Robotics
LeRobot
Safetensors
molmoact2
qualiadev commited on
Commit
7f5749a
·
verified ·
1 Parent(s): b2b1817

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +128 -0
  3. model.safetensors +3 -0
  4. train_config.json +266 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: JonathanGiegold/cylinder-in-hole-70
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: molmoact2
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - lerobot
9
+ - molmoact2
10
+ - robotics
11
+ ---
12
+
13
+ # Model Card for molmoact2
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
+ lerobot-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
+ 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 +1,128 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "molmoact2",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.images.front": {
6
+ "type": "VISUAL",
7
+ "shape": [
8
+ 3,
9
+ 480,
10
+ 640
11
+ ]
12
+ },
13
+ "observation.images.side": {
14
+ "type": "VISUAL",
15
+ "shape": [
16
+ 3,
17
+ 480,
18
+ 640
19
+ ]
20
+ },
21
+ "observation.state": {
22
+ "type": "STATE",
23
+ "shape": [
24
+ 6
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "use_peft": false,
39
+ "push_to_hub": true,
40
+ "repo_id": "qualiadev/molmoact2-so100-101-cylinder-in-hole-70-3ecfb2d3",
41
+ "private": null,
42
+ "tags": null,
43
+ "license": null,
44
+ "pretrained_path": null,
45
+ "checkpoint_path": "allenai/MolmoAct2-SO100_101",
46
+ "checkpoint_revision": null,
47
+ "checkpoint_force_download": false,
48
+ "chunk_size": 30,
49
+ "n_action_steps": 30,
50
+ "action_mode": "both",
51
+ "inference_action_mode": null,
52
+ "discrete_action_tokenizer": "allenai/MolmoAct2-FAST-Tokenizer",
53
+ "discrete_generation_max_steps": null,
54
+ "norm_tag": null,
55
+ "setup_type": "",
56
+ "control_mode": "",
57
+ "image_keys": [],
58
+ "normalize_language": true,
59
+ "add_setup_tokens": true,
60
+ "add_control_tokens": true,
61
+ "normalize_gripper": true,
62
+ "num_state_tokens": 256,
63
+ "max_sequence_length": null,
64
+ "expected_max_action_dim": 32,
65
+ "num_flow_timesteps": 8,
66
+ "flow_matching_cutoff": 1.0,
67
+ "flow_matching_time_offset": 0.001,
68
+ "flow_matching_time_scale": 0.999,
69
+ "flow_matching_beta_alpha": 1.0,
70
+ "flow_matching_beta_beta": 1.5,
71
+ "num_inference_steps": null,
72
+ "mask_action_dim_padding": true,
73
+ "enable_inference_cuda_graph": true,
74
+ "per_episode_seed": false,
75
+ "eval_seed": null,
76
+ "rtc_config": null,
77
+ "enable_lora_vlm": false,
78
+ "lora_rank": 64,
79
+ "lora_alpha": 16,
80
+ "lora_dropout": 0.05,
81
+ "lora_bias": "none",
82
+ "enable_lora_action_expert": false,
83
+ "enable_knowledge_insulation": false,
84
+ "freeze_embedding": true,
85
+ "train_action_expert_only": false,
86
+ "gradient_checkpointing": false,
87
+ "model_dtype": "bfloat16",
88
+ "softmax_auxiliary_loss": true,
89
+ "softmax_auxiliary_loss_scale": 0.0001,
90
+ "discrete_loss_token_weighting": "root_subsegments_root_tokens",
91
+ "optimizer_lr": 1e-05,
92
+ "optimizer_vit_lr": 5e-06,
93
+ "optimizer_connector_lr": 5e-06,
94
+ "optimizer_action_expert_lr": 5e-05,
95
+ "optimizer_betas": [
96
