Hugging Face's logo

Whalswp
/
experiment_v1

Safetensors
Model card Files
xet
 Community
Whalswp commited on
Commit
ce4f07f
·
verified ·
1 Parent(s): 4e651fa

Add files using upload-large-folder tool

Browse files
Files changed (17) hide show
  1. RKDv2.1/DA_MeanPool+LP/default/RKD v2.1 DA MF.yaml +62 -0
  2. RKDv2.1/DA_MeanPool+LP/default/phase2/experiment_cfg/metadata.json +431 -0
  3. RKDv2.1/DA_MeanPool+LP/default/resolved_config.yaml +63 -0
  4. RKDv2.1/KL_DTW_APHead_ComponentProbMix/default/RKD v2.1 KL-DTW APHead ComponentProbMix.yaml +72 -0
  5. RKDv2.1/KL_DTW_APHead_ComponentProbMix/default/phase2/experiment_cfg/metadata.json +431 -0
  6. RKDv2.1/KL_DTW_APHead_ComponentProbMix/default/resolved_config.yaml +73 -0
  7. RKDv2.1/KL_DTW_APHead_ComponentProbMix/launch_logs/train_gpu2-3_bs64_numgpu2_20260625_153201.log +0 -0
  8. RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix/default/RKD v2.1 KL-DTW APHead RowWiseControlGatedAdaptiveProbMix.yaml +78 -0
  9. RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix/default/phase2/experiment_cfg/metadata.json +431 -0
  10. RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix/default/resolved_config.yaml +79 -0
  11. RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix/launch_logs/train_gpu3_bs128_20260625_152846.log +0 -0
  12. clvla/1/default/clvla.yaml +47 -0
  13. clvla/1/default/phase2/experiment_cfg/metadata.json +431 -0
  14. clvla/1/default/resolved_config.yaml +48 -0
  15. clvla/2/default/clvla.yaml +47 -0
  16. clvla/2/default/phase2/experiment_cfg/metadata.json +431 -0
  17. clvla/2/default/resolved_config.yaml +48 -0
RKDv2.1/DA_MeanPool+LP/default/RKD v2.1 DA MF.yaml ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: RKD v2.1 DA, Student:MeanPool + LP
2
+ policy_type: groot_MGRKD
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/DA_MeanPool+LP
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model:
12
+ rkd_student_reconstruction_head_pooling: mean
13
+ phases:
14
+ - name: phase2_mgrkd_da_student_meanpool_only
15
+ max_steps: 30000
16
+ save_steps: 0
17
+ trainable:
18
+ preset: processing_line_only
19
+ tune_llm: false
20
+ tune_visual: false
21
+ tune_projector: false
22
+ tune_diffusion_model: false
23
+ losses:
24
+ rkd_enabled: true
25
+ rkd_fm_loss_weight: 0.0
26
+ rkd_loss_weight: 1.0
27
+ rkd_relation_mode: flatten
28
+ rkd_loss_type: distance_angle
29
+ rkd_teacher_source: action_encoder
30
+ rkd_action_encoder_projector_enabled: true
31
+ rkd_action_encoder_projector_dim: 512
32
+ rkd_action_encoder_projector_pooling: flatten
33
+ rkd_student_reconstruction_head_enabled: true
34
+ rkd_student_reconstruction_head_dim: 512
35
+ rkd_student_reconstruction_head_hidden_dim: 512
36
+ rkd_distance_loss_weight: 1.0
37
+ rkd_angle_loss_weight: 2.0
38
+ rkd_exclude_diagonal: true
39
+ - name: phase3_fm_mgrkd_da_student_meanpool_fixed_0p5
40
+ max_steps: 30000
41
+ save_steps: 0
42
+ trainable:
43
+ tune_llm: false
44
+ tune_visual: false
45
+ tune_projector: true
46
+ tune_diffusion_model: true
47
+ losses:
48
+ rkd_enabled: true
49
+ rkd_fm_loss_weight: 1.0
50
+ rkd_loss_weight: 0.5
51
+ rkd_relation_mode: flatten
52
+ rkd_loss_type: distance_angle
53
+ rkd_teacher_source: action_encoder
54
+ rkd_action_encoder_projector_enabled: true
55
+ rkd_action_encoder_projector_dim: 512
56
+ rkd_action_encoder_projector_pooling: flatten
57
+ rkd_student_reconstruction_head_enabled: true
58
+ rkd_student_reconstruction_head_dim: 512
59
+ rkd_student_reconstruction_head_hidden_dim: 512
60
+ rkd_distance_loss_weight: 1.0
61
+ rkd_angle_loss_weight: 2.0
62
+ rkd_exclude_diagonal: true
