Hugging Face's logo

Corematic-Europe
/
Strawberries_model

Robotics
Safetensors
Model card Files Community
rbo-corematic commited on
Commit
da8c67a
·
verified ·
1 Parent(s): a578250

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +178 -0
  2. model.safetensors +3 -0
  3. train_config.json +112 -0
config.json ADDED
@@ -0,0 +1,178 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "smolvla",
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
+ 6
14
+ ]
15
+ },
16
+ "observation.images.wrist": {
17
+ "type": "VISUAL",
18
+ "shape": [
19
+ 3,
20
+ 480,
21
+ 640
22
+ ]
23
+ },
24
+ "observation.images.top": {
25
+ "type": "VISUAL",
26
+ "shape": [
27
+ 3,
28
+ 480,
29
+ 640
30
+ ]
31
+ }
32
+ },
33
+ "output_features": {
34
+ "action": {
35
+ "type": "ACTION",
36
+ "shape": [
37
+ 6
38
+ ]
39
+ }
40
+ },
41
+ "chunk_size": 50,
42
+ "n_action_steps": 50,
43
+ "max_state_dim": 6,
44
+ "max_action_dim": 6,
45
+ "resize_imgs_with_padding": [
46
+ 640,
47
+ 480
48
+ ],
49
+ "empty_cameras": 0,
50
+ "adapt_to_pi_aloha": false,
51
+ "use_delta_joint_actions_aloha": false,
52
+ "tokenizer_max_length": 48,
53
+ "num_steps": 10,
54
+ "use_cache": true,
55
+ "freeze_vision_encoder": true,
56
+ "train_expert_only": true,
57
+ "train_state_proj": true,
58
+ "optimizer_lr": 0.0001,
59
+ "optimizer_betas": [
60
+ 0.9,
61
+ 0.95
62
+ ],
63
+ "optimizer_eps": 1e-08,
64
+ "optimizer_weight_decay": 1e-10,
65
+ "optimizer_grad_clip_norm": 10,
66
+ "scheduler_warmup_steps": 1000,
67
+ "scheduler_decay_steps": 30000,
68
+ "scheduler_decay_lr": 2.5e-06,
69
+ "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
70
+ "load_vlm_weights": true,
71
+ "attention_mode": "cross_attn",
72
+ "prefix_length": 0,
73
+ "pad_language_to": "max_length",
74
+ "num_expert_layers": 0,
75
+ "num_vlm_layers": 16,
76
+ "self_attn_every_n_layers": 2,
77
+ "expert_width_multiplier": 0.75,
78
+ "robot_config": {
79
+ "robot_type": "so101_follower",
80
+ "codebase_version": "v2.1",
81
+ "total_episodes": 86,
82
+ "total_frames": 23660,
83
+ "total_tasks": 1,
84
+ "total_videos": 172,
85
+ "total_chunks": 1,
86
+ "chunks_size": 1000,
87
+ "fps": 30,
88
+ "splits": {
89
+ "train": "0:86"
90
+ },
91
+ "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
92
+ "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
93
+ "action_names": [
94
+ "shoulder_pan.pos",
95
+ "shoulder_lift.pos",
96
+ "elbow_flex.pos",
97
+ "wrist_flex.pos",
98
+ "wrist_roll.pos",
99
+ "gripper.pos"
100
+ ],
101
+ "camera_views": ["wrist", "top"],
102
+ "features": {
103
+ "action": {
104
+ "dtype": "float32",
105
+ "shape": [6],
106
+ "names": [
107
+ "shoulder_pan.pos",
108
+ "shoulder_lift.pos",
109
+ "elbow_flex.pos",
110
+ "wrist_flex.pos",
111
+ "wrist_roll.pos",
112
+ "gripper.pos"
113
+ ]
114
+ },
115
+ "observation.state": {
116
+ "dtype": "float32",
117
+ "shape": [6],
118
+ "names": [
119
+ "shoulder_pan.pos",
120
+ "shoulder_lift.pos",
121
+ "elbow_flex.pos",
122
+ "wrist_flex.pos",
123
+ "wrist_roll.pos",
124
+ "gripper.pos"
125
+ ]
126
+ },
127
+ "timestamp": {
128
+ "dtype": "float32",
129
+ "shape": [1],
130
+ "names": null
131
+ },
132
+ "episode_index": {
133
+ "dtype": "int64",
134
+ "shape": [1],
135
+ "names": null
136
+ },
137
+ "frame_index": {
138
+ "dtype": "int64",
139
+ "shape": [1],
140
+ "names": null
141
+ },
142
+ "task_index": {
143
+ "dtype": "int64",
144
+ "shape": [1],
145
+ "names": null
146
+ },
147
+ "index": {
148
+ "dtype": "int64",
149
+ "shape": [1],
150
+ "names": null
151
+ },
152
+ "observation.images.wrist": {
153
+ "dtype": "video",
154
+ "shape": [480, 640, 3],
155
+ "names": ["height", "width", "channels"],
156
+ "info": {
157
+ "video.fps": 30,
158
+ "video.codec": "av1",
159
+ "video.pix_fmt": "yuv420p",
160
+ "video.is_depth_map": false,
161
+ "has_audio": false
162
+ }
163
+ },
164
+ "observation.images.top": {
165
+ "dtype": "video",
166
+ "shape": [480, 640, 3],
167
+ "names": ["height", "width", "channels"],