+ 0.9,
97
+ 0.95
98
+ ],
99
+ "optimizer_eps": 1e-06,
100
+ "optimizer_weight_decay": 0.0,
101
+ "optimizer_grad_clip_norm": 1.0,
102
+ "scheduler_warmup_steps": 200,
103
+ "scheduler_decay_steps": null,
104
+ "scheduler_decay_lr": 1e-06,
105
+ "normalization_mapping": {
106
+ "VISUAL": "IDENTITY",
107
+ "STATE": "QUANTILES",
108
+ "ACTION": "QUANTILES"
109
+ },
110
+ "dataset_feature_names": {
111
+ "action": [
112
+ "shoulder_pan.pos",
113
+ "shoulder_lift.pos",
114
+ "elbow_flex.pos",
115
+ "wrist_flex.pos",
116
+ "wrist_roll.pos",
117
+ "gripper.pos"
118
+ ],
119
+ "observation.state": [
120
+ "shoulder_pan.pos",
121
+ "shoulder_lift.pos",
122
+ "elbow_flex.pos",
123
+ "wrist_flex.pos",
124
+ "wrist_roll.pos",
125
+ "gripper.pos"
126
+ ]
127
+ }
128
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa05010b2829ab4497287e32f60b45ebd989e898053bea868e9b56621b9d9a6e
3
+ size 10884573720
train_config.json ADDED
@@ -0,0 +1,266 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "JonathanGiegold/cylinder-in-hole-70",
4
+ "root": "/home/ubuntu/vla/model/molmoact2/dataset/JonathanGiegold/cylinder-in-hole-70",
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
+ "affine": {
62
+ "weight": 1.0,
63
+ "type": "RandomAffine",
64
+ "kwargs": {
65
+ "degrees": [
66
+ -5.0,
67
+ 5.0
68
+ ],
69
+ "translate": [
70
+ 0.05,
71
+ 0.05
72
+ ]
73
+ }
74
+ }
75
+ }
76
+ },
77
+ "revision": null,
78
+ "use_imagenet_stats": true,
79
+ "video_backend": "pyav",
80
+ "return_uint8": false,
81
+ "streaming": false
82
+ },
83
+ "env": null,
84
+ "policy": {
85
+ "type": "molmoact2",
86
+ "n_obs_steps": 1,
87
+ "input_features": {
88
+ "observation.images.front": {
89
+ "type": "VISUAL",
90
+ "shape": [
91
+ 3,
92
+ 480,
93
+ 640
94
+ ]
95
+ },
96
+ "observation.images.side": {
97
+ "type": "VISUAL",
98
+ "shape": [
99
+ 3,
100
+ 480,
101
+ 640
102
+ ]
103
+ },
104
+ "observation.state": {
105
+ "type": "STATE",
106
+ "shape": [
107
+ 6
108
+ ]
109
+ }
110
+ },
111
+ "output_features": {
112
+ "action": {
113
+ "type": "ACTION",
114
+ "shape": [
115
+ 6
116
+ ]
117
+ }
118
+ },
119
+ "device": "cuda",
120
+ "use_amp": false,
121
+ "use_peft": false,
122
+ "push_to_hub": true,
123
+ "repo_id": "qualiadev/molmoact2-so100-101-cylinder-in-hole-70-3ecfb2d3",
124
+ "private": null,
125
+ "tags": null,
126
+ "license": null,
127
+ "pretrained_path": null,
128
+ "checkpoint_path": "allenai/MolmoAct2-SO100_101",
129
+ "checkpoint_revision": null,
130
+ "checkpoint_force_download": false,
131
+ "chunk_size": 30,
132
+ "n_action_steps": 30,
133
+ "action_mode": "both",
134
+ "inference_action_mode": null,
135
+ "discrete_action_tokenizer": "allenai/MolmoAct2-FAST-Tokenizer",
136
+ "discrete_generation_max_steps": null,
137
+ "norm_tag": null,
138
+ "setup_type": "",
139
+ "control_mode": "",
140
+ "image_keys": [],
141
+ "normalize_language": true,
142
+ "add_setup_tokens": true,
143
+ "add_control_tokens": true,
144
+ "normalize_gripper": true,
145
+ "num_state_tokens": 256,
146
+ "max_sequence_length": null,
147
+ "expected_max_action_dim": 32,
148
+ "num_flow_timesteps": 8,
149
+ "flow_matching_cutoff": 1.0,
150
