Hugging Face's logo

iantc104
/
av_real_open_box_v1_eye

Robotics
Safetensors
lerobot
act
model_hub_mixin
pytorch_model_hub_mixin
Model card Files Community
iantc104 commited on
Commit
12c8977
·
verified ·
1 Parent(s): 34530d5

Upload YAML configuration

Browse files
Files changed (1) hide show
  1. config.yaml +170 -0
config.yaml ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ resume: false
2
+ device: cuda
3
+ use_amp: false
4
+ seed: 0
5
+ dataset_repo_id: iantc104/gaze_real_open_box_v1
6
+ video_backend: pyav
7
+ training:
8
+ offline_steps: 15625
9
+ num_workers: 8
10
+ batch_size: 16
11
+ eval_freq: -1
12
+ log_freq: 100
13
+ save_checkpoint: true
14
+ save_freq: 1000
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: 2.5e-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
+ - 0.030303030303030304
61
+ - 0.06060606060606061
62
+ - 0.09090909090909091
63
+ - 0.12121212121212122
64
+ - 0.15151515151515152
65
+ - 0.18181818181818182
66
+ - 0.21212121212121213
67
+ - 0.24242424242424243
68
+ - 0.2727272727272727
69
+ - 0.30303030303030304
70
+ - 0.3333333333333333
71
+ - 0.36363636363636365
72
+ - 0.3939393939393939
73
+ - 0.42424242424242425
74
+ - 0.45454545454545453
75
+ - 0.48484848484848486
76
+ - 0.5151515151515151
77
+ - 0.5454545454545454
78
+ - 0.5757575757575758
79
+ - 0.6060606060606061
80
+ - 0.6363636363636364
81
+ - 0.6666666666666666
82
+ - 0.696969696969697
83
+ - 0.7272727272727273
84
+ - 0.7575757575757576
85
+ - 0.7878787878787878
86
+ - 0.8181818181818182
87
+ - 0.8484848484848485
88
+ - 0.8787878787878788
89
+ - 0.9090909090909091
90
+ - 0.9393939393939394
91
+ - 0.9696969696969697
92
+ eval:
93
+ n_episodes: 1
94
+ batch_size: 1
95
+ use_async_envs: false
96
+ wandb:
97
+ enable: true
98
+ disable_artifact: true
99
+ project: av_aloha_real
100
+ notes: ''
101
+ fps: 33
102
+ env:
103
+ name: real_world
104
+ task: null
105
+ state_dim: 21
106
+ action_dim: 21
107
+ fps: ${fps}
108
+ override_dataset_stats:
109
+ observation.images.left_eye_cam:
110
+ mean:
111
+ - - - 0.485
112
+ - - - 0.456
113
+ - - - 0.406
114
+ std:
115
+ - - - 0.229
116
+ - - - 0.224
117
+ - - - 0.225
118
+ observation.images.right_eye_cam:
119
+ mean:
120
+ - - - 0.485
121
+ - - - 0.456
122
+ - - - 0.406
123
+ std:
124
+ - - - 0.229
125
+ - - - 0.224
126
+ - - - 0.225
127
+ policy:
128
+ name: act
129
+ n_obs_steps: 1
130
+ chunk_size: 33
131
+ n_action_steps: 33
132
+ input_shapes:
133
+ observation.images.left_eye_cam:
134
+ - 3
135
+ - 480
136
+ - 640
137
+ observation.images.right_eye_cam:
138
+ - 3
139
+ - 480
140
+ - 640
141
+ observation.state:
142
+ - ${env.state_dim}
143
+ output_shapes:
144
+ action:
145
+ - ${env.action_dim}
146
+ input_normalization_modes:
147
+ observation.images.left_eye_cam: mean_std
148
+ observation.images.right_eye_cam: mean_std
149
+ observation.state: mean_std
150
+ output_normalization_modes:
151
+ action: mean_std
152
+ image_size:
153
+ - 480
154
+ - 640
155
+ vision_backbone: resnet18
156
+ pretrained_backbone_weights: ResNet18_Weights.IMAGENET1K_V1
157
+ replace_final_stride_with_dilation: false
158
+ pre_norm: false
159
+ dim_model: 512
160
+ n_heads: 8
161
+ dim_feedforward: 3200
162
+ feedforward_activation: relu
163
+ n_encoder_layers: 4
164
+ n_decoder_layers: 1
165
+ use_vae: true
166
+ latent_dim: 32
167
+ n_vae_encoder_layers: 4
168
+ temporal_ensemble_coeff: null
169
+ dropout: 0.1
170
+ kl_weight: 10.0