RKDv2.1/DA_MeanPool+LP/default/phase2/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.3139495849609375,
8
+ 0.4876587688922882,
9
+ 0.7196521759033203
10
+ ],
11
+ "min": [
12
+ -4.94293737411499,
13
+ -6.890198230743408,
14
+ 0.6996827125549316
15
+ ],
16
+ "mean": [
17
+ 2.749288365724937,
18
+ -2.0455216239909983,
19
+ 0.700532551020533
20
+ ],
21
+ "std": [
22
+ 1.6786295897046262,
23
+ 1.312897969434244,
24
+ 0.00133751758906258
25
+ ],
26
+ "q01": [
27
+ -1.4241907881148037,
28
+ -5.1716547935950885,
29
+ 0.6999670898768614
30
+ ],
31
+ "q99": [
32
+ 6.198111545831555,
33
+ -0.5873531334104489,
34
+ 0.7038833014242587
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.23084999587450247,
54
+ 0.6238431088581847
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6556404371644047,
60
+ 0.3572252105732042
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 1.7482354593273349e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9014528393745422,
78
+ 0.8003877401351929,
79
+ 0.9829942584037781
80
+ ],
81
+ "min": [
82
+ -0.37471598386764526,
83
+ -0.8472502827644348,
84
+ -0.25070279836654663
85
+ ],
86
+ "mean": [
87
+ 0.28667332601863704,
88
+ -0.038315473170876704,
89
+ 0.45974658906777444
90
+ ],
91
+ "std": [
92
+ 0.17067058828305154,
93
+ 0.23569952037784195,
94
+ 0.21950702354181487
95
+ ],
96
+ "q01": [
97
+ 0.015172948963784228,
98
+ -0.44450502063220615,
99
+ 0.23314444103329338
100
+ ],
101
+ "q99": [
102
+ 0.5609069689572412,
103
+ 0.38454179128009547,
104
+ 0.7028102736473852
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9984455108642578,
111
+ 0.9434727430343628,
112
+ 0.9062229990959167
113
+ ],
114
+ "min": [
115
+ -0.9999930262565613,
116
+ -0.9988337755203247,
117
+ -0.9618149995803833,
118
+ 2.1872274658107926e-07
119
+ ],
120
+ "mean": [
121
+ -0.25672297401336,
122
+ 0.024425335806687216,
123
+ -0.08740977164483847,
124
+ 0.16608739644018397
125
+ ],
126
+ "std": [
127
+ 0.7932585208392383,
128
+ 0.3102260195580416,
129
+ 0.3772472576315148,
130
+ 0.17451959300158773
131
+ ],
132
+ "q01": [
133
+ -0.99379006987882,
134
+ -0.5478102050038828,
135
+ -0.6101777952971636,
136
+ 0.002274085014134748
137
+ ],
138
+ "q99": [
139
+ 0.8804856038896288,
140
+ 0.5910611092873037,
141
+ 0.5216058042993562,
142
+ 0.5231965974012255
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.055869169533252716,
148
+ 0.010916369967162609
149
+ ],
150
+ "min": [
151
+ -0.011436971835792065,
152
+ -0.05664053186774254
153
+ ],
154
+ "mean": [
155
+ 0.031651551516052555,
156
+ -0.03161481387920421
157
+ ],
158
+ "std": [
159
+ 0.013119769222217526,
160
+ 0.01306078225192402
161
+ ],
162
+ "q01": [
163
+ 0.0060999246446777336,
164
+ -0.04062601653964198
165
+ ],
166
+ "q99": [
167
+ 0.04054891621517283,
168
+ -0.0059163567113456345
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.008144896753718373,
188
+ -0.00018893135146078263,
189
+ -0.0008739062727221845,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11030045424364411,
194
+ 0.10148082570313594,
195
+ 0.08975698373467506,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.07287373067581518,
200
+ -0.09446994199569948,
201
+ -0.07702818259249572,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.04485485414407898,
206
+ 0.09626941605965177,
207
+ 0.07610665906122742,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9221974313425939
220
+ ],
221
+ "std": [
222
+ 0.38670194342612674
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.384693946883099
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ 0.01976129923017491,
244
+ -0.020870631344490034,
245
+ -0.05993476517349181
246
+ ],
247
+ "std": [
248
