Hugging Face's logo

ducido
/
m3_20eps_base_40k

Safetensors
Model card Files
xet
 Community
ducido commited on
Commit
7bdb560
·
verified ·
1 Parent(s): 43a41d5

Initial commit

Browse files
Files changed (3) hide show
  1. config.json +77 -0
  2. model.safetensors +3 -0
  3. train_config.json +270 -0
config.json ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "litevla-ms",
3
+ "n_obs_steps": 1,
4
+ "normalization_mapping": {
5
+ "VISUAL": "IDENTITY",
6
+ "STATE": "MEAN_STD",
7
+ "ACTION": "MEAN_STD"
8
+ },
9
+ "input_features": {
10
+ "observation.state": {
11
+ "type": "STATE",
12
+ "shape": [
13
+ 5
14
+ ]
15
+ },
16
+ "observation.image": {
17
+ "type": "VISUAL",
18
+ "shape": [
19
+ 3,
20
+ 256,
21
+ 256
22
+ ]
23
+ }
24
+ },
25
+ "output_features": {
26
+ "action": {
27
+ "type": "ACTION",
28
+ "shape": [
29
+ 5
30
+ ]
31
+ }
32
+ },
33
+ "device": "cuda",
34
+ "use_amp": false,
35
+ "gradient_accumulation_steps": 1,
36
+ "chunk_size": 50,
37
+ "n_action_steps": 1,
38
+ "max_state_dim": 32,
39
+ "max_action_dim": 32,
40
+ "resize_imgs_with_padding": [
41
+ 512,
42
+ 512
43
+ ],
44
+ "empty_cameras": 0,
45
+ "adapt_to_pi_aloha": false,
46
+ "use_delta_joint_actions_aloha": false,
47
+ "tokenizer_max_length": 48,
48
+ "num_steps": 10,
49
+ "use_cache": true,
50
+ "freeze_vision_encoder": true,
51
+ "train_expert_only": false,
52
+ "train_state_proj": true,
53
+ "optimizer_lr": 0.0001,
54
+ "optimizer_betas": [
55
+ 0.9,
56
+ 0.95
57
+ ],
58
+ "optimizer_eps": 1e-08,
59
+ "optimizer_weight_decay": 1e-10,
60
+ "optimizer_grad_clip_norm": 10,
61
+ "scheduler_warmup_steps": 1000,
62
+ "scheduler_decay_steps": 100000,
63
+ "scheduler_decay_lr": 2.5e-06,
64
+ "vlm_model_name": "/projects/extern/kisski/kisski-spath/dir.project/VLA_Imit/LiteVLA-MS/CP/SmolVLM2-500M-Video-Instruct",
65
+ "load_vlm_weights": true,
66
+ "add_image_special_tokens": false,
67
+ "attention_mode": "cross_attn",
68
+ "prefix_length": 0,
69
+ "pad_language_to": "max_length",
70
+ "num_expert_layers": 0,
71
+ "num_vlm_layers": 16,
72
+ "self_attn_every_n_layers": 2,
73
+ "expert_width_multiplier": 0.75,
74
+ "min_period": 0.004,
75
+ "max_period": 4.0,
76
+ "of_path": null
77
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a00a4faf6f0832e6ce0271d755b3dd77fe62e29b33efad59a124f9d0d738f1d2
3
+ size 934390336
train_config.json ADDED
@@ -0,0 +1,270 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": ".",
4
+ "root": "/projects/extern/kisski/kisski-spath/dir.project/VLA_Imit/LiteVLA-MS/CP/lerobot/ducido/m3_20eps_lerobot",
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": true,
8
+ "max_num_transforms": 3,
9
+ "random_order": false,
10
+ "image_tfs": {
11
+ "hue": {
12
+ "weight": 1.0,
13
+ "type": "ColorJitter",
14
+ "kwargs": {
15
+ "hue": [
16
+ -0.05,
17
+ 0.05
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
+ "sharpness": {
32
+ "weight": 1.0,
33
+ "type": "SharpnessJitter",
34
+ "kwargs": {
35
+ "sharpness": [
36
+ 0.5,
37
+ 1.5
38
+ ]
39
+ }
40
+ },
41
+ "brightness": {
42
+ "weight": 1.0,
43
+ "type": "ColorJitter",
44
+ "kwargs": {
45
+ "brightness": [
46
+ 0.8,
47
+ 1.2
48
+ ]
49
+ }
50
+ },
51
+ "saturation": {
52
+ "weight": 1.0,
53
+ "type": "ColorJitter",
54
+ "kwargs": {
55
+ "saturation": [
56
+ 0.5,
57
+ 1.5
58
+ ]
59
+ }
60
+ },
61
+ "crop_resize": {
62
+ "weight": 1.0,
63
+ "type": "RandomResizedCrop",
64
+ "kwargs": {
65
+ "size": [
66
+ 256,
67
+ 256
68
+ ],
69
+ "ratio": [
70
+ 1,
71
+ 1
72
+ ],
73
+ "scale": [
74
+ 0.9,
75
+ 0.95
76
+ ]
77
+ }
78
+ },
79
+ "rotate": {
80
+ "weight": 1.0,
81
+ "type": "RandomRotate",
82
+ "kwargs": {
83
+ "degrees": [
84
+ -5,
85
+ 5
86
+ ]
87
+ }
88
+ }
89
+ },
90
+ "wrist_tfs": {
91
+ "hue": {
92
+ "weight": 1.0,
93
+ "type": "ColorJitter",
94
+ "kwargs": {
95
+ "hue": [
96
+ -0.05,
97
+ 0.05
98
+ ]
99
