Hugging Face's logo

robot-learning-group47
/
eval2_smolvla_vqa_coadapt

Safetensors
Model card Files
xet
 Community
nkim7 commited on
Commit
dc55817
·
verified ·
1 Parent(s): 0164eff

Root: step010000 (final checkpoint)

Browse files
Files changed (7) hide show
  1. config.json +117 -0
  2. model.safetensors +3 -0
  3. policy_postprocessor.json +32 -0
  4. policy_postprocessor_step_0_unnormalizer_processor.safetensors +3 -0
  5. policy_preprocessor.json +105 -0
  6. policy_preprocessor_step_5_normalizer_processor.safetensors +3 -0
  7. train_config.json +221 -0
config.json ADDED
@@ -0,0 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "smolvla",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.camera1": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 256,
16
+ 256
17
+ ]
18
+ },
19
+ "observation.images.camera2": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 256,
24
+ 256
25
+ ]
26
+ },
27
+ "observation.images.camera3": {
28
+ "type": "VISUAL",
29
+ "shape": [
30
+ 3,
31
+ 256,
32
+ 256
33
+ ]
34
+ },
35
+ "observation.images.empty_camera_0": {
36
+ "type": "VISUAL",
37
+ "shape": [
38
+ 3,
39
+ 480,
40
+ 640
41
+ ]
42
+ },
43
+ "observation.images.empty_camera_1": {
44
+ "type": "VISUAL",
45
+ "shape": [
46
+ 3,
47
+ 480,
48
+ 640
49
+ ]
50
+ }
51
+ },
52
+ "output_features": {
53
+ "action": {
54
+ "type": "ACTION",
55
+ "shape": [
56
+ 6
57
+ ]
58
+ }
59
+ },
60
+ "device": "cuda",
61
+ "use_amp": true,
62
+ "use_peft": false,
63
+ "push_to_hub": true,
64
+ "repo_id": "robot-learning-group47/smolvla-eval2-pathC",
65
+ "private": null,
66
+ "tags": null,
67
+ "license": null,
68
+ "pretrained_path": "/home/shadeform/outputs/smolvla-eval2-pathC-init",
69
+ "chunk_size": 50,
70
+ "n_action_steps": 50,
71
+ "normalization_mapping": {
72
+ "VISUAL": "IDENTITY",
73
+ "STATE": "MEAN_STD",
74
+ "ACTION": "MEAN_STD"
75
+ },
76
+ "max_state_dim": 32,
77
+ "max_action_dim": 32,
78
+ "resize_imgs_with_padding": [
79
+ 256,
80
+ 256
81
+ ],
82
+ "empty_cameras": 2,
83
+ "adapt_to_pi_aloha": false,
84
+ "use_delta_joint_actions_aloha": false,
85
+ "tokenizer_max_length": 48,
86
+ "num_steps": 10,
87
+ "use_cache": true,
88
+ "freeze_vision_encoder": true,
89
+ "train_expert_only": false,
90
+ "train_state_proj": true,
91
+ "optimizer_lr": 0.0001,
92
+ "optimizer_betas": [
93
+ 0.9,
94
+ 0.95
95
+ ],
96
+ "optimizer_eps": 1e-08,
97
+ "optimizer_weight_decay": 1e-10,
98
+ "optimizer_grad_clip_norm": 10.0,
99
+ "scheduler_warmup_steps": 1000,
100
+ "scheduler_decay_steps": 30000,
101
+ "scheduler_decay_lr": 2.5e-06,
102
+ "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
103
+ "load_vlm_weights": true,
104
+ "add_image_special_tokens": false,
105
+ "attention_mode": "cross_attn",
106
+ "prefix_length": 0,
107
+ "pad_language_to": "max_length",
108
+ "num_expert_layers": 0,
109
+ "num_vlm_layers": 16,
110
+ "self_attn_every_n_layers": 2,
111
+ "expert_width_multiplier": 0.75,
112
+ "min_period": 0.004,
113
+ "max_period": 4.0,
114
+ "rtc_config": null,
115
+ "compile_model": false,
116
+ "compile_mode": "max-autotune"
117
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e45de5d2745de44186f186a4919abeeee2a797c531d83d93595505609c769416
3
+ size 906712520
policy_postprocessor.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "policy_postprocessor",
3
+ "steps": [
4
+ {
5
+ "registry_name": "unnormalizer_processor",
6
+ "config": {
7
+ "eps": 1e-08,
8
+ "features": {
9
+ "action": {
10
+ "type": "ACTION",
11
+ "shape": [
12
+ 6
13
+ ]
14
+ }
15
+ },
16
+ "norm_map": {
17
+ "VISUAL": "IDENTITY",
18
+ "STATE": "MEAN_STD",
19
+ "ACTION": "MEAN_STD"
20
+ }
21
+ },
22