+ 0.4347024990804053,
249
+ 0.4221662976569997,
250
+ 0.3845506843044066
251
+ ],
252
+ "q01": [
253
+ -0.8835792939767807,
254
+ -0.9280919251713778,
255
+ -0.783361071470956
256
+ ],
257
+ "q99": [
258
+ 0.7622084438449307,
259
+ 0.8625125252609077,
260
+ 0.7470974753250706
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008089673068544335,
276
+ -0.026498254627927348,
277
+ 0.0029083551743059005
278
+ ],
279
+ "std": [
280
+ 0.11216734503733773,
281
+ 0.13206002613798629,
282
+ 0.12615870412692864
283
+ ],
284
+ "q01": [
285
+ -0.2814627972846672,
286
+ -0.4342386516115439,
287
+ -0.34489755628340574
288
+ ],
289
+ "q99": [
290
+ 0.31879353266939386,
291
+ 0.28411797683220563,
292
+ 0.3597718671669894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3824228874769045
304
+ ],
305
+ "std": [
306
+ 0.9240015096469786
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.5597272305813592
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
RKDv2.1/DA_MeanPool+LP/default/resolved_config.yaml ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: RKD v2.1 DA, Student:MeanPool + LP
2
+ policy_type: groot_MGRKD
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/DA_MeanPool+LP
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model:
12
+ rkd_student_reconstruction_head_pooling: mean
13
+ phases:
14
+ - name: phase2_mgrkd_da_student_meanpool_only
15
+ max_steps: 30000
16
+ save_steps: 0
17
+ trainable:
18
+ preset: processing_line_only
19
+ tune_llm: false
20
+ tune_visual: false
21
+ tune_projector: false
22
+ tune_diffusion_model: false
23
+ losses:
24
+ rkd_enabled: true
25
+ rkd_fm_loss_weight: 0.0
26
+ rkd_loss_weight: 1.0
27
+ rkd_relation_mode: flatten
28
+ rkd_loss_type: distance_angle
29
+ rkd_teacher_source: action_encoder
30
+ rkd_action_encoder_projector_enabled: true
31
+ rkd_action_encoder_projector_dim: 512
32
+ rkd_action_encoder_projector_pooling: flatten
33
+ rkd_student_reconstruction_head_enabled: true
34
+ rkd_student_reconstruction_head_dim: 512
35
+ rkd_student_reconstruction_head_hidden_dim: 512
36
+ rkd_distance_loss_weight: 1.0
37
+ rkd_angle_loss_weight: 2.0
38
+ rkd_exclude_diagonal: true
39
+ - name: phase3_fm_mgrkd_da_student_meanpool_fixed_0p5
40
+ max_steps: 30000
41
+ save_steps: 0
42
+ trainable:
43
+ tune_llm: false
44
+ tune_visual: false
45
+ tune_projector: true
46
+ tune_diffusion_model: true
47
+ losses:
48
+ rkd_enabled: true
49
+ rkd_fm_loss_weight: 1.0
50
+ rkd_loss_weight: 0.5
51
+ rkd_relation_mode: flatten
52
+ rkd_loss_type: distance_angle
53
+ rkd_teacher_source: action_encoder
54
+ rkd_action_encoder_projector_enabled: true
55
+ rkd_action_encoder_projector_dim: 512
56
+ rkd_action_encoder_projector_pooling: flatten
57
+ rkd_student_reconstruction_head_enabled: true
58
+ rkd_student_reconstruction_head_dim: 512
59
+ rkd_student_reconstruction_head_hidden_dim: 512
60
+ rkd_distance_loss_weight: 1.0
61
+ rkd_angle_loss_weight: 2.0
62
+ rkd_exclude_diagonal: true
63
+ resolved_sweep: {}
RKDv2.1/KL_DTW_APHead_ComponentProbMix/default/RKD v2.1 KL-DTW APHead ComponentProbMix.yaml ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: RKD v2.1 KL-DTW ComponentProbMix, Student:APHead
2
+ policy_type: groot_MGRKD
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/KL_DTW_APHead_ComponentProbMix
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 2
9
+ batch_size: 64
10
+ seed: 42
11
+ model:
12
+ rkd_student_reconstruction_head_pooling: attention
13
+ phases:
14
+ - name: phase2_mgrkd_kl_dtw_aphead_component_prob_mix_only
15
+ max_steps: 30000
16
+ save_steps: 0
17
+ trainable:
18
+ preset: processing_line_only
19
+ tune_llm: false
20
