Robotics
LeRobot
Safetensors
molmoact2
autmoate commited on
Commit
6d851a7
·
verified ·
1 Parent(s): 8f5ae4e

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +60 -1
  2. config.json +140 -0
  3. model.safetensors +3 -0
  4. train_config.json +278 -0
README.md CHANGED
@@ -1,3 +1,62 @@
1
  ---
2
- license: mit
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ datasets: autmoate/coffee_task_by_task_1_put_mug_large
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: molmoact2
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - robotics
9
+ - molmoact2
10
+ - lerobot
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 +1,140 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "molmoact2",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.top": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 720,
16
+ 1280
17
+ ]
18
+ },
19
+ "observation.images.side": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 720,
24
+ 1280
25
+ ]
26
+ },
27
+ "observation.images.wrist": {
28
+ "type": "VISUAL",
29
+ "shape": [
30
+ 3,
31
+ 720,
32
+ 1280
33
+ ]
34
+ }
35
+ },
36
+ "output_features": {
37
+ "action": {
38
+ "type": "ACTION",
39
+ "shape": [
40
+ 6
41
+ ]
42
+ }
43
+ },
44
+ "device": "cuda",
45
+ "use_amp": false,
46
+ "use_peft": false,
47
+ "push_to_hub": true,
48
+ "repo_id": "autmoate/molmoact2_put_mug_large_both_bs16_20k",
49
+ "private": null,
50
+ "tags": null,
51
+ "license": null,
52
+ "pretrained_path": null,
53
+ "checkpoint_path": "allenai/MolmoAct2-SO100_101",
54
+ "checkpoint_revision": null,
55
+ "checkpoint_force_download": false,
56
+ "chunk_size": 10,
57
+ "n_action_steps": 10,
58
+ "action_mode": "both",
59
+ "inference_action_mode": null,
60
+ "discrete_action_tokenizer": "allenai/MolmoAct2-FAST-Tokenizer",
61
+ "discrete_generation_max_steps": null,
62
+ "norm_tag": null,
63
+ "setup_type": "single SO-100 follower arm with top, wrist, and side RGB cameras",
64
+ "control_mode": "absolute joint pose",
65
+ "image_keys": [
66
+ "observation.images.top",
67
+ "observation.images.wrist",
68
+ "observation.images.side"
69
+ ],
70
+ "normalize_language": true,
71
+ "add_setup_tokens": true,
72
+ "add_control_tokens": true,
73
+ "normalize_gripper": true,
74
+ "num_state_tokens": 256,
75
+ "max_sequence_length": null,
76
+ "expected_max_action_dim": 32,
77
+ "num_flow_timesteps": 8,
78
+ "flow_matching_cutoff": 1.0,
79
+ "flow_matching_time_offset": 0.001,
80
+ "flow_matching_time_scale": 0.999,
81
+ "flow_matching_beta_alpha": 1.0,
82
+ "flow_matching_beta_beta": 1.5,
83
+ "num_inference_steps": null,
84
+ "mask_action_dim_padding": true,
85
+ "enable_inference_cuda_graph": true,
86
+ "per_episode_seed": false,
87
+ "eval_seed": null,
88
+ "rtc_config": null,
89
+ "enable_lora_vlm": true,
90
+ "lora_rank": 64,
91
+ "lora_alpha": 16,
92
+ "lora_dropout": 0.05,
93
+ "lora_bias": "none",
94
+ "enable_lora_action_expert": false,
95
+ "enable_knowledge_insulation": false,
96
+ "freeze_embedding": true,
97
+ "train_action_expert_only": false,
98
+ "gradient_checkpointing": true,
99
+ "model_dtype": "bfloat16",
100
+ "softmax_auxiliary_loss": true,
101
+ "softmax_auxiliary_loss_scale": 0.0001,
102
