Hugging Face's logo

Robomaster
/
lbm

Model card Files
xet
Community
Robomaster commited on
Commit
2fc98c3
·
verified ·
1 Parent(s): ad1dcdf

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. .DS_Store +0 -0
  2. config.json +240 -0
  3. model.pt +3 -0
  4. train_config.json +683 -0
.DS_Store ADDED
Binary file (6.15 kB). View file
 
config.json ADDED
@@ -0,0 +1,240 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "lbm_policy",
3
+ "chunk_size": 16,
4
+ "n_action_steps": 8,
5
+ "n_obs_steps": 2,
6
+ "freq": 10.0,
7
+ "use_relative_action": true,
8
+ "use_relative_pose": false,
9
+ "normalization_mapping": {
10
+ "visual": "mean_std",
11
+ "state": "min_max",
12
+ "action_pose_gripper": "min_max"
13
+ },
14
+ "input_features": {
15
+ "observation_images_left_wrist": {
16
+ "shape": [
17
+ 3,
18
+ 224,
19
+ 224
20
+ ],
21
+ "dtype": "visual",
22
+ "fps": null,
23
+ "rep": null,
24
+ "rep_kwargs": null,
25
+ "compose": null
26
+ },
27
+ "observation_images_right_wrist": {
28
+ "shape": [
29
+ 3,
30
+ 224,
31
+ 224
32
+ ],
33
+ "dtype": "visual",
34
+ "fps": null,
35
+ "rep": null,
36
+ "rep_kwargs": null,
37
+ "compose": null
38
+ },
39
+ "observation_images_head": {
40
+ "shape": [
41
+ 3,
42
+ 224,
43
+ 224
44
+ ],
45
+ "dtype": "visual",
46
+ "fps": null,
47
+ "rep": null,
48
+ "rep_kwargs": null,
49
+ "compose": null
50
+ },
51
+ "observation_images_chest": {
52
+ "shape": [
53
+ 3,
54
+ 224,
55
+ 224
56
+ ],
57
+ "dtype": "visual",
58
+ "fps": null,
59
+ "rep": null,
60
+ "rep_kwargs": null,
61
+ "compose": null
62
+ },
63
+ "observation_states": {
64
+ "shape": [
65
+ 34
66
+ ],
67
+ "dtype": "state",
68
+ "fps": null,
69
+ "rep": null,
70
+ "rep_kwargs": null,
71
+ "compose": {
72
+ "observation_states_ee_pose_left": {
73
+ "shape": [
74
+ 9
75
+ ],
76
+ "dtype": "state",
77
+ "fps": null,
78
+ "rep": "rotation_6d",
79
+ "rep_kwargs": null,
80
+ "compose": null
81
+ },
82
+ "observation_states_joint_angle_left": {
83
+ "shape": [
84
+ 7
85
+ ],
86
+ "dtype": "state",
87
+ "fps": null,
88
+ "rep": null,
89
+ "rep_kwargs": null,
90
+ "compose": null
91
+ },
92
+ "observation_states_gripper_left": {
93
+ "shape": [
94
+ 1
95
+ ],
96
+ "dtype": "state",
97
+ "fps": null,
98
+ "rep": null,
99
+ "rep_kwargs": null,
100
+ "compose": null
101
+ },
102
+ "observation_states_ee_pose_right": {
103
+ "shape": [
104
+ 9
105
+ ],
106
+ "dtype": "state",
107
+ "fps": null,
108
+ "rep": "rotation_6d",
109
+ "rep_kwargs": null,
110
+ "compose": null
111
+ },
112
+ "observation_states_joint_angle_right": {
113
+ "shape": [
114
+ 7
115
+ ],
116
+ "dtype": "state",
117
+ "fps": null,
118
+ "rep": null,
119
+ "rep_kwargs": null,
120
+ "compose": null
121
+ },
122
+ "observation_states_gripper_right": {
123
+ "shape": [
124
+ 1
125
+ ],
126
+ "dtype": "state",
127
+ "fps": null,
128
+ "rep": null,
129
+ "rep_kwargs": null,
130
+ "compose": null
131
+ }
132
+ }
133
+ }
134
+ },
135