+ tune_visual: false
21
+ tune_projector: false
22
+ tune_diffusion_model: false
23
+ losses:
24
+ rkd_enabled: true
25
+ rkd_fm_loss_weight: 0.0
26
+ rkd_loss_weight: 1.0
27
+ rkd_relation_mode: flatten
28
+ rkd_loss_type: kl
29
+ rkd_teacher_source: dtw_action
30
+ rkd_action_temp: 0.1
31
+ rkd_vlm_temp: 0.1
32
+ rkd_action_encoder_projector_enabled: false
33
+ rkd_raw_action_projector_enabled: false
34
+ rkd_student_reconstruction_head_enabled: true
35
+ rkd_student_reconstruction_head_dim: 512
36
+ rkd_student_reconstruction_head_hidden_dim: 512
37
+ rkd_dtw_mix_mode: component_prob_mix
38
+ rkd_dtw_pos_weight: 1.0
39
+ rkd_dtw_rot_weight: 1.0
40
+ rkd_dtw_gripper_weight: 0.5
41
+ rkd_dtw_normalize_components: true
42
+ rkd_dtw_timing_log_interval: 1000
43
+ rkd_exclude_diagonal: true
44
+ - name: phase3_fm_mgrkd_kl_dtw_aphead_component_prob_mix_fixed_0p5
45
+ max_steps: 30000
46
+ save_steps: 0
47
+ trainable:
48
+ tune_llm: false
49
+ tune_visual: false
50
+ tune_projector: true
51
+ tune_diffusion_model: true
52
+ losses:
53
+ rkd_enabled: true
54
+ rkd_fm_loss_weight: 1.0
55
+ rkd_loss_weight: 0.5
56
+ rkd_relation_mode: flatten
57
+ rkd_loss_type: kl
58
+ rkd_teacher_source: dtw_action
59
+ rkd_action_temp: 0.1
60
+ rkd_vlm_temp: 0.1
61
+ rkd_action_encoder_projector_enabled: false
62
+ rkd_raw_action_projector_enabled: false
63
+ rkd_student_reconstruction_head_enabled: true
64
+ rkd_student_reconstruction_head_dim: 512
65
+ rkd_student_reconstruction_head_hidden_dim: 512
66
+ rkd_dtw_mix_mode: component_prob_mix
67
+ rkd_dtw_pos_weight: 1.0
68
+ rkd_dtw_rot_weight: 1.0
69
+ rkd_dtw_gripper_weight: 0.5
70
+ rkd_dtw_normalize_components: true
71
+ rkd_dtw_timing_log_interval: 1000
72
+ rkd_exclude_diagonal: true
RKDv2.1/KL_DTW_APHead_ComponentProbMix/default/phase2/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.3139495849609375,
8
+ 0.4876587688922882,
9
+ 0.7196521759033203
10
+ ],
11
+ "min": [
12
+ -4.94293737411499,
13
+ -6.890198230743408,
14
+ 0.6996827125549316
15
+ ],
16
+ "mean": [
17
+ 2.749288365724937,
18
+ -2.0455216239909983,
19
+ 0.700532551020533
20
+ ],
21
+ "std": [
22
+ 1.6786295897046262,
23
+ 1.312897969434244,
24
+ 0.00133751758906258
25
+ ],
26
+ "q01": [
27
+ -1.4241907881148037,
28
+ -5.1716547935950885,
29
+ 0.6999670898768614
30
+ ],
31
+ "q99": [
32
+ 6.198111545831555,
33
+ -0.5873531334104489,
34
+ 0.7038833014242587
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.23084999587450247,
54
+ 0.6238431088581847
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6556404371644047,
60
+ 0.3572252105732042
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 1.7482354593273349e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9014528393745422,
78
+ 0.8003877401351929,
79
+ 0.9829942584037781
80
+ ],
81
+ "min": [
82
+ -0.37471598386764526,
83
+ -0.8472502827644348,
84
+ -0.25070279836654663
85
+ ],
86
+ "mean": [
87
+ 0.28667332601863704,
88
+ -0.038315473170876704,
89
+ 0.45974658906777444
90
+ ],
91
+ "std": [
92
+ 0.17067058828305154,
93
+ 0.23569952037784195,
94
+ 0.21950702354181487
95
+ ],
96
+ "q01": [
97
+ 0.015172948963784228,
98
+ -0.44450502063220615,
99
+ 0.23314444103329338
100
+ ],
101
+ "q99": [
102
+ 0.5609069689572412,
103
+ 0.38454179128009547,
104
+ 0.7028102736473852
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9984455108642578,
111
+ 0.9434727430343628,
112
+ 0.9062229990959167
113
+ ],
114
+ "min": [
115
+ -0.9999930262565613,
116
+ -0.9988337755203247,
117
+ -0.9618149995803833,
118
+ 2.1872274658107926e-07
119
+ ],
120
+ "mean": [
121
+ -0.25672297401336,
122
+ 0.024425335806687216,
123
+ -0.08740977164483847,
