Hugging Face's logo

continuallearning
/
dit_mt_libero_90_pretrain

Robotics
LeRobot
Safetensors
Model card Files
xet
 Community
1
Ralf Römer commited on
Commit
f0c31d7
·
1 Parent(s): 152bbe9

Initial commit of pretrained policy

Browse files
Files changed (3) hide show
  1. config.json +88 -0
  2. model.safetensors +3 -0
  3. train_config.json +197 -0
config.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "dit_mt",
3
+ "n_obs_steps": 5,
4
+ "normalization_mapping": {
5
+ "VISUAL": "MEAN_STD",
6
+ "STATE": "MIN_MAX",
7
+ "ACTION": "MIN_MAX"
8
+ },
9
+ "input_features": {
10
+ "observation.images.image": {
11
+ "type": "VISUAL",
12
+ "shape": [
13
+ 3,
14
+ 256,
15
+ 256
16
+ ]
17
+ },
18
+ "observation.images.wrist_image": {
19
+ "type": "VISUAL",
20
+ "shape": [
21
+ 3,
22
+ 256,
23
+ 256
24
+ ]
25
+ },
26
+ "observation.state": {
27
+ "type": "STATE",
28
+ "shape": [
29
+ 8
30
+ ]
31
+ },
32
+ "observation.state.joint": {
33
+ "type": "STATE",
34
+ "shape": [
35
+ 7
36
+ ]
37
+ }
38
+ },
39
+ "output_features": {
40
+ "action": {
41
+ "type": "ACTION",
42
+ "shape": [
43
+ 7
44
+ ]
45
+ }
46
+ },
47
+ "device": "cuda",
48
+ "use_amp": false,
49
+ "horizon": 16,
50
+ "n_action_steps": 8,
51
+ "use_pretrained_dino": true,
52
+ "vision_backbone_size": 18,
53
+ "pretrained_backbone_weights": null,
54
+ "avg_pool": true,
55
+ "vision_backbone_norm_name": "group_norm",
56
+ "vision_backbone_norm_num_groups": 16,
57
+ "use_film": true,
58
+ "tokenizer_max_length": 48,
59
+ "freeze_language_pretrained": true,
60
+ "dim_model": 512,
61
+ "n_heads": 8,
62
+ "dropout": 0.1,
63
+ "dim_feedforward": 2048,
64
+ "feedforward_activation": "gelu",
65
+ "n_encoder_layers": 6,
66
+ "n_decoder_layers": 6,
67
+ "noise_scheduler_type": "DDIM",
68
+ "num_train_timesteps": 100,
69
+ "beta_schedule": "squaredcos_cap_v2",
70
+ "beta_start": 0.0001,
71
+ "beta_end": 0.02,
72
+ "prediction_type": "epsilon",
73
+ "clip_sample": true,
74
+ "set_alpha_to_one": true,
75
+ "steps_offset": 0,
76
+ "num_inference_steps": 10,
77
+ "do_mask_loss_for_padding": false,
78
+ "optimizer_lr": 0.0001,
79
+ "optimizer_betas": [
80
+ 0.95,
81
+ 0.999
82
+ ],
83
+ "optimizer_eps": 1e-08,
84
+ "optimizer_weight_decay": 1e-06,
85
+ "optimizer_grad_clip_norm": 10.0,
86
+ "scheduler_name": "cosine",
87
+ "scheduler_warmup_steps": 500
88
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c5a83347ff554525bee55f668302e65ec3d8656282e2b23962af7a741101d2c
3
+ size 804841588
train_config.json ADDED
@@ -0,0 +1,197 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "libero_90_image",
4
+ "root": "./data/datasets/libero_90_image",
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": false,
8
+ "max_num_transforms": 3,
9
+ "random_order": false,
10
+ "tfs": {
11
+ "brightness": {
12
+ "weight": 1.0,
13
+ "type": "ColorJitter",
14
+ "kwargs": {
15
+ "brightness": [
16
+ 0.8,
17
+ 1.2
18
+ ]
19
+ }
20
+ },
21
+ "contrast": {
22
+ "weight": 1.0,
23
+ "type": "ColorJitter",
24
+ "kwargs": {
25
+ "contrast": [
26
+ 0.8,
27
+ 1.2
28
+ ]
29
+ }
30
+ },
31
+ "saturation": {
32
+ "weight": 1.0,
33
+ "type": "ColorJitter",
34
+ "kwargs": {
35
+ "saturation": [
36
+ 0.5,
37
+ 1.5
38
+ ]
39
+ }
40
+ },
41
+ "hue": {
42
+ "weight": 1.0,
43
+ "type": "ColorJitter",
44
+ "kwargs": {
45
+ "hue": [
46
+ -0.05,
47
+ 0.05