+ "output_features": {
136
+ "action": {
137
+ "shape": [
138
+ 20
139
+ ],
140
+ "dtype": "action_pose_gripper",
141
+ "fps": null,
142
+ "rep": "rotation_6d",
143
+ "rep_kwargs": null,
144
+ "compose": {
145
+ "action_left": {
146
+ "shape": [
147
+ 10
148
+ ],
149
+ "dtype": "action_pose_gripper",
150
+ "fps": null,
151
+ "rep": "rotation_6d",
152
+ "rep_kwargs": null,
153
+ "compose": null
154
+ },
155
+ "action_right": {
156
+ "shape": [
157
+ 10
158
+ ],
159
+ "dtype": "action_pose_gripper",
160
+ "fps": null,
161
+ "rep": "rotation_6d",
162
+ "rep_kwargs": null,
163
+ "compose": null
164
+ }
165
+ }
166
+ }
167
+ },
168
+ "device": "cuda",
169
+ "use_amp": false,
170
+ "mixed_precision": null,
171
+ "data_transform_override": null,
172
+ "max_action_dim": 20,
173
+ "share_obs_encoder": true,
174
+ "obs_encoder_group": null,
175
+ "noise_scheduler": {
176
+ "type": "fm_beta",
177
+ "num_train_timesteps": 10,
178
+ "alpha": 1.5,
179
+ "beta": 1.0,
180
+ "s": 0.999,
181
+ "clip_sample": false,
182
+ "clip_sample_range": 1.0,
183
+ "flow_sig_min": 0.0,
184
+ "num_inference_steps": 10,
185
+ "prediction_type": "velocity",
186
+ "path_reversed": false
187
+ },
188
+ "vision_encoder": {
189
+ "type": "clip_hf_vision_encoder",
190
+ "model_name": "openai/clip-vit-base-patch16",
191
+ "tune_vision_encoder": true,
192
+ "extract_cls_token": true,
193
+ "output_projection_dim": null
194
+ },
195
+ "text_encoder": {
196
+ "type": "clip_hf_text_encoder",
197
+ "model_name": "openai/clip-vit-base-patch32",
198
+ "tune_text_encoder": false,
199
+ "tune_projection_layer": true,
200
+ "output_projection_dim": 768,
201
+ "extract_eos_token": true
202
+ },
203
+ "action_head": {
204
+ "type": "lbm_action_head",
205
+ "noise_scheduler": {
206
+ "type": "flow_matching_beta",
207
+ "num_inference_steps": 10,
208
+ "alpha": 1.5,
209
+ "beta": 1.0,
210
+ "s": 0.999,
211
+ "input_perturb": 0.0,
212
+ "sampling_path": "euler",
213
+ "clip_sample": false,
214
+ "clip_sample_range": null,
215
+ "path_reversed": false
216
+ },
217
+ "num_layers": 16,
218
+ "hidden_size": 768,
219
+ "num_attention_heads": 16,
220
+ "dropout": 0.1,
221
+ "final_dropout": true,
222
+ "max_timestep_buckets": 1000,
223
+ "conditioning_dim": 6980,
224
+ "max_state_dim": 34,
225
+ "max_action_dim": 20,
226
+ "max_chunk_size": 16
227
+ },
228
+ "proprioception_dim": 34,
229
+ "optimizer_lr": 0.0001,
230
+ "optimizer_betas": [
231
+ 0.9,
232
+ 0.95
233
+ ],
234
+ "optimizer_eps": 1e-08,
235
+ "optimizer_weight_decay": 1e-06,
236
+ "optimizer_vision_lr": 0.0001,
237
+ "optimizer_vision_weight_decay": 1e-06,
238
+ "scheduler_name": "cosine",
239
+ "scheduler_warmup_steps": 500
240
+ }
model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7c5053db697a968846179ce5da22108e2e25fdb0dfa16431233c1b95e547a6f
3
+ size 3399392491
train_config.json ADDED
@@ -0,0 +1,683 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "type": "lerobot",