+ "flow_matching_time_offset": 0.001,
151
+ "flow_matching_time_scale": 0.999,
152
+ "flow_matching_beta_alpha": 1.0,
153
+ "flow_matching_beta_beta": 1.5,
154
+ "num_inference_steps": null,
155
+ "mask_action_dim_padding": true,
156
+ "enable_inference_cuda_graph": true,
157
+ "per_episode_seed": false,
158
+ "eval_seed": null,
159
+ "rtc_config": null,
160
+ "enable_lora_vlm": false,
161
+ "lora_rank": 64,
162
+ "lora_alpha": 16,
163
+ "lora_dropout": 0.05,
164
+ "lora_bias": "none",
165
+ "enable_lora_action_expert": false,
166
+ "enable_knowledge_insulation": false,
167
+ "freeze_embedding": true,
168
+ "train_action_expert_only": false,
169
+ "gradient_checkpointing": false,
170
+ "model_dtype": "bfloat16",
171
+ "softmax_auxiliary_loss": true,
172
+ "softmax_auxiliary_loss_scale": 0.0001,
173
+ "discrete_loss_token_weighting": "root_subsegments_root_tokens",
174
+ "optimizer_lr": 1e-05,
175
+ "optimizer_vit_lr": 5e-06,
176
+ "optimizer_connector_lr": 5e-06,
177
+ "optimizer_action_expert_lr": 5e-05,
178
+ "optimizer_betas": [
179
+ 0.9,
180
+ 0.95
181
+ ],
182
+ "optimizer_eps": 1e-06,
183
+ "optimizer_weight_decay": 0.0,
184
+ "optimizer_grad_clip_norm": 1.0,
185
+ "scheduler_warmup_steps": 200,
186
+ "scheduler_decay_steps": null,
187
+ "scheduler_decay_lr": 1e-06,
188
+ "normalization_mapping": {
189
+ "VISUAL": "IDENTITY",
190
+ "STATE": "QUANTILES",
191
+ "ACTION": "QUANTILES"
192
+ },
193
+ "dataset_feature_names": {
194
+ "action": [
195
+ "shoulder_pan.pos",
196
+ "shoulder_lift.pos",
197
+ "elbow_flex.pos",
198
+ "wrist_flex.pos",
199
+ "wrist_roll.pos",
200
+ "gripper.pos"
201
+ ],
202
+ "observation.state": [
203
+ "shoulder_pan.pos",
204
+ "shoulder_lift.pos",
205
+ "elbow_flex.pos",
206
+ "wrist_flex.pos",
207
+ "wrist_roll.pos",
208
+ "gripper.pos"
209
+ ]
210
+ }
211
+ },
212
+ "reward_model": null,
213
+ "output_dir": "/ephemeral/output/run1",
214
+ "job_name": "c8df4e56-65e6-419a-a1f6-27d63ecfb2d3",
215
+ "resume": false,
216
+ "seed": 1000,
217
+ "cudnn_deterministic": false,
218
+ "num_workers": 4,
219
+ "batch_size": 1,
220
+ "prefetch_factor": 4,
221
+ "persistent_workers": true,
222
+ "steps": 10000000,
223
+ "eval_freq": 20000,
224
+ "log_freq": 10,
225
+ "tolerance_s": 0.0001,
226
+ "save_checkpoint": true,
227
+ "save_freq": 20000,
228
+ "use_policy_training_preset": true,
229
+ "optimizer": {
230
+ "type": "adamw",
231
+ "lr": 1e-05,
232
+ "weight_decay": 0.0,
233
+ "grad_clip_norm": 1.0,
234
+ "betas": [
235
+ 0.9,
236
+ 0.95
237
+ ],
238
+ "eps": 1e-06
239
+ },
240
+ "scheduler": {
241
+ "type": "molmoact2_cosine_decay_with_warmup",
242
+ "num_warmup_steps": 200,
243
+ "num_decay_steps": null,
244
+ "peak_lr": 1e-05,
245
+ "decay_lr": 1e-06
246
+ },
247
+ "eval": {
248
+ "n_episodes": 50,
249
+ "batch_size": 19,
250
+ "use_async_envs": true
251
+ },
252
+ "wandb": {
253
+ "enable": false,
254
+ "disable_artifact": false,
255
+ "project": "lerobot",
256
+ "entity": null,
257
+ "notes": null,
258
+ "run_id": null,
259
+ "mode": null,
260
+ "add_tags": true
261
+ },
262
+ "peft": null,
263
+ "sample_weighting": null,
264
+ "rename_map": {},
265
+ "checkpoint_path": null
266
+ }