Hugging Face's logo

HumanoidTeam
/
binary_cube_rainbow_local_processing_joint_fixes_100k

Safetensors
Model card Files Community
diogormalmeida commited on
Commit
01fc9ac
·
verified ·
1 Parent(s): 32b8845

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +62 -0
  2. model.safetensors +3 -0
  3. train_config.json +167 -0
config.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "normalization_mapping": {
5
+ "VISUAL": "MEAN_STD",
6
+ "STATE": "MEAN_STD",
7
+ "ACTION": "MEAN_STD"
8
+ },
9
+ "input_features": {
10
+ "observation.state": {
11
+ "type": "STATE",
12
+ "shape": [
13
+ 16
14
+ ]
15
+ },
16
+ "observation.image.head": {
17
+ "type": "VISUAL",
18
+ "shape": [
19
+ 3,
20
+ 480,
21
+ 848
22
+ ]
23
+ },
24
+ "observation.image.wrist_right": {
25
+ "type": "VISUAL",
26
+ "shape": [
27
+ 3,
28
+ 480,
29
+ 848
30
+ ]
31
+ }
32
+ },
33
+ "output_features": {
34
+ "action": {
35
+ "type": "ACTION",
36
+ "shape": [
37
+ 16
38
+ ]
39
+ }
40
+ },
41
+ "chunk_size": 100,
42
+ "n_action_steps": 100,
43
+ "vision_backbone": "resnet18",
44
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
45
+ "replace_final_stride_with_dilation": false,
46
+ "pre_norm": false,
47
+ "dim_model": 512,
48
+ "n_heads": 8,
49
+ "dim_feedforward": 3200,
50
+ "feedforward_activation": "relu",
51
+ "n_encoder_layers": 4,
52
+ "n_decoder_layers": 1,
53
+ "use_vae": true,
54
+ "latent_dim": 32,
55
+ "n_vae_encoder_layers": 4,
56
+ "temporal_ensemble_coeff": null,
57
+ "dropout": 0.1,
58
+ "kl_weight": 10.0,
59
+ "optimizer_lr": 1e-05,
60
+ "optimizer_weight_decay": 0.0001,
61
+ "optimizer_lr_backbone": 1e-05
62
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03e232bcde62cd9a5f7439f919267373d347d48e156492eff5617bdf622075de
3
+ size 206783288
train_config.json ADDED
@@ -0,0 +1,167 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "HumanoidTeam/rby_binary_cube_v5",
4
+ "episodes": null,
5
+ "image_transforms": {
6
+ "enable": false,
7
+ "max_num_transforms": 3,
8
+ "random_order": false,
9
+ "tfs": {
10
+ "brightness": {
11
+ "weight": 1.0,
12
+ "type": "ColorJitter",
13
+ "kwargs": {
14
+ "brightness": [
15
+ 0.8,
16
+ 1.2
17
+ ]
18
+ }
19
+ },
20
+ "contrast": {
21
+ "weight": 1.0,
22
+ "type": "ColorJitter",
23
+ "kwargs": {
24
+ "contrast": [
25
+ 0.8,
26
+ 1.2
27
+ ]
28
+ }
29
+ },
30
+ "saturation": {
31
+ "weight": 1.0,
32
+ "type": "ColorJitter",
33
+ "kwargs": {
34
+ "saturation": [
35
+ 0.5,
36
+ 1.5
37
+ ]
38
+ }
39
+ },
40
+ "hue": {
41
+ "weight": 1.0,
42
+ "type": "ColorJitter",
43
+ "kwargs": {
44
+ "hue": [
45
+ -0.05,
46
+ 0.05
47
+ ]
48
+ }
49
+ },
50
+ "sharpness": {
51
+ "weight": 1.0,
52
+ "type": "SharpnessJitter",
53
+ "kwargs": {
54
+ "sharpness": [
55
+ 0.5,
56
+ 1.5
57
+ ]
58
+ }
59
+ }
60
+ }
61
+ },
62
+ "local_files_only": false,
63
+ "use_imagenet_stats": true,
64
+ "video_backend": "pyav"
65
+ },
66
+ "env": null,
67
+ "policy": {
68
+ "type": "act",
69
+ "n_obs_steps": 1,
70
+ "normalization_mapping": {
71
+ "VISUAL": "MEAN_STD",
72
+ "STATE": "MEAN_STD",
73
+ "ACTION": "MEAN_STD"
74
+ },
75
+ "input_features": {
76
+ "observation.state": {
77
+ "type": "STATE",
78
+ "shape": [
79
+ 16
80
+ ]
81
+ },
82
+ "observation.image.head": {
83
+ "type": "VISUAL",
84
+ "shape": [
85
+ 3,
86
+ 480,
87
+ 848
88
+ ]
89
+ },
90
+ "observation.image.wrist_right": {
91
+ "type": "VISUAL",
92
+ "shape": [
93
+ 3,
94
+ 480,
95
+ 848
96
+ ]
97
+ }
98
+ },
99
+ "output_features": {
100
+ "action": {
101
+ "type": "ACTION",
102
+ "shape": [
103
+ 16
104
+ ]
105
+ }
106
+ },
107
+ "chunk_size": 100,
108
+ "n_action_steps": 100,
109
+ "vision_backbone": "resnet18",
110
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
111
+ "replace_final_stride_with_dilation": false,
112
+ "pre_norm": false,
113
+ "dim_model": 512,
114
+ "n_heads": 8,
115
+ "dim_feedforward": 3200,
116
+ "feedforward_activation": "relu",
117
+ "n_encoder_layers": 4,
118
+ "n_decoder_layers": 1,
119
+ "use_vae": true,
120
+ "latent_dim": 32,
121
+ "n_vae_encoder_layers": 4,
122
+ "temporal_ensemble_coeff": null,
123
+ "dropout": 0.1,
124
+ "kl_weight": 10.0,
125
+ "optimizer_lr": 1e-05,
126
+ "optimizer_weight_decay": 0.0001,
127
+ "optimizer_lr_backbone": 1e-05
128
+ },
129
+ "output_dir": "outputs/train/2025-03-10/12-59-05_act",
130
+ "job_name": "act",
131
+ "resume": false,
132
+ "device": "cuda",
133
+ "use_amp": false,
134
+ "seed": 1000,
135
+ "num_workers": 4,
136
+ "batch_size": 8,
137
+ "steps": 100000,
138
+ "eval_freq": 20000,
139
+ "log_freq": 200,
140
+ "save_checkpoint": true,
141
+ "save_freq": 10000,
142
+ "use_policy_training_preset": true,
143
+ "optimizer": {
144
+ "type": "adamw",
145
+ "lr": 1e-05,
146
+ "weight_decay": 0.0001,
147
+ "grad_clip_norm": 10.0,
148
+ "betas": [
149
+ 0.9,
150
+ 0.999
151
+ ],
152
+ "eps": 1e-08
153
+ },
154
+ "scheduler": null,
155
+ "eval": {
156
+ "n_episodes": 50,
157
+ "batch_size": 50,
158
+ "use_async_envs": false
159
+ },
160
+ "wandb": {
161
+ "enable": true,
162
+ "disable_artifact": false,
163
+ "project": "'amazon_demo_rainbow'",
164
+ "entity": null,
165
+ "notes": null
166
+ }
167
+ }