4
+ "repo_id": "lerobot_dataset",
5
+ "root": "/scratch/shared/dataset_astra_lbm_20251119163923/astrabot/dataset_2025-11-04_11-14-49/",
6
+ "episodes": null,
7
+ "quality_filter": [
8
+ "smooth",
9
+ "corrected"
10
+ ],
11
+ "quality_sample_ratio": null,
12
+ "skill_id_filter": null,
13
+ "image_transforms": {
14
+ "enable": true,
15
+ "image_shape": [
16
+ 480,
17
+ 640
18
+ ],
19
+ "max_num_transforms": 3,
20
+ "random_order": false,
21
+ "tfs": {
22
+ "brightness": {
23
+ "weight": 1.0,
24
+ "always_apply": false,
25
+ "type": "ColorJitter",
26
+ "kwargs": {
27
+ "brightness": [
28
+ 0.7,
29
+ 1.3
30
+ ]
31
+ }
32
+ },
33
+ "contrast": {
34
+ "weight": 1.0,
35
+ "always_apply": false,
36
+ "type": "ColorJitter",
37
+ "kwargs": {
38
+ "contrast": [
39
+ 0.6,
40
+ 1.4
41
+ ]
42
+ }
43
+ },
44
+ "saturation": {
45
+ "weight": 1.0,
46
+ "always_apply": false,
47
+ "type": "ColorJitter",
48
+ "kwargs": {
49
+ "saturation": [
50
+ 0.5,
51
+ 1.5
52
+ ]
53
+ }
54
+ }
55
+ }
56
+ },
57
+ "video_backend": "decord",
58
+ "cache_frame_size": null,
59
+ "cache_video_workers": 16,
60
+ "normalize_images": true,
61
+ "load_additional_info": false,
62
+ "load_prompt": true,
63
+ "use_prompt_augmentation": null,
64
+ "task_skill_prompt_ratio": null,
65
+ "feature_key_remapping": null,
66
+ "image_stats_preset": null,
67
+ "local_files_only": true,
68
+ "download_videos": false,
69
+ "use_imagenet_stats": false
70
+ },
71
+ "robot_setup": {
72
+ "type": "astrabot",
73
+ "task_id": "",
74
+ "robot": {
75
+ "type": "astrabot",
76
+ "_obs_features": {
77
+ "observation_images_left_wrist": {
78
+ "shape": [
79
+ 480,
80
+ 640,
81
+ 3
82
+ ],
83
+ "dtype": "visual",
84
+ "fps": 30,
85
+ "rep": null,
86
+ "rep_kwargs": null,
87
+ "compose": null
88
+ },
89
+ "observation_images_right_wrist": {
90
+ "shape": [
91
+ 480,
92
+ 640,
93
+ 3
94
+ ],
95
+ "dtype": "visual",
96
+ "fps": 30,
97
+ "rep": null,
98
+ "rep_kwargs": null,
99
+ "compose": null
100
+ },
101
+ "observation_images_head": {
102
+ "shape": [
103
+ 480,
104
+ 640,
105
+ 3
106
+ ],
107
+ "dtype": "visual",
108
+ "fps": 30,
109
+ "rep": null,
110
+ "rep_kwargs": null,
111
+ "compose": null
112
+ },
113
+ "observation_images_chest": {
114
+ "shape": [
115
+ 480,
116
+ 640,
117
+ 3
118
+ ],
119
+ "dtype": "visual",
120
+ "fps": 30,
121
+ "rep": null,
122
+ "rep_kwargs": null,
123
+ "compose": null
124
+ },
125
+ "observation_states_joint_angle_left": {
126
+ "shape": [
127
+ 7
128
+ ],
129
+ "dtype": "state",
130
+ "fps": 50,
131
+ "rep": null,
132
+ "rep_kwargs": null,
133
+ "compose": null
134
+ },
135
+ "observation_states_joint_angle_right": {
136
+ "shape": [
137
+ 7
138
+ ],
139
+ "dtype": "state",
140
+ "fps": 50,
141
+ "rep": null,
142
+ "rep_kwargs": null,
143
+ "compose": null
144
+ },
145
+ "observation_states_ee_pose_left": {
146
+ "shape": [
147
+ 6
148
+ ],
149
+ "dtype": "state",
150
+ "fps": 50,
151
+ "rep": "euler",
