Hugging Face's logo

andabi
/
M15

Robotics
Safetensors
lerobot
act
model_hub_mixin
pytorch_model_hub_mixin
Model card Files Community
andabi commited on
Commit
5a62c3d
·
verified ·
1 Parent(s): 560b74d

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. README.md +12 -0
  2. config.json +55 -0
  3. config.yaml +182 -0
  4. model.safetensors +3 -0
README.md ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: lerobot
3
+ tags:
4
+ - act
5
+ - model_hub_mixin
6
+ - pytorch_model_hub_mixin
7
+ - robotics
8
+ ---
9
+
10
+ This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:
11
+ - Library: https://github.com/huggingface/lerobot
12
+ - Docs: [More Information Needed]
config.json ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "chunk_size": 30,
3
+ "dim_feedforward": 3200,
4
+ "dim_model": 512,
5
+ "dropout": 0.1,
6
+ "feedforward_activation": "relu",
7
+ "input_normalization_modes": {
8
+ "observation.images.bird": "mean_std",
9
+ "observation.images.wrist_left": "mean_std",
10
+ "observation.images.wrist_right": "mean_std",
11
+ "observation.state": "mean_std"
12
+ },
13
+ "input_shapes": {
14
+ "observation.images.bird": [
15
+ 3,
16
+ 480,
17
+ 640
18
+ ],
19
+ "observation.images.wrist_left": [
20
+ 3,
21
+ 480,
22
+ 640
23
+ ],
24
+ "observation.images.wrist_right": [
25
+ 3,
26
+ 480,
27
+ 640
28
+ ],
29
+ "observation.state": [
30
+ 12
31
+ ]
32
+ },
33
+ "kl_weight": 10.0,
34
+ "latent_dim": 32,
35
+ "n_action_steps": 30,
36
+ "n_decoder_layers": 1,
37
+ "n_encoder_layers": 4,
38
+ "n_heads": 8,
39
+ "n_obs_steps": 1,
40
+ "n_vae_encoder_layers": 4,
41
+ "output_normalization_modes": {
42
+ "action": "mean_std"
43
+ },
44
+ "output_shapes": {
45
+ "action": [
46
+ 12
47
+ ]
48
+ },
49
+ "pre_norm": false,
50
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
51
+ "replace_final_stride_with_dilation": false,
52
+ "temporal_ensemble_coeff": null,
53
+ "use_vae": true,
54
+ "vision_backbone": "resnet18"
55
+ }
config.yaml ADDED
@@ -0,0 +1,182 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ resume: false
2
+ device: cuda
3
+ use_amp: false
4
+ seed: 1000
5
+ dataset_repo_id:
6
+ - andabi/D8
7
+ - andabi/D12
8
+ - andabi/D15
9
+ - andabi/D16
10
+ video_backend: pyav
11
+ training:
12
+ offline_steps: 180000
13
+ num_workers: 4
14
+ batch_size: 16
15
+ eval_freq: -1
16
+ log_freq: 100
17
+ save_checkpoint: true
18
+ save_freq: 10000
19
+ online_steps: 0
20
+ online_rollout_n_episodes: 1
21
+ online_rollout_batch_size: 1
22
+ online_steps_between_rollouts: 1
23
+ online_sampling_ratio: 0.5
24
+ online_env_seed: null
25
+ online_buffer_capacity: null
26
+ online_buffer_seed_size: 0
27
+ do_online_rollout_async: false
28
+ image_transforms:
29
+ enable: false
30
+ max_num_transforms: 3
31
+ random_order: false
32
+ brightness:
33
+ weight: 1
34
+ min_max:
35
+ - 0.8
36
+ - 1.2
37
+ contrast:
38
+ weight: 1
39
+ min_max:
40
+ - 0.8
41
+ - 1.2
42
+ saturation:
43
+ weight: 1
44
+ min_max:
45
+ - 0.5
46
+ - 1.5
47
+ hue:
48