+ }
100
+ },
101
+ "contrast": {
102
+ "weight": 1.0,
103
+ "type": "ColorJitter",
104
+ "kwargs": {
105
+ "contrast": [
106
+ 0.8,
107
+ 1.2
108
+ ]
109
+ }
110
+ },
111
+ "sharpness": {
112
+ "weight": 1.0,
113
+ "type": "SharpnessJitter",
114
+ "kwargs": {
115
+ "sharpness": [
116
+ 0.5,
117
+ 1.5
118
+ ]
119
+ }
120
+ },
121
+ "brightness": {
122
+ "weight": 1.0,
123
+ "type": "ColorJitter",
124
+ "kwargs": {
125
+ "brightness": [
126
+ 0.8,
127
+ 1.2
128
+ ]
129
+ }
130
+ },
131
+ "saturation": {
132
+ "weight": 1.0,
133
+ "type": "ColorJitter",
134
+ "kwargs": {
135
+ "saturation": [
136
+ 0.5,
137
+ 1.5
138
+ ]
139
+ }
140
+ }
141
+ }
142
+ },
143
+ "revision": null,
144
+ "use_imagenet_stats": true,
145
+ "video_backend": "torchcodec",
146
+ "vqa_data_path": null
147
+ },
148
+ "env": null,
149
+ "policy": {
150
+ "type": "litevla-ms",
151
+ "n_obs_steps": 1,
152
+ "normalization_mapping": {
153
+ "VISUAL": "IDENTITY",
154
+ "STATE": "MEAN_STD",
155
+ "ACTION": "MEAN_STD"
156
+ },
157
+ "input_features": {
158
+ "observation.state": {
159
+ "type": "STATE",
160
+ "shape": [
161
+ 5
162
+ ]
163
+ },
164
+ "observation.image": {
165
+ "type": "VISUAL",
166
+ "shape": [
167
+ 3,
168
+ 256,
169
+ 256
170
+ ]
171
+ }
172
+ },
173
+ "output_features": {
174
+ "action": {
175
+ "type": "ACTION",
176
+ "shape": [
177
+ 5
178
+ ]
179
+ }
180
+ },
181
+ "device": "cuda",
182
+ "use_amp": false,
183
+ "gradient_accumulation_steps": 1,
184
+ "chunk_size": 50,
185
+ "n_action_steps": 1,
186
+ "max_state_dim": 32,
187
+ "max_action_dim": 32,
188
+ "resize_imgs_with_padding": [
189
+ 512,
190
+ 512
191
+ ],
192
+ "empty_cameras": 0,
193
+ "adapt_to_pi_aloha": false,
194
+ "use_delta_joint_actions_aloha": false,
195
+ "tokenizer_max_length": 48,
196
+ "num_steps": 10,
197
+ "use_cache": true,
198
+ "freeze_vision_encoder": true,
199
+ "train_expert_only": false,
200
+ "train_state_proj": true,
201
+ "optimizer_lr": 0.0001,
202
+ "optimizer_betas": [
203
+ 0.9,
204
+ 0.95
205
+ ],
206
+ "optimizer_eps": 1e-08,
207
+ "optimizer_weight_decay": 1e-10,
208
+ "optimizer_grad_clip_norm": 10,
209
+ "scheduler_warmup_steps": 1000,
210
+ "scheduler_decay_steps": 100000,
211
+ "scheduler_decay_lr": 2.5e-06,
212
+ "vlm_model_name": "/projects/extern/kisski/kisski-spath/dir.project/VLA_Imit/LiteVLA-MS/CP/SmolVLM2-500M-Video-Instruct",
213
+ "load_vlm_weights": true,
214
+ "add_image_special_tokens": false,
215
+ "attention_mode": "cross_attn",
216
+ "prefix_length": 0,
217
+ "pad_language_to": "max_length",
218
+ "num_expert_layers": 0,
219
+ "num_vlm_layers": 16,
220
+ "self_attn_every_n_layers": 2,
221
+ "expert_width_multiplier": 0.75,
222
+ "min_period": 0.004,
223
+ "max_period": 4.0,
224
+ "of_path": null
225
+ },
226
+ "output_dir": "outputs/train/2026-04-06/15-19-49_m3_20eps_base",
227
+ "job_name": "m3_20eps_base",
228
+ "resume": false,
229
+ "seed": 42,
230
+ "num_workers": 8,
231
+ "batch_size": 64,
232
+ "steps": 100000,
233
+ "eval_freq": 20000,
234
+ "log_freq": 200,
235
+ "save_checkpoint": true,
236
+ "save_freq": 10000,
237
+ "use_policy_training_preset": true,
238
+ "optimizer": {
239
+ "type": "adamw",
240
+ "lr": 0.0001,
241
+ "weight_decay": 1e-10,
242
+ "grad_clip_norm": 10,
243
+ "betas": [
244
+ 0.9,
245
+ 0.95
246
+ ],
247
+ "eps": 1e-08
248
+ },
249
+ "scheduler": {
250
+ "type": "cosine_decay_with_warmup",
251
+ "num_warmup_steps": 1000,
252
+ "num_decay_steps": 100000,
253
+ "peak_lr": 0.0001,
254
+ "decay_lr": 2.5e-06
255
+ },
256
+ "eval": {
257
+ "n_episodes": 50,
258
+ "batch_size": 50,
259
+ "use_async_envs": false
260
+ },
261
+ "wandb": {
262
+ "enable": true,
263
+ "disable_artifact": true,
264
+ "project": "LiteVLA-MS",
265
+ "entity": "Robotics_VLA",
266
+ "notes": null,
267
+ "run_id": null,
268
+ "mode": "offline"
269
+ }
270
+ }