152
+ "rep_kwargs": {
153
+ "seq": "ZYX",
154
+ "degrees": false,
155
+ "euler_in_rpy": true
156
+ },
157
+ "compose": null
158
+ },
159
+ "observation_states_ee_pose_right": {
160
+ "shape": [
161
+ 6
162
+ ],
163
+ "dtype": "state",
164
+ "fps": 50,
165
+ "rep": "euler",
166
+ "rep_kwargs": {
167
+ "seq": "ZYX",
168
+ "degrees": false,
169
+ "euler_in_rpy": true
170
+ },
171
+ "compose": null
172
+ },
173
+ "observation_states_gripper_left": {
174
+ "shape": [
175
+ 1
176
+ ],
177
+ "dtype": "state",
178
+ "fps": 50,
179
+ "rep": null,
180
+ "rep_kwargs": null,
181
+ "compose": null
182
+ },
183
+ "observation_states_gripper_right": {
184
+ "shape": [
185
+ 1
186
+ ],
187
+ "dtype": "state",
188
+ "fps": 50,
189
+ "rep": null,
190
+ "rep_kwargs": null,
191
+ "compose": null
192
+ }
193
+ },
194
+ "name": "astrabot",
195
+ "control_class": null,
196
+ "_actions": {
197
+ "action_left": {
198
+ "shape": [
199
+ 7
200
+ ],
201
+ "dtype": "action_pose_gripper",
202
+ "fps": 200,
203
+ "rep": "euler",
204
+ "rep_kwargs": {
205
+ "seq": "ZYX",
206
+ "degrees": false,
207
+ "euler_in_rpy": true
208
+ },
209
+ "compose": null
210
+ },
211
+ "action_right": {
212
+ "shape": [
213
+ 7
214
+ ],
215
+ "dtype": "action_pose_gripper",
216
+ "fps": 200,
217
+ "rep": "euler",
218
+ "rep_kwargs": {
219
+ "seq": "ZYX",
220
+ "degrees": false,
221
+ "euler_in_rpy": true
222
+ },
223
+ "compose": null
224
+ }
225
+ },
226
+ "action_relative_to_pose": {
227
+ "action_left": "observation_states_ee_pose_left",
228
+ "action_right": "observation_states_ee_pose_right"
229
+ }
230
+ },
231
+ "data_transform": {
232
+ "type": "astrabot",
233
+ "sensor_preprocessor": {
234
+ "type": "dict_shared",
235
+ "transform_fn": [
236
+ [
237
+ [
238
+ "observation_states_ee_pose_left",
239
+ "observation_states_ee_pose_right"
240
+ ],
241
+ [
242
+ {
243
+ "type": "pose_conversion",
244
+ "from_repr": "euler",
245
+ "to_repr": "rotation_6d",
246
+ "from_kwargs": {
247
+ "seq": "ZYX",
248
+ "degrees": false,
249
+ "euler_in_rpy": true
250
+ },
251
+ "to_kwargs": {},
252
+ "kwargs": {}
253
+ }
254
+ ]
255
+ ],
256
+ [
257
+ [
258
+ "action_left",
259
+ "action_right"
260
+ ],
261
+ [
262
+ {
263
+ "type": "action_conversion",
264
+ "from_repr": "euler",
265
+ "to_repr": "rotation_6d",
266
+ "from_kwargs": {
267
+ "seq": "ZYX",
268
+ "degrees": false,
269
+ "euler_in_rpy": true
270
+ },
271
+ "to_kwargs": {},
272
+ "kwargs": {}
273
+ }
274
+ ]
275
+ ]
276
+ ],
277
+ "tolerant_missing_keys": true
278
+ },
279
+ "dataset_postprocessor": {
280
+ "type": "dict_shared",
281
+ "transform_fn": [
282
+ [
283
+ [
284
+ "observation_images_left_wrist",
285
+ "observation_images_right_wrist",
286
+ "observation_images_head",
287
+ "observation_images_chest"
288
+ ],
289
+ [
290
+ {
291
+ "type": "image_random_crop",
292
+ "crop_ratio": 0.9,
293
+ "output_shape": [
294
+ 224,
295
+ 224
296
+ ],
297
+ "crop_square": true,