+ weight: 1
49
+ min_max:
50
+ - -0.05
51
+ - 0.05
52
+ sharpness:
53
+ weight: 1
54
+ min_max:
55
+ - 0.8
56
+ - 1.2
57
+ lr: 1.0e-05
58
+ lr_backbone: 1.0e-05
59
+ weight_decay: 0.0001
60
+ grad_clip_norm: 10
61
+ delta_timestamps:
62
+ action:
63
+ - 0.0
64
+ - 0.06666666666666667
65
+ - 0.13333333333333333
66
+ - 0.2
67
+ - 0.26666666666666666
68
+ - 0.3333333333333333
69
+ - 0.4
70
+ - 0.4666666666666667
71
+ - 0.5333333333333333
72
+ - 0.6
73
+ - 0.6666666666666666
74
+ - 0.7333333333333333
75
+ - 0.8
76
+ - 0.8666666666666667
77
+ - 0.9333333333333333
78
+ - 1.0
79
+ - 1.0666666666666667
80
+ - 1.1333333333333333
81
+ - 1.2
82
+ - 1.2666666666666666
83
+ - 1.3333333333333333
84
+ - 1.4
85
+ - 1.4666666666666666
86
+ - 1.5333333333333334
87
+ - 1.6
88
+ - 1.6666666666666667
89
+ - 1.7333333333333334
90
+ - 1.8
91
+ - 1.8666666666666667
92
+ - 1.9333333333333333
93
+ eval:
94
+ n_episodes: 50
95
+ batch_size: 50
96
+ use_async_envs: false
97
+ wandb:
98
+ enable: true
99
+ disable_artifact: false
100
+ project: lerobot
101
+ notes: ''
102
+ fps: 15
103
+ env:
104
+ name: real_world
105
+ task: null
106
+ state_dim: 12
107
+ action_dim: 12
108
+ fps: ${fps}
109
+ override_dataset_stats:
110
+ observation.images.bird:
111
+ mean:
112
+ - - - 0.485
113
+ - - - 0.456
114
+ - - - 0.406
115
+ std:
116
+ - - - 0.229
117
+ - - - 0.224
118
+ - - - 0.225
119
+ observation.images.wrist_left:
120
+ mean:
121
+ - - - 0.485
122
+ - - - 0.456
123
+ - - - 0.406
124
+ std:
125
+ - - - 0.229
126
+ - - - 0.224
127
+ - - - 0.225
128
+ observation.images.wrist_right:
129
+ mean:
130
+ - - - 0.485
131
+ - - - 0.456
132
+ - - - 0.406
133
+ std:
134
+ - - - 0.229
135
+ - - - 0.224
136
+ - - - 0.225
137
+ policy:
138
+ name: act
139
+ n_obs_steps: 1
140
+ chunk_size: 30
141
+ n_action_steps: 30
142
+ input_shapes:
143
+ observation.images.bird:
144
+ - 3
145
+ - 480
146
+ - 640
147
+ observation.images.wrist_left:
148
+ - 3
149
+ - 480
150
+ - 640
151
+ observation.images.wrist_right:
152
+ - 3
153
+ - 480
154
+ - 640
155
+ observation.state:
156
+ - ${env.state_dim}
157
+ output_shapes:
158
+ action:
159
+ - ${env.action_dim}
160
+ input_normalization_modes:
161
+ observation.images.bird: mean_std
162
+ observation.images.wrist_left: mean_std
163
+ observation.images.wrist_right: mean_std
164
+ observation.state: mean_std
165
+ output_normalization_modes:
166
+ action: mean_std
167
+ vision_backbone: resnet18
168
+ pretrained_backbone_weights: ResNet18_Weights.IMAGENET1K_V1
169
+ replace_final_stride_with_dilation: false
170
+ pre_norm: false
171
+ dim_model: 512
172
+ n_heads: 8
173
+ dim_feedforward: 3200
174
+ feedforward_activation: relu
175
+ n_encoder_layers: 4
176
+ n_decoder_layers: 1
177
+ use_vae: true
178
+ latent_dim: 32
179
+ n_vae_encoder_layers: 4
180
+ temporal_ensemble_coeff: null
181
+ dropout: 0.1
182
+ kl_weight: 10.0
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f1c423cb443c663e2aaaa8a0514accb1fc62a467941a8b5142d5225ff160474
3
+ size 206463968