124
+ 0.16608739644018397
125
+ ],
126
+ "std": [
127
+ 0.7932585208392383,
128
+ 0.3102260195580416,
129
+ 0.3772472576315148,
130
+ 0.17451959300158773
131
+ ],
132
+ "q01": [
133
+ -0.99379006987882,
134
+ -0.5478102050038828,
135
+ -0.6101777952971636,
136
+ 0.002274085014134748
137
+ ],
138
+ "q99": [
139
+ 0.8804856038896288,
140
+ 0.5910611092873037,
141
+ 0.5216058042993562,
142
+ 0.5231965974012255
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.055869169533252716,
148
+ 0.010916369967162609
149
+ ],
150
+ "min": [
151
+ -0.011436971835792065,
152
+ -0.05664053186774254
153
+ ],
154
+ "mean": [
155
+ 0.031651551516052555,
156
+ -0.03161481387920421
157
+ ],
158
+ "std": [
159
+ 0.013119769222217526,
160
+ 0.01306078225192402
161
+ ],
162
+ "q01": [
163
+ 0.0060999246446777336,
164
+ -0.04062601653964198
165
+ ],
166
+ "q99": [
167
+ 0.04054891621517283,
168
+ -0.0059163567113456345
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.008144896753718373,
188
+ -0.00018893135146078263,
189
+ -0.0008739062727221845,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11030045424364411,
194
+ 0.10148082570313594,
195
+ 0.08975698373467506,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.07287373067581518,
200
+ -0.09446994199569948,
201
+ -0.07702818259249572,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.04485485414407898,
206
+ 0.09626941605965177,
207
+ 0.07610665906122742,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9221974313425939
220
+ ],
221
+ "std": [
222
+ 0.38670194342612674
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.384693946883099
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ 0.01976129923017491,
244
+ -0.020870631344490034,
245
+ -0.05993476517349181
246
+ ],
247
+ "std": [
248
+ 0.4347024990804053,
249
+ 0.4221662976569997,
250
+ 0.3845506843044066
251
+ ],
252
+ "q01": [
253
+ -0.8835792939767807,
254
+ -0.9280919251713778,
255
+ -0.783361071470956
256
+ ],
257
+ "q99": [
258
+ 0.7622084438449307,
259
+ 0.8625125252609077,
260
+ 0.7470974753250706
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008089673068544335,
276
+ -0.026498254627927348,
277
+ 0.0029083551743059005
278
+ ],
279
+ "std": [
280
+ 0.11216734503733773,
281
+ 0.13206002613798629,
282
+ 0.12615870412692864
283
+ ],
284
+ "q01": [
285
+ -0.2814627972846672,
286
+ -0.4342386516115439,
287
+ -0.34489755628340574
288
+ ],
289
+ "q99": [
290
+ 0.31879353266939386,
291
+ 0.28411797683220563,
292
+ 0.3597718671669894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3824228874769045
304
+ ],
305
+ "std": [
306
+ 0.9240015096469786
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.5597272305813592
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
RKDv2.1/KL_DTW_APHead_ComponentProbMix/default/resolved_config.yaml ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: RKD v2.1 KL-DTW ComponentProbMix, Student:APHead
2
+ policy_type: groot_MGRKD
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/KL_DTW_APHead_ComponentProbMix
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 2
9
+ batch_size: 64
10
+ seed: 42
11
+ model:
12
+ rkd_student_reconstruction_head_pooling: attention
13
+ phases:
14
+ - name: phase2_mgrkd_kl_dtw_aphead_component_prob_mix_only
15
+ max_steps: 30000
16
+ save_steps: 0
17
+ trainable:
18
+ preset: processing_line_only
19
+ tune_llm: false
20
+ tune_visual: false
21
+ tune_projector: false
22
+ tune_diffusion_model: false
23
+ losses:
24
+ rkd_enabled: true
25
+ rkd_fm_loss_weight: 0.0
26
+ rkd_loss_weight: 1.0
27
+ rkd_relation_mode: flatten
28
+ rkd_loss_type: kl
29
+ rkd_teacher_source: dtw_action
30