298
+ "random_crop": true
299
+ }
300
+ ]
301
+ ]
302
+ ],
303
+ "tolerant_missing_keys": true
304
+ },
305
+ "model_preprocessor": {
306
+ "type": "repack_dict",
307
+ "pattern": {
308
+ "observation_states": [
309
+ "observation_states_ee_pose_left",
310
+ "observation_states_joint_angle_left",
311
+ "observation_states_gripper_left",
312
+ "observation_states_ee_pose_right",
313
+ "observation_states_joint_angle_right",
314
+ "observation_states_gripper_right",
315
+ -1
316
+ ],
317
+ "action": [
318
+ "action_left",
319
+ "action_right",
320
+ -1
321
+ ]
322
+ },
323
+ "exclude": [],
324
+ "transform_fn": {},
325
+ "tolerant_missing_keys": true
326
+ },
327
+ "output_processor": {
328
+ "type": "eef_action_output",
329
+ "from_repr": "rotation_6d",
330
+ "from_kwargs": {},
331
+ "to_repr": "euler",
332
+ "to_kwargs": {
333
+ "seq": "ZYX",
334
+ "degrees": false,
335
+ "euler_in_rpy": true
336
+ },
337
+ "split_pattern": {
338
+ "action": {
339
+ "action_left": [
340
+ 0,
341
+ 10
342
+ ],
343
+ "action_right": [
344
+ 10,
345
+ 20
346
+ ]
347
+ }
348
+ }
349
+ }
350
+ },
351
+ "ignore_feature_keys": null,
352
+ "name": "",
353
+ "description": "",
354
+ "ext_sensor": null
355
+ },
356
+ "env": null,
357
+ "policy": {
358
+ "type": "lbm_policy",
359
+ "chunk_size": 16,
360
+ "n_action_steps": 8,
361
+ "n_obs_steps": 2,
362
+ "freq": 10.0,
363
+ "use_relative_action": true,
364
+ "use_relative_pose": false,
365
+ "normalization_mapping": {
366
+ "visual": "mean_std",
367
+ "state": "min_max",
368
+ "action_pose_gripper": "min_max"
369
+ },
370
+ "input_features": {
371
+ "observation_images_left_wrist": {
372
+ "shape": [
373
+ 3,
374
+ 224,
375
+ 224
376
+ ],
377
+ "dtype": "visual",
378
+ "fps": null,
379
+ "rep": null,
380
+ "rep_kwargs": null,
381
+ "compose": null
382
+ },
383
+ "observation_images_right_wrist": {
384
+ "shape": [
385
+ 3,
386
+ 224,
387
+ 224
388
+ ],
389
+ "dtype": "visual",
390
+ "fps": null,
391
+ "rep": null,
392
+ "rep_kwargs": null,
393
+ "compose": null
394
+ },
395
+ "observation_images_head": {
396
+ "shape": [
397
+ 3,
398
+ 224,
399
+ 224
400
+ ],
401
+ "dtype": "visual",
402
+ "fps": null,
403
+ "rep": null,
404
+ "rep_kwargs": null,
405
+ "compose": null
406
+ },
407
+ "observation_images_chest": {
408
+ "shape": [
409
+ 3,
410
+ 224,
411
+ 224
412
+ ],
413
+ "dtype": "visual",
414
+ "fps": null,
415
+ "rep": null,
416
+ "rep_kwargs": null,
417
+ "compose": null
418
+ },
419
+ "observation_states": {
420
+ "shape": [
421
+ 34
422
+ ],
423
+ "dtype": "state",
424
+ "fps": null,
425
+ "rep": null,
426
+ "rep_kwargs": null,
427
+ "compose": {
428
+ "observation_states_ee_pose_left": {
429
+ "shape": [
430
+ 9
431
+ ],
432
+ "dtype": "state",
433
+ "fps": null,
434
+ "rep": "rotation_6d",
435
+ "rep_kwargs": null,
436
+ "compose": null
437
+ },
438
+ "observation_states_joint_angle_left": {
439
+ "shape": [
440
+ 7
441
+ ],
442
+ "dtype": "state",
