Hugging Face's logo

pmarsella
/
diffusion_pusht

Safetensors
Model card Files Community
pmarsella commited on
Commit
c0a724b
·
verified ·
1 Parent(s): 4255f8f

Upload 3 files

Browse files
Files changed (3) hide show
  1. config.json +74 -0
  2. model.safetensors +3 -0
  3. train_config.json +183 -0
config.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "diffusion",
3
+ "n_obs_steps": 2,
4
+ "normalization_mapping": {
5
+ "VISUAL": "MEAN_STD",
6
+ "STATE": "MIN_MAX",
7
+ "ACTION": "MIN_MAX"
8
+ },
9
+ "input_features": {
10
+ "observation.state": {
11
+ "type": "STATE",
12
+ "shape": [
13
+ 6
14
+ ]
15
+ },
16
+ "observation.images.phone": {
17
+ "type": "VISUAL",
18
+ "shape": [
19
+ 3,
20
+ 480,
21
+ 640
22
+ ]
23
+ }
24
+ },
25
+ "output_features": {
26
+ "action": {
27
+ "type": "ACTION",
28
+ "shape": [
29
+ 6
30
+ ]
31
+ }
32
+ },
33
+ "horizon": 16,
34
+ "n_action_steps": 8,
35
+ "drop_n_last_frames": 7,
36
+ "vision_backbone": "resnet18",
37
+ "crop_shape": [
38
+ 84,
39
+ 84
40
+ ],
41
+ "crop_is_random": true,
42
+ "pretrained_backbone_weights": null,
43
+ "use_group_norm": true,
44
+ "spatial_softmax_num_keypoints": 32,
45
+ "use_separate_rgb_encoder_per_camera": false,
46
+ "down_dims": [
47
+ 512,
48
+ 1024,
49
+ 2048
50
+ ],
51
+ "kernel_size": 5,
52
+ "n_groups": 8,
53
+ "diffusion_step_embed_dim": 128,
54
+ "use_film_scale_modulation": true,
55
+ "noise_scheduler_type": "DDPM",
56
+ "num_train_timesteps": 100,
57
+ "beta_schedule": "squaredcos_cap_v2",
58
+ "beta_start": 0.0001,
59
+ "beta_end": 0.02,
60
+ "prediction_type": "epsilon",
61
+ "clip_sample": true,
62
+ "clip_sample_range": 1.0,
63
+ "num_inference_steps": null,
64
+ "do_mask_loss_for_padding": false,
65
+ "optimizer_lr": 0.0001,
66
+ "optimizer_betas": [
67
+ 0.95,
68
+ 0.999
69
+ ],
70
+ "optimizer_eps": 1e-08,
71
+ "optimizer_weight_decay": 1e-06,
72
+ "scheduler_name": "cosine",
73
+ "scheduler_warmup_steps": 500
74
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02522f30f9bc7fdfc197a24377e94a10b3e3ecc0e04ac2b76b39c6fcf1f42a9e
3
+ size 1051837384
train_config.json ADDED
@@ -0,0 +1,183 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "ellen2imagine/pusht_green_same_init2",
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": "diffusion",
69
+ "n_obs_steps": 2,
70
+ "normalization_mapping": {
71
+ "VISUAL": "MEAN_STD",
72
+ "STATE": "MIN_MAX",
73
+ "ACTION": "MIN_MAX"
74
+ },
75
+ "input_features": {
76
+ "observation.state": {
77
+ "type": "STATE",
78
+ "shape": [
79
+ 6
80
+ ]
81
+ },
82
+ "observation.images.phone": {
83
+ "type": "VISUAL",
84
+ "shape": [
85
+ 3,
86
+ 480,
87
+ 640
88
+ ]
89
+ }
90
+ },
91
+ "output_features": {
92
+ "action": {
93
+ "type": "ACTION",
94
+ "shape": [
95
+ 6
96
+ ]
97
+ }
98
+ },
99
+ "horizon": 16,
100
+ "n_action_steps": 8,
101
+ "drop_n_last_frames": 7,
102
+ "vision_backbone": "resnet18",
103
+ "crop_shape": [
104
+ 84,
105
+ 84
106
+ ],
107
+ "crop_is_random": true,
108
+ "pretrained_backbone_weights": null,
109
+ "use_group_norm": true,
110
+ "spatial_softmax_num_keypoints": 32,
111
+ "use_separate_rgb_encoder_per_camera": false,
112
+ "down_dims": [
113
+ 512,
114
+ 1024,
115
+ 2048
116
+ ],
117
+ "kernel_size": 5,
118
+ "n_groups": 8,
119
+ "diffusion_step_embed_dim": 128,
120
+ "use_film_scale_modulation": true,
121
+ "noise_scheduler_type": "DDPM",
122
+ "num_train_timesteps": 100,
123
+ "beta_schedule": "squaredcos_cap_v2",
124
+ "beta_start": 0.0001,
125
+ "beta_end": 0.02,
126
+ "prediction_type": "epsilon",
127
+ "clip_sample": true,
128
+ "clip_sample_range": 1.0,
129
+ "num_inference_steps": null,
130
+ "do_mask_loss_for_padding": false,
131
+ "optimizer_lr": 0.0001,
132
+ "optimizer_betas": [
133
+ 0.95,
134
+ 0.999
135
+ ],
136
+ "optimizer_eps": 1e-08,
137
+ "optimizer_weight_decay": 1e-06,
138
+ "scheduler_name": "cosine",
139
+ "scheduler_warmup_steps": 500
140
+ },
141
+ "output_dir": "outputs/train/push_t_same_init_diffusion",
142
+ "job_name": "act_koch_test",
143
+ "resume": false,
144
+ "device": "cuda",
145
+ "use_amp": true,
146
+ "seed": 1000,
147
+ "num_workers": 32,
148
+ "batch_size": 8,
149
+ "steps": 50000,
150
+ "eval_freq": 20000,
151
+ "log_freq": 200,
152
+ "save_checkpoint": true,
153
+ "save_freq": 2000,
154
+ "use_policy_training_preset": true,
155
+ "optimizer": {
156
+ "type": "adam",
157
+ "lr": 0.0001,
158
+ "weight_decay": 1e-06,
159
+ "grad_clip_norm": 10.0,
160
+ "betas": [
161
+ 0.95,
162
+ 0.999
163
+ ],
164
+ "eps": 1e-08
165
+ },
166
+ "scheduler": {
167
+ "type": "diffuser",
168
+ "num_warmup_steps": 500,
169
+ "name": "cosine"
170
+ },
171
+ "eval": {
172
+ "n_episodes": 50,
173
+ "batch_size": 50,
174
+ "use_async_envs": false
175
+ },
176
+ "wandb": {
177
+ "enable": false,
178
+ "disable_artifact": false,
179
+ "project": "lerobot",
180
+ "entity": null,
181
+ "notes": null
182
+ }
183
+ }