_num_transforms": 3,
22
+ "random_order": false,
23
+ "tfs": {
24
+ "brightness": {
25
+ "weight": 1.0,
26
+ "type": "ColorJitter",
27
+ "kwargs": {
28
+ "brightness": [
29
+ 0.8,
30
+ 1.2
31
+ ]
32
+ }
33
+ },
34
+ "contrast": {
35
+ "weight": 1.0,
36
+ "type": "ColorJitter",
37
+ "kwargs": {
38
+ "contrast": [
39
+ 0.8,
40
+ 1.2
41
+ ]
42
+ }
43
+ },
44
+ "saturation": {
45
+ "weight": 1.0,
46
+ "type": "ColorJitter",
47
+ "kwargs": {
48
+ "saturation": [
49
+ 0.5,
50
+ 1.5
51
+ ]
52
+ }
53
+ },
54
+ "hue": {
55
+ "weight": 1.0,
56
+ "type": "ColorJitter",
57
+ "kwargs": {
58
+ "hue": [
59
+ -0.05,
60
+ 0.05
61
+ ]
62
+ }
63
+ },
64
+ "sharpness": {
65
+ "weight": 1.0,
66
+ "type": "SharpnessJitter",
67
+ "kwargs": {
68
+ "sharpness": [
69
+ 0.5,
70
+ 1.5
71
+ ]
72
+ }
73
+ },
74
+ "affine": {
75
+ "weight": 1.0,
76
+ "type": "RandomAffine",
77
+ "kwargs": {
78
+ "degrees": [
79
+ -5.0,
80
+ 5.0
81
+ ],
82
+ "translate": [
83
+ 0.05,
84
+ 0.05
85
+ ]
86
+ }
87
+ }
88
+ }
89
+ },
90
+ "revision": null,
91
+ "use_imagenet_stats": true,
92
+ "video_backend": "torchcodec",
93
+ "streaming": false
94
+ },
95
+ "env": null,
96
+ "policy": {
97
+ "type": "xvla",
98
+ "n_obs_steps": 1,
99
+ "input_features": {
100
+ "observation.images.image": {
101
+ "type": "VISUAL",
102
+ "shape": [
103
+ 3,
104
+ 256,
105
+ 256
106
+ ]
107
+ },
108
+ "observation.images.image2": {
109
+ "type": "VISUAL",
110
+ "shape": [
111
+ 3,
112
+ 256,
113
+ 256
114
+ ]
115
+ },
116
+ "observation.state": {
117
+ "type": "STATE",
118
+ "shape": [
119
+ 8
120
+ ]
121
+ },
122
+ "observation.images.image3": {
123
+ "type": "VISUAL",
124
+ "shape": [
125
+ 3,
126
+ 224,
127
+ 224
128
+ ]
129
+ }
130
+ },
131
+ "output_features": {
132
+ "action": {
133
+ "type": "ACTION",
134
+ "shape": [
135
+ 6
136
+ ]
137
+ }
138
+ },
139
+ "device": "cuda",
140
+ "use_amp": false,
141
+ "use_peft": false,
142
+ "push_to_hub": true,
143
+ "repo_id": "AriRyo/blockpick_gray_xvla_12",
144
+ "private": null,
145
+ "tags": null,
146
+ "license": null,
147
+ "pretrained_path": "lerobot/xvla-base",
148
+ "chunk_size": 30,
149
+ "n_action_steps": 30,
150
+ "dtype": "bfloat16",
151
+ "normalization_mapping": {
152
+ "STATE": "IDENTITY",
153
+ "ACTION": "MEAN_STD",
154
+ "VISUAL": "IDENTITY"
155
+ },
156
+ "florence_config": {
157
+ "model_type": "florence2",
158
+ "bos_token_id": 0,
159
+ "eos_token_id": 2,
160
+ "ignore_index": -100,
161
+ "pad_token_id": 1,
162
+ "projection_dim": 1024,
163
+ "text_config": {
164
+ "vocab_size": 51289,
165
+ "activation_dropout": 0.1,
166
+ "activation_function": "gelu",
167
+ "add_bias_logits": false,
168
+ "add_final_layer_norm": false,
169
+ "attention_dropout": 0.1,
170
+ "bos_token_id": 0,
171
+ "classif_dropout": 0.1,
172
+ "classifier_dropout": 0.0,
173
+ "d_model": 1024,
174
+ "decoder_attention_heads": 16,
175
+ "decoder_ffn_dim": 4096,
176
+ "decoder_layerdrop": 0.0,
177
+ "decoder_layers": 12,
178
+ "decoder_start_token_id": 2,
179
+ "dropout": 0.1,
180
+ "early_stopping": true,
181
+ "encoder_attention_heads": 16,
182
+ "encoder_ffn_dim": 4096,
183
+ "encoder_layerdrop": 0.0,
184
+ "encoder_layers": 12,
185
+ "eos_token_id": 2,
186
+ "forced_eos_token_id": 2,
187
+ "forced_bos_token_id": 0,
188
+ "gradient_checkpointing": false,
189
+ "init_std": 0.02,
190
+ "is_encoder_decoder": true,
191
+ "label2id": {
192
+ "LABEL_0": 0,
193
+ "LABEL_1": 1,
194
+ "LABEL_2": 2
195
+ },
196
+ "max_position_embeddings": 4096,
197
+ "no_repeat_ngram_size": 3,
198
+ "normalize_before": false,
199
+ "num_hidden_layers": 12,
200
+ "pad_token_id": 1,
201
+ "scale_embedding": false,
202
+ "num_beams": 3
203
+ },
204
+ "vision_config": {
205
+ "model_type": "davit",
206
+ "drop_path_rate": 0.1,
207
+ "patch_size": [
208
+ 7,
209
+ 3,
210
+ 3,
211
+ 3
212
+ ],
213
+ "patch_stride": [
