Hugging Face's logo

andabi
/
M8

Robotics
Safetensors
lerobot
act
model_hub_mixin
pytorch_model_hub_mixin
Model card Files Community
andabi commited on
Commit
44fccb9
·
verified ·
1 Parent(s): 9417f95

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 +178 -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 +1,178 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ resume: false
2
+ device: cuda
3
+ use_amp: false
4
+ seed: 1000
5
+ dataset_repo_id: andabi/D9
6
+ video_backend: pyav
7
+ training:
8
+ offline_steps: 200000
9
+ num_workers: 4
10
+ batch_size: 16
11
+ eval_freq: -1
12
+ log_freq: 100
13
+ save_checkpoint: true
14
+ save_freq: 10000
15
+ online_steps: 0
16
+ online_rollout_n_episodes: 1
17
+ online_rollout_batch_size: 1
18
+ online_steps_between_rollouts: 1
19
+ online_sampling_ratio: 0.5
20
+ online_env_seed: null
21
+ online_buffer_capacity: null
22
+ online_buffer_seed_size: 0
23
+ do_online_rollout_async: false
24
+ image_transforms:
25
+ enable: false
26
+ max_num_transforms: 3
27
+ random_order: false
28
+ brightness:
29
+ weight: 1
30
+ min_max:
31
+ - 0.8
32
+ - 1.2
33
+ contrast:
34
+ weight: 1
35
+ min_max:
36
+ - 0.8
37
+ - 1.2
38
+ saturation:
39
+ weight: 1
40
+ min_max:
41
+ - 0.5
42
+ - 1.5
43
+ hue:
44
+ weight: 1
45
+ min_max:
46
+ - -0.05
47
+ - 0.05
48
+ sharpness:
49
+ weight: 1
50
+ min_max:
51
+ - 0.8
52
+ - 1.2
53
+ lr: 1.0e-05
54
+ lr_backbone: 1.0e-05
55
+ weight_decay: 0.0001
56
+ grad_clip_norm: 10
57
+ delta_timestamps:
58
+ action:
59
+ - 0.0
60
+ - 0.06666666666666667
61
+ - 0.13333333333333333
62
+ - 0.2
63
+ - 0.26666666666666666
64
+ - 0.3333333333333333
65
+ - 0.4
66
+ - 0.4666666666666667
67
+ - 0.5333333333333333
68
+ - 0.6
69
+ - 0.6666666666666666
70
+ - 0.7333333333333333
71
+ - 0.8
72
+ - 0.8666666666666667
73
+ - 0.9333333333333333
74
+ - 1.0
75
+ - 1.0666666666666667
76
+ - 1.1333333333333333
77
+ - 1.2
78
+ - 1.2666666666666666
79
+ - 1.3333333333333333
80
+ - 1.4
81
+ - 1.4666666666666666
82
+ - 1.5333333333333334
83
+ - 1.6
84
+ - 1.6666666666666667
85
+ - 1.7333333333333334
86
+ - 1.8
87
+ - 1.8666666666666667
88
+ - 1.9333333333333333
89
+ eval:
90
+ n_episodes: 50
91
+ batch_size: 50
92
+ use_async_envs: false
93
+ wandb:
94
+ enable: true
95
+ disable_artifact: false
96
+ project: lerobot
97
+ notes: ''
98
+ fps: 15
99
+ env:
100
+ name: real_world
101
+ task: null
102
+ state_dim: 12
103
+ action_dim: 12
104
+ fps: ${fps}
105
+ override_dataset_stats:
106
+ observation.images.bird:
107
+ mean:
108
+ - - - 0.485
109
+ - - - 0.456
110
+ - - - 0.406
111
+ std:
112
+ - - - 0.229
113
+ - - - 0.224
114
+ - - - 0.225
115
+ observation.images.wrist_left:
116
+ mean:
117
+ - - - 0.485
118
+ - - - 0.456
119
+ - - - 0.406
120
+ std:
121
+ - - - 0.229
122
+ - - - 0.224
123
+ - - - 0.225
124
+ observation.images.wrist_right:
125
+ mean:
126
+ - - - 0.485
127
+ - - - 0.456
128
+ - - - 0.406
129
+ std:
130
+ - - - 0.229
131
+ - - - 0.224
132
+ - - - 0.225
133
+ policy:
134
+ name: act
135
+ n_obs_steps: 1
136
+ chunk_size: 30
137
+ n_action_steps: 30
138
+ input_shapes:
139
+ observation.images.bird:
140
+ - 3
141
+ - 480
142
+ - 640
143
+ observation.images.wrist_left:
144
+ - 3
145
+ - 480
146
+ - 640
147
+ observation.images.wrist_right:
148
+ - 3
149
+ - 480
150
+ - 640
151
+ observation.state:
152
+ - ${env.state_dim}
153
+ output_shapes:
154
+ action:
155
+ - ${env.action_dim}
156
+ input_normalization_modes:
157
+ observation.images.bird: mean_std
158
+ observation.images.wrist_left: mean_std
159
+ observation.images.wrist_right: mean_std
160
+ observation.state: mean_std
161
+ output_normalization_modes:
162
+ action: mean_std
163
+ vision_backbone: resnet18
164
+ pretrained_backbone_weights: ResNet18_Weights.IMAGENET1K_V1
165
+ replace_final_stride_with_dilation: false
166
+ pre_norm: false
167
+ dim_model: 512
168
+ n_heads: 8
169
+ dim_feedforward: 3200
170
+ feedforward_activation: relu
171
+ n_encoder_layers: 4
172
+ n_decoder_layers: 1
173
+ use_vae: true
174
+ latent_dim: 32
175
+ n_vae_encoder_layers: 4
176
+ temporal_ensemble_coeff: null
177
+ dropout: 0.1
178
+ kl_weight: 10.0
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1eb4358e08706ef31e8155c2b6caedc439f4be0542a590ea3ae0eb36cd75a9fa
3
+ size 206463968