443
+ "fps": null,
444
+ "rep": null,
445
+ "rep_kwargs": null,
446
+ "compose": null
447
+ },
448
+ "observation_states_gripper_left": {
449
+ "shape": [
450
+ 1
451
+ ],
452
+ "dtype": "state",
453
+ "fps": null,
454
+ "rep": null,
455
+ "rep_kwargs": null,
456
+ "compose": null
457
+ },
458
+ "observation_states_ee_pose_right": {
459
+ "shape": [
460
+ 9
461
+ ],
462
+ "dtype": "state",
463
+ "fps": null,
464
+ "rep": "rotation_6d",
465
+ "rep_kwargs": null,
466
+ "compose": null
467
+ },
468
+ "observation_states_joint_angle_right": {
469
+ "shape": [
470
+ 7
471
+ ],
472
+ "dtype": "state",
473
+ "fps": null,
474
+ "rep": null,
475
+ "rep_kwargs": null,
476
+ "compose": null
477
+ },
478
+ "observation_states_gripper_right": {
479
+ "shape": [
480
+ 1
481
+ ],
482
+ "dtype": "state",
483
+ "fps": null,
484
+ "rep": null,
485
+ "rep_kwargs": null,
486
+ "compose": null
487
+ }
488
+ }
489
+ }
490
+ },
491
+ "output_features": {
492
+ "action": {
493
+ "shape": [
494
+ 20
495
+ ],
496
+ "dtype": "action_pose_gripper",
497
+ "fps": null,
498
+ "rep": "rotation_6d",
499
+ "rep_kwargs": null,
500
+ "compose": {
501
+ "action_left": {
502
+ "shape": [
503
+ 10
504
+ ],
505
+ "dtype": "action_pose_gripper",
506
+ "fps": null,
507
+ "rep": "rotation_6d",
508
+ "rep_kwargs": null,
509
+ "compose": null
510
+ },
511
+ "action_right": {
512
+ "shape": [
513
+ 10
514
+ ],
515
+ "dtype": "action_pose_gripper",
516
+ "fps": null,
517
+ "rep": "rotation_6d",
518
+ "rep_kwargs": null,
519
+ "compose": null
520
+ }
521
+ }
522
+ }
523
+ },
524
+ "device": "cuda",
525
+ "use_amp": false,
526
+ "mixed_precision": null,
527
+ "data_transform_override": null,
528
+ "max_action_dim": 20,
529
+ "share_obs_encoder": true,
530
+ "obs_encoder_group": null,
531
+ "noise_scheduler": {
532
+ "type": "fm_beta",
533
+ "num_train_timesteps": 10,
534
+ "alpha": 1.5,
535
+ "beta": 1.0,
536
+ "s": 0.999,
537
+ "clip_sample": false,
538
+ "clip_sample_range": 1.0,
539
+ "flow_sig_min": 0.0,
540
+ "num_inference_steps": 10,
541
+ "prediction_type": "velocity",
542
+ "path_reversed": false
543
+ },
544
+ "vision_encoder": {
545
+ "type": "clip_hf_vision_encoder",
546
+ "model_name": "openai/clip-vit-base-patch16",
547
+ "tune_vision_encoder": true,
548
+ "extract_cls_token": true,
549
+ "output_projection_dim": null
550
+ },
551
+ "text_encoder": {
552
+ "type": "clip_hf_text_encoder",
553
+ "model_name": "openai/clip-vit-base-patch32",
554
+ "tune_text_encoder": false,
555
+ "tune_projection_layer": true,
556
+ "output_projection_dim": 768,
557
+ "extract_eos_token": true
558
+ },
559
+ "action_head": {
560
+ "type": "lbm_action_head",
561
+ "noise_scheduler": {
562
+ "type": "flow_matching_beta",
563
+ "num_inference_steps": 10,
564
+ "alpha": 1.5,
565
+ "beta": 1.0,
566
+ "s": 0.999,
567
+ "input_perturb": 0.0,
568
+ "sampling_path": "euler",
569
+ "clip_sample": false,
570
+ "clip_sample_range": null,