+ rkd_action_temp: 0.1
31
+ rkd_vlm_temp: 0.1
32
+ rkd_action_encoder_projector_enabled: false
33
+ rkd_raw_action_projector_enabled: false
34
+ rkd_student_reconstruction_head_enabled: true
35
+ rkd_student_reconstruction_head_dim: 512
36
+ rkd_student_reconstruction_head_hidden_dim: 512
37
+ rkd_dtw_mix_mode: component_prob_mix
38
+ rkd_dtw_pos_weight: 1.0
39
+ rkd_dtw_rot_weight: 1.0
40
+ rkd_dtw_gripper_weight: 0.5
41
+ rkd_dtw_normalize_components: true
42
+ rkd_dtw_timing_log_interval: 1000
43
+ rkd_exclude_diagonal: true
44
+ - name: phase3_fm_mgrkd_kl_dtw_aphead_component_prob_mix_fixed_0p5
45
+ max_steps: 30000
46
+ save_steps: 0
47
+ trainable:
48
+ tune_llm: false
49
+ tune_visual: false
50
+ tune_projector: true
51
+ tune_diffusion_model: true
52
+ losses:
53
+ rkd_enabled: true
54
+ rkd_fm_loss_weight: 1.0
55
+ rkd_loss_weight: 0.5
56
+ rkd_relation_mode: flatten
57
+ rkd_loss_type: kl
58
+ rkd_teacher_source: dtw_action
59
+ rkd_action_temp: 0.1
60
+ rkd_vlm_temp: 0.1
61
+ rkd_action_encoder_projector_enabled: false
62
+ rkd_raw_action_projector_enabled: false
63
+ rkd_student_reconstruction_head_enabled: true
64
+ rkd_student_reconstruction_head_dim: 512
65
+ rkd_student_reconstruction_head_hidden_dim: 512
66
+ rkd_dtw_mix_mode: component_prob_mix
67
+ rkd_dtw_pos_weight: 1.0
68
+ rkd_dtw_rot_weight: 1.0
69
+ rkd_dtw_gripper_weight: 0.5
70
+ rkd_dtw_normalize_components: true
71
+ rkd_dtw_timing_log_interval: 1000
72
+ rkd_exclude_diagonal: true
73
+ resolved_sweep: {}
RKDv2.1/KL_DTW_APHead_ComponentProbMix/launch_logs/train_gpu2-3_bs64_numgpu2_20260625_153201.log ADDED
The diff for this file is too large to render. See raw diff
 
RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix/default/RKD v2.1 KL-DTW APHead RowWiseControlGatedAdaptiveProbMix.yaml ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: RKD v2.1 KL-DTW RowWiseControlGatedAdaptiveProbMix, Student:APHead
2
+ policy_type: groot_MGRKD
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model:
12
+ rkd_student_reconstruction_head_pooling: attention
13
+ phases:
14
+ - name: phase2_mgrkd_kl_dtw_aphead_rowwise_control_gated_only
15
+ max_steps: 30000
16
+ save_steps: 0
17
+ trainable:
18
+ preset: processing_line_only
19
+ tune_llm: false
20
+ tune_visual: false
21
+ tune_projector: false
22
+ tune_diffusion_model: false
23
+ losses:
24
+ rkd_enabled: true
25
+ rkd_fm_loss_weight: 0.0
26
+ rkd_loss_weight: 1.0
27
+ rkd_relation_mode: flatten
28
+ rkd_loss_type: kl
29
+ rkd_teacher_source: dtw_action
30
+ rkd_action_temp: 0.1
31
+ rkd_vlm_temp: 0.1
32
+ rkd_action_encoder_projector_enabled: false
33
+ rkd_raw_action_projector_enabled: false
34
+ rkd_student_reconstruction_head_enabled: true
35
+ rkd_student_reconstruction_head_dim: 512
36
+ rkd_student_reconstruction_head_hidden_dim: 512
37
+ rkd_dtw_mix_mode: rowwise_control_gated_adaptive_prob_mix
38
+ rkd_dtw_pos_weight: 1.0
39
+ rkd_dtw_rot_weight: 1.0
40
+ rkd_dtw_gripper_weight: 0.5
41
+ rkd_dtw_base_weight: 1.0
42
+ rkd_dtw_normalize_components: true
43
+ rkd_dtw_adaptive_mix_lambda: 1.0
44
+ rkd_dtw_adaptive_floor: 0.0
45
+ rkd_dtw_timing_log_interval: 1000
46
+ rkd_exclude_diagonal: true
47
+ - name: phase3_fm_mgrkd_kl_dtw_aphead_rowwise_control_gated_fixed_0p5
48
+ max_steps: 30000
49
+ save_steps: 0
50
+ trainable:
51
+ tune_llm: false
52
+ tune_visual: false
53
+ tune_projector: true
54
+ tune_diffusion_model: true
55
+ losses:
56
+ rkd_enabled: true
57
+ rkd_fm_loss_weight: 1.0
58
+ rkd_loss_weight: 0.5
59
+ rkd_relation_mode: flatten
60
+ rkd_loss_type: kl
61
+ rkd_teacher_source: dtw_action
62
+ rkd_action_temp: 0.1
63
+ rkd_vlm_temp: 0.1