168
+ "info": {
169
+ "video.fps": 30,
170
+ "video.codec": "av1",
171
+ "video.pix_fmt": "yuv420p",
172
+ "video.is_depth_map": false,
173
+ "has_audio": false
174
+ }
175
+ }
176
+ }
177
+ }
178
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9f15de7415427b7754ad0cc1336e57089a819af17c9755ee408e66f93628d4d
3
+ size 906713296
train_config.json ADDED
@@ -0,0 +1,112 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"batch_size": 64,
2
+ "dataset": {"episodes": null,
3
+ "image_transforms": {"enable": false,
4
+ "max_num_transforms": 3,
5
+ "random_order": false,
6
+ "tfs": {"brightness": {"kwargs": {"brightness": [0.8,
7
+ 1.2]},
8
+ "type": "ColorJitter",
9
+ "weight": 1.0},
10
+ "contrast": {"kwargs": {"contrast": [0.8,
11
+ 1.2]},
12
+ "type": "ColorJitter",
13
+ "weight": 1.0},
14
+ "hue": {"kwargs": {"hue": [-0.05,
15
+ 0.05]},
16
+ "type": "ColorJitter",
17
+ "weight": 1.0},
18
+ "saturation": {"kwargs": {"saturation": [0.5,
19
+ 1.5]},
20
+ "type": "ColorJitter",
21
+ "weight": 1.0},
22
+ "sharpness": {"kwargs": {"sharpness": [0.5,
23
+ 1.5]},
24
+ "type": "SharpnessJitter",
25
+ "weight": 1.0}}},
26
+ "repo_id": "florian-moyen/strawberries_merged",
27
+ "revision": null,
28
+ "root": null,
29
+ "use_imagenet_stats": true,
30
+ "video_backend": "torchcodec"},
31
+ "env": null,
32
+ "eval": {"batch_size": 50, "n_episodes": 50, "use_async_envs": false},
33
+ "eval_freq": 20000,
34
+ "job_name": "strawberries_job",
35
+ "log_freq": 200,
36
+ "num_workers": 4,
37
+ "optimizer": {"betas": [0.9, 0.95],
38
+ "eps": 1e-08,
39
+ "grad_clip_norm": 10.0,
40
+ "lr": 0.0001,
41
+ "type": "adamw",
42
+ "weight_decay": 1e-10},
43
+ "output_dir": "outputs/train/strawberries_model",
44
+ "policy": {"adapt_to_pi_aloha": false,
45
+ "add_image_special_tokens": false,
46
+ "attention_mode": "cross_attn",
47
+ "chunk_size": 50,
48
+ "device": "cuda",
49
+ "empty_cameras": 0,
50
+ "expert_width_multiplier": 0.75,
51
+ "freeze_vision_encoder": true,
52
+ "input_features": {"observation.image": {"shape": [3, 256, 256],
53
+ "type": "VISUAL"},
54
+ "observation.image2": {"shape": [3, 256, 256],
55
+ "type": "VISUAL"},
56
+ "observation.image3": {"shape": [3, 256, 256],
57
+ "type": "VISUAL"},
58
+ "observation.state": {"shape": [6],
59
+ "type": "STATE"}},
60
+ "load_vlm_weights": true,
61
+ "max_action_dim": 32,
62
+ "max_period": 4.0,
63
+ "max_state_dim": 32,
64
+ "min_period": 0.004,
65
+ "n_action_steps": 50,
66
+ "n_obs_steps": 1,
67
+ "normalization_mapping": {"ACTION": "MEAN_STD",
68
+ "STATE": "MEAN_STD",
69
+ "VISUAL": "IDENTITY"},
70
+ "num_expert_layers": 0,
71
+ "num_steps": 10,
72
+ "num_vlm_layers": 16,
73
+ "optimizer_betas": [0.9, 0.95],
74
+ "optimizer_eps": 1e-08,
75
+ "optimizer_grad_clip_norm": 10.0,
76
+ "optimizer_lr": 0.0001,
77
+ "optimizer_weight_decay": 1e-10,
78
+ "output_features": {"action": {"shape": [6],
79
+ "type": "ACTION"}},
80
+ "pad_language_to": "max_length",
81
+ "prefix_length": 0,
82
+ "resize_imgs_with_padding": [512, 512],
83
+ "scheduler_decay_lr": 2.5e-06,
84
+ "scheduler_decay_steps": 30000,
85
+ "scheduler_warmup_steps": 1000,
86
+ "self_attn_every_n_layers": 2,
87
+ "tokenizer_max_length": 48,
88
+ "train_expert_only": true,
89
+ "train_state_proj": true,
90
+ "type": "smolvla",
91
+ "use_amp": false,
92
+ "use_cache": true,
93
+ "use_delta_joint_actions_aloha": false,
94
+ "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct"},
95
+ "resume": false,
96
+ "save_checkpoint": true,
97
+ "save_freq": 20000,
98
+ "scheduler": {"decay_lr": 2.5e-06,
99
+ "num_decay_steps": 30000,
100
+ "num_warmup_steps": 1000,
101
+ "peak_lr": 0.0001,
102
+ "type": "cosine_decay_with_warmup"},
103
+ "seed": 1000,
104
+ "steps": 20000,
105
+ "use_policy_training_preset": true,
106
+ "wandb": {"disable_artifact": false,
107
+ "enable": true,
108
+ "entity": null,
109
+ "mode": null,
110
+ "notes": null,
111
+ "project": "lerobot",
112
+ "run_id": null}}