571
+ "path_reversed": false
572
+ },
573
+ "num_layers": 16,
574
+ "hidden_size": 768,
575
+ "num_attention_heads": 16,
576
+ "dropout": 0.1,
577
+ "final_dropout": true,
578
+ "max_timestep_buckets": 1000,
579
+ "conditioning_dim": 6980,
580
+ "max_state_dim": 34,
581
+ "max_action_dim": 20,
582
+ "max_chunk_size": 16
583
+ },
584
+ "proprioception_dim": 34,
585
+ "optimizer_lr": 0.0001,
586
+ "optimizer_betas": [
587
+ 0.9,
588
+ 0.95
589
+ ],
590
+ "optimizer_eps": 1e-08,
591
+ "optimizer_weight_decay": 1e-06,
592
+ "optimizer_vision_lr": 0.0001,
593
+ "optimizer_vision_weight_decay": 1e-06,
594
+ "scheduler_name": "cosine",
595
+ "scheduler_warmup_steps": 500
596
+ },
597
+ "torch_compile": false,
598
+ "output_dir": "outputs/train/2025-11-19/21-44-38_astra_lbm",
599
+ "job_name": "astra_lbm",
600
+ "resume": false,
601
+ "reset_normalizer": false,
602
+ "seed": 42,
603
+ "dataloader": {
604
+ "batch_size": 50,
605
+ "global_batch_size": null,
606
+ "num_workers": 60,
607
+ "shuffle": true,
608
+ "pin_memory": true,
609
+ "persistent_workers": true,
610
+ "collate_fn": "prompts_as_list"
611
+ },
612
+ "long_horizon_training": false,
613
+ "fsdp": {
614
+ "enable": false,
615
+ "fsdp_version": 1,
616
+ "sharding_strategy": "FULL_SHARD",
617
+ "backward_prefetch": "BACKWARD_POST",
618
+ "cpu_offload": false,
619
+ "use_orig_params": true,
620
+ "sync_module_states": true,
621
+ "limit_all_gathers": true,
622
+ "auto_wrap_policy": "size_based_wrap",
623
+ "min_params": null,
624
+ "activation_checkpointing": false,
625
+ "mixed_precision_policy": "fp32"
626
+ },
627
+ "ema": {
628
+ "use_ema": true,
629
+ "decay": 0.99,
630
+ "device": null,
631
+ "update_every": 1,
632
+ "start_after": 0,
633
+ "dtype": null
634
+ },
635
+ "gradient_accumulate_every": 1,
636
+ "debug": false,
637
+ "steps": 100000,
638
+ "log_freq": 20,
639
+ "eval_freq": 2000,
640
+ "resample_freq": 10000,
641
+ "save_checkpoint": true,
642
+ "max_saved_checkpoints": 3,
643
+ "save_freq": 10000,
644
+ "use_policy_training_preset": true,
645
+ "optimizer": {
646
+ "type": "adamw",
647
+ "lr": 0.0001,
648
+ "weight_decay": 1e-06,
649
+ "grad_clip_norm": 1.0,
650
+ "betas": [
651
+ 0.9,
652
+ 0.95
653
+ ],
654
+ "eps": 1e-08
655
+ },
656
+ "scheduler": {
657
+ "type": "diffuser",
658
+ "num_warmup_steps": 500,
659
+ "name": "cosine"
660
+ },
661
+ "evaluation": {
662
+ "n_episodes": 50,
663
+ "batch_size": 50,
664
+ "use_async_envs": false
665
+ },
666
+ "wandb": {
667
+ "enable": true,
668
+ "disable_artifact": true,
669
+ "project": "astra_lbm",
670
+ "entity": null,
671
+ "notes": null,
672
+ "run_id": "nqokucby"
673
+ },
674
+ "mlflow": {
675
+ "enable": true,
676
+ "tracking_uri": "http://mlflow.tail44e147.ts.net",
677
+ "experiment_name": "astra_lbm",
678
+ "model_name": "lbm_policy",
679
+ "pretrained_path": null,
680
+ "background_logging": true,
681
+ "upload_training_state": false
682
+ }
683
+ }