64
+ rkd_action_encoder_projector_enabled: false
65
+ rkd_raw_action_projector_enabled: false
66
+ rkd_student_reconstruction_head_enabled: true
67
+ rkd_student_reconstruction_head_dim: 512
68
+ rkd_student_reconstruction_head_hidden_dim: 512
69
+ rkd_dtw_mix_mode: rowwise_control_gated_adaptive_prob_mix
70
+ rkd_dtw_pos_weight: 1.0
71
+ rkd_dtw_rot_weight: 1.0
72
+ rkd_dtw_gripper_weight: 0.5
73
+ rkd_dtw_base_weight: 1.0
74
+ rkd_dtw_normalize_components: true
75
+ rkd_dtw_adaptive_mix_lambda: 1.0
76
+ rkd_dtw_adaptive_floor: 0.0
77
+ rkd_dtw_timing_log_interval: 1000
78
+ rkd_exclude_diagonal: true
RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix/default/phase2/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.3139495849609375,
8
+ 0.4876587688922882,
9
+ 0.7196521759033203
10
+ ],
11
+ "min": [
12
+ -4.94293737411499,
13
+ -6.890198230743408,
14
+ 0.6996827125549316
15
+ ],
16
+ "mean": [
17
+ 2.749288365724937,
18
+ -2.0455216239909983,
19
+ 0.700532551020533
20
+ ],
21
+ "std": [
22
+ 1.6786295897046262,
23
+ 1.312897969434244,
24
+ 0.00133751758906258
25
+ ],
26
+ "q01": [
27
+ -1.4241907881148037,
28
+ -5.1716547935950885,
29
+ 0.6999670898768614
30
+ ],
31
+ "q99": [
32
+ 6.198111545831555,
33
+ -0.5873531334104489,
34
+ 0.7038833014242587
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.23084999587450247,
54
+ 0.6238431088581847
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6556404371644047,
60
+ 0.3572252105732042
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 1.7482354593273349e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9014528393745422,
78
+ 0.8003877401351929,
79
+ 0.9829942584037781
80
+ ],
81
+ "min": [
82
+ -0.37471598386764526,
83
+ -0.8472502827644348,
84
+ -0.25070279836654663
85
+ ],
86
+ "mean": [
87
+ 0.28667332601863704,
88
+ -0.038315473170876704,
89
+ 0.45974658906777444
90
+ ],
91
+ "std": [
92
+ 0.17067058828305154,
93
+ 0.23569952037784195,
94
+ 0.21950702354181487
95
+ ],
96
+ "q01": [
97
+ 0.015172948963784228,
98
+ -0.44450502063220615,
99
+ 0.23314444103329338
100
+ ],
101
+ "q99": [
102
+ 0.5609069689572412,
103
+ 0.38454179128009547,
104
+ 0.7028102736473852
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9984455108642578,
111
+ 0.9434727430343628,
112
+ 0.9062229990959167
113
+ ],
114
+ "min": [
115
+ -0.9999930262565613,
116
+ -0.9988337755203247,
117
+ -0.9618149995803833,
118
+ 2.1872274658107926e-07
119
+ ],
120
+ "mean": [
121
+ -0.25672297401336,
122
+ 0.024425335806687216,
123
+ -0.08740977164483847,
124
+ 0.16608739644018397
125
+ ],
126
+ "std": [
127
+ 0.7932585208392383,
128
+ 0.3102260195580416,
129
+ 0.3772472576315148,
130
+ 0.17451959300158773
131
+ ],
132
+ "q01": [
133
+ -0.99379006987882,
134
+ -0.5478102050038828,
135
+ -0.6101777952971636,
136
+ 0.002274085014134748
137
+ ],
138
+ "q99": [
139
+ 0.8804856038896288,
140
+ 0.5910611092873037,
141
+ 0.5216058042993562,
142
+ 0.5231965974012255
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.055869169533252716,
148
+ 0.010916369967162609
149
+ ],
150
+ "min": [
151
+ -0.011436971835792065,
152
+ -0.05664053186774254
153
+ ],
154
+ "mean": [
155
+ 0.031651551516052555,
156
+ -0.03161481387920421
157
+ ],
158
+ "std": [
159
+ 0.013119769222217526,
160
+ 0.01306078225192402
161
+ ],
162
+ "q01": [
163
+ 0.0060999246446777336,
164
+ -0.04062601653964198
165
+ ],
166
+ "q99": [
167
+ 0.04054891621517283,
168
+ -0.0059163567113456345
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.008144896753718373,
188
+ -0.00018893135146078263,
189
+ -0.0008739062727221845,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11030045424364411,