214
+ 4,
215
+ 2,
216
+ 2,
217
+ 2
218
+ ],
219
+ "patch_padding": [
220
+ 3,
221
+ 1,
222
+ 1,
223
+ 1
224
+ ],
225
+ "patch_prenorm": [
226
+ false,
227
+ true,
228
+ true,
229
+ true
230
+ ],
231
+ "enable_checkpoint": false,
232
+ "dim_embed": [
233
+ 256,
234
+ 512,
235
+ 1024,
236
+ 2048
237
+ ],
238
+ "num_heads": [
239
+ 8,
240
+ 16,
241
+ 32,
242
+ 64
243
+ ],
244
+ "num_groups": [
245
+ 8,
246
+ 16,
247
+ 32,
248
+ 64
249
+ ],
250
+ "depths": [
251
+ 1,
252
+ 1,
253
+ 9,
254
+ 1
255
+ ],
256
+ "window_size": 12,
257
+ "projection_dim": 1024,
258
+ "visual_temporal_embedding": {
259
+ "type": "COSINE",
260
+ "max_temporal_embeddings": 100
261
+ },
262
+ "image_pos_embed": {
263
+ "type": "learned_abs_2d",
264
+ "max_pos_embeddings": 50
265
+ },
266
+ "image_feature_source": [
267
+ "spatial_avg_pool",
268
+ "temporal_avg_pool"
269
+ ]
270
+ },
271
+ "vocab_size": 51289,
272
+ "torch_dtype": "float32",
273
+ "is_encoder_decoder": true
274
+ },
275
+ "tokenizer_name": "facebook/bart-large",
276
+ "tokenizer_max_length": 1024,
277
+ "tokenizer_padding_side": "right",
278
+ "pad_language_to": "max_length",
279
+ "hidden_size": 1024,
280
+ "depth": 24,
281
+ "num_heads": 16,
282
+ "mlp_ratio": 4.0,
283
+ "num_domains": 30,
284
+ "len_soft_prompts": 32,
285
+ "dim_time": 32,
286
+ "max_len_seq": 512,
287
+ "use_hetero_proj": false,
288
+ "action_mode": "auto",
289
+ "num_denoising_steps": 10,
290
+ "use_proprio": true,
291
+ "max_state_dim": 20,
292
+ "max_action_dim": 20,
293
+ "domain_feature_key": null,
294
+ "resize_imgs_with_padding": [
295
+ 224,
296
+ 224
297
+ ],
298
+ "num_image_views": 3,
299
+ "empty_cameras": 0,
300
+ "freeze_vision_encoder": false,
301
+ "freeze_language_encoder": false,
302
+ "train_policy_transformer": true,
303
+ "train_soft_prompts": true,
304
+ "optimizer_lr": 0.0001,
305
+ "optimizer_betas": [
306
+ 0.9,
307
+ 0.95
308
+ ],
309
+ "optimizer_eps": 1e-08,
310
+ "optimizer_weight_decay": 0.0001,
311
+ "optimizer_grad_clip_norm": 10.0,
312
+ "optimizer_soft_prompt_lr_scale": 1.0,
313
+ "optimizer_soft_prompt_warmup_lr_scale": null,
314
+ "scheduler_warmup_steps": 1000,
315
+ "scheduler_decay_steps": 30000,
316
+ "scheduler_decay_lr": 2.5e-06
317
+ },
318
+ "output_dir": "outputs/train/blockpick_gray_xvla_12",
319
+ "job_name": "blockpick_gray_xvla_12",
320
+ "resume": false,
321
+ "seed": 1000,
322
+ "cudnn_deterministic": false,
323
+ "num_workers": 4,
324
+ "batch_size": 8,
325
+ "steps": 4000,
326
+ "eval_freq": 20000,
327
+ "log_freq": 200,
328
+ "tolerance_s": 0.0001,
329
+ "save_checkpoint": true,
330
+ "save_freq": 1000,
331
+ "use_policy_training_preset": true,
332
+ "optimizer": {
333
+ "type": "xvla-adamw",
334
+ "lr": 0.0001,
335
+ "weight_decay": 0.0001,
336
+ "grad_clip_norm": 10.0,
337
+ "betas": [
338
+ 0.9,
339
+ 0.95
340
+ ],
341
+ "eps": 1e-08,
342
+ "soft_prompt_lr_scale": 1.0,
343
+ "soft_prompt_warmup_lr_scale": null
344
+ },
345
+ "scheduler": {
346
+ "type": "cosine_decay_with_warmup",
347
+ "num_warmup_steps": 1000,
348
+ "num_decay_steps": 30000,
349
+ "peak_lr": 0.0001,
350
+ "decay_lr": 2.5e-06
351
+ },
352
+ "eval": {
353
+ "n_episodes": 50,
354
+ "batch_size": 50,
355
+ "use_async_envs": false
356
+ },
357
+ "wandb": {
358
+ "enable": true,
359
+ "disable_artifact": true,
360
+ "project": "lerobot",
361
+ "entity": null,
362
+ "notes": null,
363
+ "run_id": "u3rp9mmq",
364
+ "mode": null,
365
+ "add_tags": true
366
+ },
367
+ "peft": null,
368
+ "use_rabc": false,
369
+ "rabc_progress_path": null,
370
+ "rabc_kappa": 0.01,
371
+ "rabc_epsilon": 1e-06,
372
+ "rabc_head_mode": "sparse",
373
+ "rename_map": {
374
+ "observation.images.above": "observation.images.image",
375
+ "observation.images.side": "observation.images.image2"
376
+ },
377
+ "checkpoint_path": null
378
+ }