48
+ ]
49
+ }
50
+ },
51
+ "sharpness": {
52
+ "weight": 1.0,
53
+ "type": "SharpnessJitter",
54
+ "kwargs": {
55
+ "sharpness": [
56
+ 0.5,
57
+ 1.5
58
+ ]
59
+ }
60
+ }
61
+ }
62
+ },
63
+ "revision": null,
64
+ "use_imagenet_stats": true,
65
+ "video_backend": "pyav"
66
+ },
67
+ "env": null,
68
+ "policy": {
69
+ "type": "dit_mt",
70
+ "n_obs_steps": 5,
71
+ "normalization_mapping": {
72
+ "VISUAL": "MEAN_STD",
73
+ "STATE": "MIN_MAX",
74
+ "ACTION": "MIN_MAX"
75
+ },
76
+ "input_features": {
77
+ "observation.images.image": {
78
+ "type": "VISUAL",
79
+ "shape": [
80
+ 3,
81
+ 256,
82
+ 256
83
+ ]
84
+ },
85
+ "observation.images.wrist_image": {
86
+ "type": "VISUAL",
87
+ "shape": [
88
+ 3,
89
+ 256,
90
+ 256
91
+ ]
92
+ },
93
+ "observation.state": {
94
+ "type": "STATE",
95
+ "shape": [
96
+ 8
97
+ ]
98
+ },
99
+ "observation.state.joint": {
100
+ "type": "STATE",
101
+ "shape": [
102
+ 7
103
+ ]
104
+ }
105
+ },
106
+ "output_features": {
107
+ "action": {
108
+ "type": "ACTION",
109
+ "shape": [
110
+ 7
111
+ ]
112
+ }
113
+ },
114
+ "device": "cuda",
115
+ "use_amp": false,
116
+ "horizon": 16,
117
+ "n_action_steps": 8,
118
+ "use_pretrained_dino": true,
119
+ "vision_backbone_size": 18,
120
+ "pretrained_backbone_weights": null,
121
+ "avg_pool": true,
122
+ "vision_backbone_norm_name": "group_norm",
123
+ "vision_backbone_norm_num_groups": 16,
124
+ "use_film": true,
125
+ "tokenizer_max_length": 48,
126
+ "freeze_language_pretrained": true,
127
+ "dim_model": 512,
128
+ "n_heads": 8,
129
+ "dropout": 0.1,
130
+ "dim_feedforward": 2048,
131
+ "feedforward_activation": "gelu",
132
+ "n_encoder_layers": 6,
133
+ "n_decoder_layers": 6,
134
+ "noise_scheduler_type": "DDIM",
135
+ "num_train_timesteps": 100,
136
+ "beta_schedule": "squaredcos_cap_v2",
137
+ "beta_start": 0.0001,
138
+ "beta_end": 0.02,
139
+ "prediction_type": "epsilon",
140
+ "clip_sample": true,
141
+ "set_alpha_to_one": true,
142
+ "steps_offset": 0,
143
+ "num_inference_steps": 10,
144
+ "do_mask_loss_for_padding": false,
145
+ "optimizer_lr": 0.0001,
146
+ "optimizer_betas": [
147
+ 0.95,
148
+ 0.999
149
+ ],
150
+ "optimizer_eps": 1e-08,
151
+ "optimizer_weight_decay": 1e-06,
152
+ "optimizer_grad_clip_norm": 10.0,
153
+ "scheduler_name": "cosine",
154
+ "scheduler_warmup_steps": 500
155
+ },
156
+ "output_dir": "outputs/dit_mt_libero_90_pretrain",
157
+ "job_name": "dit_mt_libero_90_pretrain",
158
+ "resume": true,
159
+ "seed": 1000,
160
+ "num_workers": 4,
161
+ "batch_size": 32,
162
+ "steps": 200000,
163
+ "eval_freq": 0,
164
+ "log_freq": 500,
165
+ "save_checkpoint": true,
166
+ "save_freq": 100000,
167
+ "use_policy_training_preset": true,
168
+ "optimizer": {
169
+ "type": "adamw",
170
+ "lr": 0.0001,
171
+ "weight_decay": 1e-06,
172
+ "grad_clip_norm": 10.0,
173
+ "betas": [
174
+ 0.95,
175
+ 0.999
176
+ ],
177
+ "eps": 1e-08
178
+ },
179
+ "scheduler": {
180
+ "type": "diffuser",
181
+ "num_warmup_steps": 500,
182
+ "name": "cosine"
183
+ },
184
+ "eval": {
185
+ "n_episodes": 50,
186
+ "batch_size": 50,
187
+ "use_async_envs": false,
188
+ "max_episodes_rendered": 4
189
+ },
190
+ "wandb": {
191
+ "enable": true,
192
+ "disable_artifact": true,
193
+ "project": "paper",
194
+ "entity": "470620104-technical-university-of-munich",
195
+ "notes": null
196
+ }
197
+ }