+ "state_file": "policy_postprocessor_step_0_unnormalizer_processor.safetensors"
23
+ },
24
+ {
25
+ "registry_name": "device_processor",
26
+ "config": {
27
+ "device": "cpu",
28
+ "float_dtype": null
29
+ }
30
+ }
31
+ ]
32
+ }
policy_postprocessor_step_0_unnormalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21963dba500d0c035464b590466400e6012903b9cdda03ed3a23fe6f2eba8f61
3
+ size 6560
policy_preprocessor.json ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "policy_preprocessor",
3
+ "steps": [
4
+ {
5
+ "registry_name": "rename_observations_processor",
6
+ "config": {
7
+ "rename_map": {
8
+ "observation.images.front": "observation.images.camera1"
9
+ }
10
+ }
11
+ },
12
+ {
13
+ "registry_name": "to_batch_processor",
14
+ "config": {}
15
+ },
16
+ {
17
+ "registry_name": "smolvla_new_line_processor",
18
+ "config": {}
19
+ },
20
+ {
21
+ "registry_name": "tokenizer_processor",
22
+ "config": {
23
+ "max_length": 48,
24
+ "task_key": "task",
25
+ "padding_side": "right",
26
+ "padding": "max_length",
27
+ "truncation": true,
28
+ "tokenizer_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct"
29
+ }
30
+ },
31
+ {
32
+ "registry_name": "device_processor",
33
+ "config": {
34
+ "device": "cuda",
35
+ "float_dtype": null
36
+ }
37
+ },
38
+ {
39
+ "registry_name": "normalizer_processor",
40
+ "config": {
41
+ "eps": 1e-08,
42
+ "features": {
43
+ "observation.state": {
44
+ "type": "STATE",
45
+ "shape": [
46
+ 6
47
+ ]
48
+ },
49
+ "observation.images.camera1": {
50
+ "type": "VISUAL",
51
+ "shape": [
52
+ 3,
53
+ 256,
54
+ 256
55
+ ]
56
+ },
57
+ "observation.images.camera2": {
58
+ "type": "VISUAL",
59
+ "shape": [
60
+ 3,
61
+ 256,
62
+ 256
63
+ ]
64
+ },
65
+ "observation.images.camera3": {
66
+ "type": "VISUAL",
67
+ "shape": [
68
+ 3,
69
+ 256,
70
+ 256
71
+ ]
72
+ },
73
+ "observation.images.empty_camera_0": {
74
+ "type": "VISUAL",
75
+ "shape": [
76
+ 3,
77
+ 480,
78
+ 640
79
+ ]
80
+ },
81
+ "observation.images.empty_camera_1": {
82
+ "type": "VISUAL",
83
+ "shape": [
84
+ 3,
85
+ 480,
86
+ 640
87
+ ]
88
+ },
89
+ "action": {
90
+ "type": "ACTION",
91
+ "shape": [
92
+ 6
93
+ ]
94
+ }
95
+ },
96
+ "norm_map": {
97
+ "VISUAL": "IDENTITY",
98
+ "STATE": "MEAN_STD",
99
+ "ACTION": "MEAN_STD"
100
+ }
101
+ },
102
+ "state_file": "policy_preprocessor_step_5_normalizer_processor.safetensors"
103
+ }
104
+ ]
105
+ }
policy_preprocessor_step_5_normalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21963dba500d0c035464b590466400e6012903b9cdda03ed3a23fe6f2eba8f61
3
+ size 6560
train_config.json ADDED
@@ -0,0 +1,221 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "robot-learning-group47/eval2_compositional_augmented",
4
+ "root": null,
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": true,
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
+ 0.7,
17
+ 1.3
18
+ ]
19
+ }
20
+ },
21
+ "contrast": {
22
+ "weight": 1.0,
23
+ "type": "ColorJitter",
24
+ "kwargs": {
25
+ "contrast": [
26
+ 0.7,
27
+ 1.3
28
+ ]
29
+ }
30
+ },
31
+ "sharpness": {
32
+ "weight": 0.5,
33
+ "type": "SharpnessJitter",
34
+ "kwargs": {
35
+ "sharpness": [
36
+ 0.5,
37
+ 1.5
38
+ ]
39
+ }
40
+ }
41
+ }
42
+ },
43
+ "revision": "main",
44
+ "use_imagenet_stats": true,
45
+ "video_backend": "pyav",
46
+ "streaming": false
47
+ },
48
+ "env": null,
49
+ "policy": {
50
+ "type": "smolvla",
51
+ "n_obs_steps": 1,
52
+ "input_features": {
53
+ "observation.state": {
54
+ "type": "STATE",
55
+ "shape": [
56
+ 6
57
+ ]
58
+ },
59
+ "observation.images.camera1": {
60
+ "type": "VISUAL",
61
+ "shape": [
62
+ 3,
63
+ 256,
64
+ 256
65
+ ]
66
+ },
67
+ "observation.images.camera2": {
68
+ "type": "VISUAL",
69
