Hugging Face's logo

tmeynier
/
test_hepha_model_2

Robotics
LeRobot
Safetensors
dot
Model card Files
xet
Community
tmeynier commited on
Commit
49ebe23
·
verified ·
1 Parent(s): 9dbd412

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +114 -0
  3. model.safetensors +3 -0
  4. train_config.json +225 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: tmeynier/test_hepha_record_22
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: dot
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - lerobot
9
+ - robotics
10
+ - dot
11
+ ---
12
+
13
+ # Model Card for dot
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+ _Model type not recognized — please update this template._
19
+
20
+
21
+ This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
22
+ See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
23
+
24
+ ---
25
+
26
+ ## How to Get Started with the Model
27
+
28
+ For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
29
+ Below is the short version on how to train and run inference/eval:
30
+
31
+ ### Train from scratch
32
+
33
+ ```bash
34
+ python -m lerobot.scripts.train \
35
+ --dataset.repo_id=${HF_USER}/<dataset> \
36
+ --policy.type=act \
37
+ --output_dir=outputs/train/<desired_policy_repo_id> \
38
+ --job_name=lerobot_training \
39
+ --policy.device=cuda \
40
+ --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
41
+ --wandb.enable=true
42
+ ```
43
+
44
+ *Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`.*
45
+
46
+ ### Evaluate the policy/run inference
47
+
48
+ ```bash
49
+ python -m lerobot.record \
50
+ --robot.type=so100_follower \
51
+ --dataset.repo_id=<hf_user>/eval_<dataset> \
52
+ --policy.path=<hf_user>/<desired_policy_repo_id> \
53
+ --episodes=10
54
+ ```
55
+
56
+ Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
57
+
58
+ ---
59
+
60
+ ## Model Details
61
+
62
+ * **License:** apache-2.0
config.json ADDED
@@ -0,0 +1,114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "dot",
3
+ "n_obs_steps": 3,
4
+ "normalization_mapping": {
5
+ "VISUAL": "MEAN_STD",
6
+ "STATE": "MIN_MAX",
7
+ "ENV": "MIN_MAX",
8
+ "ACTION": "MIN_MAX"
9
+ },
10
+ "input_features": {
11
+ "observation.state": {
12
+ "type": "STATE",
13
+ "shape": [
14
+ 5
15
+ ]
16
+ },
17
+ "observation.images.top_view": {
18
+ "type": "VISUAL",
19
+ "shape": [
20
+ 3,
21
+ 264,
22
+ 264
23
+ ]
24
+ },
25
+ "observation.images.gripper_cam": {
26
+ "type": "VISUAL",
27
+ "shape": [
28
+ 3,
29
+ 264,
30
+ 264
31
+ ]
32
+ }
33
+ },
34
+ "output_features": {
35
+ "action": {
36
+ "type": "ACTION",
37
+ "shape": [
38
+ 5
39
+ ]
40
+ }
41
+ },
42
+ "device": "cuda",
43
+ "use_amp": false,
44
+ "push_to_hub": true,
45
+ "repo_id": "tmeynier/test_hepha_model_2",
46
+ "private": null,
47
+ "tags": null,
48
+ "license": null,
49
+ "pretrained_path": null,
50
+ "train_horizon": 20,
51
+ "inference_horizon": 20,
52
+ "lookback_obs_steps": 10,
53
+ "lookback_aug": 5,
54
+ "override_dataset_stats": false,
55
+ "new_dataset_stats": {
56
+ "action": {
57
+ "max": [
58
+ 0.225,
59
+ 0.271,
60
+ 0.175,
61
+ 0.7854,
62
+ 0.0333
63
+ ],
64
+ "min": [
65
+ -0.225,
66
+ -0.271,
67
+ -0.175,
68
+ -1.5708,
69
+ -0.0333
70
+ ]
71
+ },
72
+ "observation.state": {
73
+ "max": [
74
+ 0.225,
75
+ 0.271,
76
+ 0.175,
77
+ 0.7854,
78
+ 0.0333
79
+ ],
80
+ "min": [
81
+ -0.225,
82
+ -0.271,
83
+ -0.175,
84
+ -1.5708,
85
+ -0.0333
86
+ ]
87
+ }
88
+ },
89
+ "vision_backbone": "resnet18",
90
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
91
+ "pre_norm": true,
92
+ "lora_rank": 20,
93
+ "merge_lora": false,
94
+ "dim_model": 384,
95
+ "n_heads": 8,
96
+ "dim_feedforward": 1536,
97
+ "n_decoder_layers": 8,
98
+ "rescale_shape": [
99
+ 264,
100
+ 264
101
+ ],
102
+ "crop_scale": 0.8,
103
+ "state_noise": 0.01,
104
+ "noise_decay": 0.999995,