+ "discrete_loss_token_weighting": "root_subsegments_root_tokens",
103
+ "optimizer_lr": 1e-05,
104
+ "optimizer_vit_lr": 5e-06,
105
+ "optimizer_connector_lr": 5e-06,
106
+ "optimizer_action_expert_lr": 5e-05,
107
+ "optimizer_betas": [
108
+ 0.9,
109
+ 0.95
110
+ ],
111
+ "optimizer_eps": 1e-06,
112
+ "optimizer_weight_decay": 0.0,
113
+ "optimizer_grad_clip_norm": 1.0,
114
+ "scheduler_warmup_steps": 200,
115
+ "scheduler_decay_steps": null,
116
+ "scheduler_decay_lr": 1e-06,
117
+ "normalization_mapping": {
118
+ "VISUAL": "IDENTITY",
119
+ "STATE": "QUANTILES",
120
+ "ACTION": "QUANTILES"
121
+ },
122
+ "dataset_feature_names": {
123
+ "action": [
124
+ "shoulder_pan.pos",
125
+ "shoulder_lift.pos",
126
+ "elbow_flex.pos",
127
+ "wrist_flex.pos",
128
+ "wrist_roll.pos",
129
+ "gripper.pos"
130
+ ],
131
+ "observation.state": [
132
+ "shoulder_pan.pos",
133
+ "shoulder_lift.pos",
134
+ "elbow_flex.pos",
135
+ "wrist_flex.pos",
136
+ "wrist_roll.pos",
137
+ "gripper.pos"
138
+ ]
139
+ }
140
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d68560690fc049e762876082eb53eb4aa1bc133bb72a033b6f371b03c4bd577
3
+ size 11483634784
train_config.json ADDED
@@ -0,0 +1,278 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "autmoate/coffee_task_by_task_1_put_mug_large",
4
+ "root": null,
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": true,
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.state": {
89
+ "type": "STATE",
90
+ "shape": [
91
+ 6
92
+ ]
93
+ },
94
+ "observation.images.top": {
95
+ "type": "VISUAL",
96
+ "shape": [
97
+ 3,
98
+ 720,
99
+ 1280
100
+ ]
101
+ },
102
+ "observation.images.side": {
103
+ "type": "VISUAL",
104
+ "shape": [
105
+ 3,
106
+ 720,
107
+ 1280
108
+ ]
109
+ },
110
+ "observation.images.wrist": {
111
+ "type": "VISUAL",
112
+ "shape": [
113
+ 3,
114
+ 720,
115
+ 1280
116
+ ]
117
+ }
118
+ },
119
+ "output_features": {
120
+ "action": {
121
+ "type": "ACTION",
122
+ "shape": [
123
+ 6
124
+ ]
125
+ }
126
+ },
127
+ "device": "cuda",
128
+ "use_amp": false,
129
+ "use_peft": false,
130
+ "push_to_hub": true,
131
+ "repo_id": "autmoate/molmoact2_put_mug_large_both_bs16_20k",
132
+ "private": null,
133
+ "tags": null,
134
+ "license": null,
135
+ "pretrained_path": null,
136
+ "checkpoint_path": "allenai/MolmoAct2-SO100_101",
137
+ "checkpoint_revision": null,
138
+ "checkpoint_force_download": false,
139
+ "chunk_size": 10,
140
+ "n_action_steps": 10,
141
+ "action_mode": "both",
142
+ "inference_action_mode": null,
143
+ "discrete_action_tokenizer": "allenai/MolmoAct2-FAST-Tokenizer",
144
+ "discrete_generation_max_steps": null,
145
+ "norm_tag": null,
146
+ "setup_type": "single SO-100 follower arm with top, wrist, and side RGB cameras",
147
+ "control_mode": "absolute joint pose",
148
+ "image_keys": [
149
+ "observation.images.top",
150
+ "observation.images.wrist",
151
+ "observation.images.side"
152
+ ],
153
+ "normalize_language": true,
154
+ "add_setup_tokens": true,
155
+ "add_control_tokens": true,
156
+ "normalize_gripper": true,
157
+ "num_state_tokens": 256,
158
+ "max_sequence_length": null,
159
+ "expected_max_action_dim": 32,
160
+ "num_flow_timesteps": 8,
161
+ "flow_matching_cutoff": 1.0,