+ "shape": [
70
+ 3,
71
+ 256,
72
+ 256
73
+ ]
74
+ },
75
+ "observation.images.camera3": {
76
+ "type": "VISUAL",
77
+ "shape": [
78
+ 3,
79
+ 256,
80
+ 256
81
+ ]
82
+ },
83
+ "observation.images.empty_camera_0": {
84
+ "type": "VISUAL",
85
+ "shape": [
86
+ 3,
87
+ 480,
88
+ 640
89
+ ]
90
+ },
91
+ "observation.images.empty_camera_1": {
92
+ "type": "VISUAL",
93
+ "shape": [
94
+ 3,
95
+ 480,
96
+ 640
97
+ ]
98
+ }
99
+ },
100
+ "output_features": {
101
+ "action": {
102
+ "type": "ACTION",
103
+ "shape": [
104
+ 6
105
+ ]
106
+ }
107
+ },
108
+ "device": "cuda",
109
+ "use_amp": true,
110
+ "use_peft": false,
111
+ "push_to_hub": true,
112
+ "repo_id": "robot-learning-group47/smolvla-eval2-pathC",
113
+ "private": null,
114
+ "tags": null,
115
+ "license": null,
116
+ "pretrained_path": "/home/shadeform/outputs/smolvla-eval2-pathC-init",
117
+ "chunk_size": 50,
118
+ "n_action_steps": 50,
119
+ "normalization_mapping": {
120
+ "VISUAL": "IDENTITY",
121
+ "STATE": "MEAN_STD",
122
+ "ACTION": "MEAN_STD"
123
+ },
124
+ "max_state_dim": 32,
125
+ "max_action_dim": 32,
126
+ "resize_imgs_with_padding": [
127
+ 256,
128
+ 256
129
+ ],
130
+ "empty_cameras": 2,
131
+ "adapt_to_pi_aloha": false,
132
+ "use_delta_joint_actions_aloha": false,
133
+ "tokenizer_max_length": 48,
134
+ "num_steps": 10,
135
+ "use_cache": true,
136
+ "freeze_vision_encoder": true,
137
+ "train_expert_only": false,
138
+ "train_state_proj": true,
139
+ "optimizer_lr": 0.0001,
140
+ "optimizer_betas": [
141
+ 0.9,
142
+ 0.95
143
+ ],
144
+ "optimizer_eps": 1e-08,
145
+ "optimizer_weight_decay": 1e-10,
146
+ "optimizer_grad_clip_norm": 10.0,
147
+ "scheduler_warmup_steps": 1000,
148
+ "scheduler_decay_steps": 30000,
149
+ "scheduler_decay_lr": 2.5e-06,
150
+ "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
151
+ "load_vlm_weights": true,
152
+ "add_image_special_tokens": false,
153
+ "attention_mode": "cross_attn",
154
+ "prefix_length": 0,
155
+ "pad_language_to": "max_length",
156
+ "num_expert_layers": 0,
157
+ "num_vlm_layers": 16,
158
+ "self_attn_every_n_layers": 2,
159
+ "expert_width_multiplier": 0.75,
160
+ "min_period": 0.004,
161
+ "max_period": 4.0,
162
+ "rtc_config": null,
163
+ "compile_model": false,
164
+ "compile_mode": "max-autotune"
165
+ },
166
+ "output_dir": "/home/shadeform/outputs/smolvla-eval2-pathC",
167
+ "job_name": "smolvla-eval2-pathC",
168
+ "resume": false,
169
+ "seed": 1000,
170
+ "num_workers": 8,
171
+ "batch_size": 32,
172
+ "steps": 10000,
173
+ "eval_freq": 10000,
174
+ "log_freq": 50,
175
+ "tolerance_s": 0.0001,
176
+ "save_checkpoint": true,
177
+ "save_freq": 2000,
178
+ "use_policy_training_preset": true,
179
+ "optimizer": {
180
+ "type": "adamw",
181
+ "lr": 0.0001,
182
+ "weight_decay": 1e-10,
183
+ "grad_clip_norm": 10.0,
184
+ "betas": [
185
+ 0.9,
186
+ 0.95
187
+ ],
188
+ "eps": 1e-08
189
+ },
190
+ "scheduler": {
191
+ "type": "cosine_decay_with_warmup",
192
+ "num_warmup_steps": 1000,
193
+ "num_decay_steps": 30000,
194
+ "peak_lr": 0.0001,
195
+ "decay_lr": 2.5e-06
196
+ },
197
+ "eval": {
198
+ "n_episodes": 50,
199
+ "batch_size": 50,
200
+ "use_async_envs": false
201
+ },
202
+ "wandb": {
203
+ "enable": true,
204
+ "disable_artifact": false,
205
+ "project": "lerobot",
206
+ "entity": null,
207
+ "notes": null,
208
+ "run_id": "564tukj1",
209
+ "mode": null
210
+ },
211
+ "peft": null,
212
+ "use_rabc": false,
213
+ "rabc_progress_path": null,
214
+ "rabc_kappa": 0.01,
215
+ "rabc_epsilon": 1e-06,
216
+ "rabc_head_mode": "sparse",
217
+ "rename_map": {
218
+ "observation.images.front": "observation.images.camera1"
219
+ },
220
+ "checkpoint_path": null
221
+ }