105
+ "dropout": 0.1,
106
+ "alpha": 0.75,
107
+ "train_alpha": 0.9,
108
+ "predict_every_n": 1,
109
+ "return_every_n": 1,
110
+ "optimizer_lr": 0.0001,
111
+ "optimizer_min_lr": 0.0001,
112
+ "optimizer_lr_cycle_steps": 300000,
113
+ "optimizer_weight_decay": 1e-05
114
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c68571b657986a5d4d32a37bebc1514f54318feb0d32b22af04fa38e783d6eb6
3
+ size 124275968
train_config.json ADDED
@@ -0,0 +1,225 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "tmeynier/test_hepha_record_22",
4
+ "root": null,
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": "dot",
70
+ "n_obs_steps": 3,
71
+ "normalization_mapping": {
72
+ "VISUAL": "MEAN_STD",
73
+ "STATE": "MIN_MAX",
74
+ "ENV": "MIN_MAX",
75
+ "ACTION": "MIN_MAX"
76
+ },
77
+ "input_features": {
78
+ "observation.state": {
79
+ "type": "STATE",
80
+ "shape": [
81
+ 5
82
+ ]
83
+ },
84
+ "observation.images.top_view": {
85
+ "type": "VISUAL",
86
+ "shape": [
87
+ 3,
88
+ 264,
89
+ 264
90
+ ]
91
+ },
92
+ "observation.images.gripper_cam": {
93
+ "type": "VISUAL",
94
+ "shape": [
95
+ 3,
96
+ 264,
97
+ 264
98
+ ]
99
+ }
100
+ },
101
+ "output_features": {
102
+ "action": {
103
+ "type": "ACTION",
104
+ "shape": [
105
+ 5
106
+ ]
107
+ }
108
+ },
109
+ "device": "cuda",
110
+ "use_amp": false,
111
+ "push_to_hub": true,
112
+ "repo_id": "tmeynier/test_hepha_model_2",
113
+ "private": null,
114
+ "tags": null,
115
+ "license": null,
116
+ "pretrained_path": null,
117
+ "train_horizon": 20,
118
+ "inference_horizon": 20,
119
+ "lookback_obs_steps": 10,
120
+ "lookback_aug": 5,
121
+ "override_dataset_stats": false,
122
+ "new_dataset_stats": {
123
+ "action": {
124
+ "max": [
125
+ 0.225,
126
+ 0.271,
127
+ 0.175,
128
+ 0.7854,
129
+ 0.0333
130
+ ],
131
+ "min": [
132
+ -0.225,
133
+ -0.271,
134
+ -0.175,
135
+ -1.5708,
136
+ -0.0333
137
+ ]
138
+ },
139
+ "observation.state": {
140
+ "max": [
141
+ 0.225,
142
+ 0.271,
143
+ 0.175,
144
+ 0.7854,
145
+ 0.0333
146
+ ],
147
+ "min": [
148
+ -0.225,
149
+ -0.271,
150
+ -0.175,
151
+ -1.5708,
152
+ -0.0333
153
+ ]
154
+ }
155
+ },
156
+ "vision_backbone": "resnet18",
157
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
158
+ "pre_norm": true,
159
+ "lora_rank": 20,
160
+ "merge_lora": false,
161
+ "dim_model": 384,
162
+ "n_heads": 8,
163
+ "dim_feedforward": 1536,
164
+ "n_decoder_layers": 8,
165
+ "rescale_shape": [
166
+ 264,
167
+ 264
168
+ ],
169
+ "crop_scale": 0.8,
170
+ "state_noise": 0.01,
171
+ "noise_decay": 0.999995,
172
+ "dropout": 0.1,
173
+ "alpha": 0.75,
174
+ "train_alpha": 0.9,
175
+ "predict_every_n": 1,
176
+ "return_every_n": 1,
177
+ "optimizer_lr": 0.0001,
178
+ "optimizer_min_lr": 0.0001,
179
+ "optimizer_lr_cycle_steps": 300000,
180
+ "optimizer_weight_decay": 1e-05
181
+ },
182
+ "output_dir": "outputs/train/hepha_test_train_2",
183
+ "job_name": "hepha_test_train_2",
184
+ "resume": false,
185
+ "seed": 1000,
186
+ "num_workers": 4,
187
+ "batch_size": 8,
188
+ "steps": 100000,
189
+ "eval_freq": 20000,
190
+ "log_freq": 200,
191
+ "save_checkpoint": true,
192
+ "save_freq": 20000,
193
+ "use_policy_training_preset": true,
194
+ "optimizer": {
195
+ "type": "adamw",
196
+ "lr": 0.0001,
197
+ "weight_decay": 1e-05,
198
+ "grad_clip_norm": 10.0,
199
+ "betas": [
200
+ 0.9,
201
+ 0.999
202
+ ],
203
+ "eps": 1e-08
204
+ },
205
+ "scheduler": {
206
+ "type": "cosine_annealing",
207
+ "num_warmup_steps": 0,
208
+ "min_lr": 0.0001,
209
+ "T_max": 300000
210
+ },
211
+ "eval": {
212
+ "n_episodes": 50,
213
+ "batch_size": 50,
214
+ "use_async_envs": false
215
+ },
216
+ "wandb": {
217
+ "enable": true,
218
+ "disable_artifact": false,
219
+ "project": "lerobot",
220
+ "entity": null,
221
+ "notes": null,
222
+ "run_id": "xa3k1v5c",
223
+ "mode": null
224
+ }
225
+ }