162
+ "flow_matching_time_offset": 0.001,
163
+ "flow_matching_time_scale": 0.999,
164
+ "flow_matching_beta_alpha": 1.0,
165
+ "flow_matching_beta_beta": 1.5,
166
+ "num_inference_steps": null,
167
+ "mask_action_dim_padding": true,
168
+ "enable_inference_cuda_graph": true,
169
+ "per_episode_seed": false,
170
+ "eval_seed": null,
171
+ "rtc_config": null,
172
+ "enable_lora_vlm": true,
173
+ "lora_rank": 64,
174
+ "lora_alpha": 16,
175
+ "lora_dropout": 0.05,
176
+ "lora_bias": "none",
177
+ "enable_lora_action_expert": false,
178
+ "enable_knowledge_insulation": false,
179
+ "freeze_embedding": true,
180
+ "train_action_expert_only": false,
181
+ "gradient_checkpointing": true,
182
+ "model_dtype": "bfloat16",
183
+ "softmax_auxiliary_loss": true,
184
+ "softmax_auxiliary_loss_scale": 0.0001,
185
+ "discrete_loss_token_weighting": "root_subsegments_root_tokens",
186
+ "optimizer_lr": 1e-05,
187
+ "optimizer_vit_lr": 5e-06,
188
+ "optimizer_connector_lr": 5e-06,
189
+ "optimizer_action_expert_lr": 5e-05,
190
+ "optimizer_betas": [
191
+ 0.9,
192
+ 0.95
193
+ ],
194
+ "optimizer_eps": 1e-06,
195
+ "optimizer_weight_decay": 0.0,
196
+ "optimizer_grad_clip_norm": 1.0,
197
+ "scheduler_warmup_steps": 200,
198
+ "scheduler_decay_steps": null,
199
+ "scheduler_decay_lr": 1e-06,
200
+ "normalization_mapping": {
201
+ "VISUAL": "IDENTITY",
202
+ "STATE": "QUANTILES",
203
+ "ACTION": "QUANTILES"
204
+ },
205
+ "dataset_feature_names": {
206
+ "action": [
207
+ "shoulder_pan.pos",
208
+ "shoulder_lift.pos",
209
+ "elbow_flex.pos",
210
+ "wrist_flex.pos",
211
+ "wrist_roll.pos",
212
+ "gripper.pos"
213
+ ],
214
+ "observation.state": [
215
+ "shoulder_pan.pos",
216
+ "shoulder_lift.pos",
217
+ "elbow_flex.pos",
218
+ "wrist_flex.pos",
219
+ "wrist_roll.pos",
220
+ "gripper.pos"
221
+ ]
222
+ }
223
+ },
224
+ "reward_model": null,
225
+ "output_dir": "outputs/molmoact2_both_vlm_lora_bs16_20k",
226
+ "job_name": "molmoact2_both_vlm_lora_bs16_20k",
227
+ "resume": false,
228
+ "seed": 1000,
229
+ "cudnn_deterministic": false,
230
+ "num_workers": 4,
231
+ "batch_size": 16,
232
+ "prefetch_factor": 4,
233
+ "persistent_workers": true,
234
+ "steps": 20000,
235
+ "eval_freq": -1,
236
+ "log_freq": 20,
237
+ "tolerance_s": 0.0001,
238
+ "save_checkpoint": true,
239
+ "save_freq": 20000,
240
+ "use_policy_training_preset": true,
241
+ "optimizer": {
242
+ "type": "adamw",
243
+ "lr": 1e-05,
244
+ "weight_decay": 0.0,
245
+ "grad_clip_norm": 1.0,
246
+ "betas": [
247
+ 0.9,
248
+ 0.95
249
+ ],
250
+ "eps": 1e-06
251
+ },
252
+ "scheduler": {
253
+ "type": "molmoact2_cosine_decay_with_warmup",
254
+ "num_warmup_steps": 200,
255
+ "num_decay_steps": null,
256
+ "peak_lr": 1e-05,
257
+ "decay_lr": 1e-06
258
+ },
259
+ "eval": {
260
+ "n_episodes": 50,
261
+ "batch_size": 33,
262
+ "use_async_envs": true
263
+ },
264
+ "wandb": {
265
+ "enable": false,
266
+ "disable_artifact": false,
267
+ "project": "lerobot",
268
+ "entity": null,
269
+ "notes": null,
270
+ "run_id": null,
271
+ "mode": null,
272
+ "add_tags": true
273
+ },
274
+ "peft": null,
275
+ "sample_weighting": null,
276
+ "rename_map": {},
277
+ "checkpoint_path": null
278
+ }