194
+ 0.10148082570313594,
195
+ 0.08975698373467506,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.07287373067581518,
200
+ -0.09446994199569948,
201
+ -0.07702818259249572,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.04485485414407898,
206
+ 0.09626941605965177,
207
+ 0.07610665906122742,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9221974313425939
220
+ ],
221
+ "std": [
222
+ 0.38670194342612674
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.384693946883099
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ 0.01976129923017491,
244
+ -0.020870631344490034,
245
+ -0.05993476517349181
246
+ ],
247
+ "std": [
248
+ 0.4347024990804053,
249
+ 0.4221662976569997,
250
+ 0.3845506843044066
251
+ ],
252
+ "q01": [
253
+ -0.8835792939767807,
254
+ -0.9280919251713778,
255
+ -0.783361071470956
256
+ ],
257
+ "q99": [
258
+ 0.7622084438449307,
259
+ 0.8625125252609077,
260
+ 0.7470974753250706
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008089673068544335,
276
+ -0.026498254627927348,
277
+ 0.0029083551743059005
278
+ ],
279
+ "std": [
280
+ 0.11216734503733773,
281
+ 0.13206002613798629,
282
+ 0.12615870412692864
283
+ ],
284
+ "q01": [
285
+ -0.2814627972846672,
286
+ -0.4342386516115439,
287
+ -0.34489755628340574
288
+ ],
289
+ "q99": [
290
+ 0.31879353266939386,
291
+ 0.28411797683220563,
292
+ 0.3597718671669894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3824228874769045
304
+ ],
305
+ "std": [
306
+ 0.9240015096469786
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.5597272305813592
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix/default/resolved_config.yaml ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: RKD v2.1 KL-DTW RowWiseControlGatedAdaptiveProbMix, Student:APHead
2
+ policy_type: groot_MGRKD
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model:
12
+ rkd_student_reconstruction_head_pooling: attention
13
+ phases:
14
+ - name: phase2_mgrkd_kl_dtw_aphead_rowwise_control_gated_only
15
+ max_steps: 30000
16
+ save_steps: 0
17
+ trainable:
18
+ preset: processing_line_only
19
+ tune_llm: false
20
+ tune_visual: false
21
+ tune_projector: false
22
+ tune_diffusion_model: false
23
+ losses:
24
+ rkd_enabled: true
25
+ rkd_fm_loss_weight: 0.0
26
+ rkd_loss_weight: 1.0
27
+ rkd_relation_mode: flatten
28
+ rkd_loss_type: kl
29
+ rkd_teacher_source: dtw_action
30
+ rkd_action_temp: 0.1
31
+ rkd_vlm_temp: 0.1
32
+ rkd_action_encoder_projector_enabled: false
33
+ rkd_raw_action_projector_enabled: false
34
+ rkd_student_reconstruction_head_enabled: true
35
+ rkd_student_reconstruction_head_dim: 512
36
+ rkd_student_reconstruction_head_hidden_dim: 512
37
+ rkd_dtw_mix_mode: rowwise_control_gated_adaptive_prob_mix
38
+ rkd_dtw_pos_weight: 1.0
39
+ rkd_dtw_rot_weight: 1.0
40
+ rkd_dtw_gripper_weight: 0.5
41
+ rkd_dtw_base_weight: 1.0
42
+ rkd_dtw_normalize_components: true
43
+ rkd_dtw_adaptive_mix_lambda: 1.0
44
+ rkd_dtw_adaptive_floor: 0.0
45
+ rkd_dtw_timing_log_interval: 1000
46
+ rkd_exclude_diagonal: true
47
+ - name: phase3_fm_mgrkd_kl_dtw_aphead_rowwise_control_gated_fixed_0p5
48
+ max_steps: 30000
49
+ save_steps: 0
50
+ trainable:
51
+ tune_llm: false
52
+ tune_visual: false
53
+ tune_projector: true
54
+ tune_diffusion_model: true
55
+ losses:
56
+ rkd_enabled: true
57
+ rkd_fm_loss_weight: 1.0
58
+ rkd_loss_weight: 0.5
59
+ rkd_relation_mode: flatten
60
+ rkd_loss_type: kl
61
+ rkd_teacher_source: dtw_action
62
+ rkd_action_temp: 0.1
63
+ rkd_vlm_temp: 0.1
64
+ rkd_action_encoder_projector_enabled: false
65
+ rkd_raw_action_projector_enabled: false
66
+ rkd_student_reconstruction_head_enabled: true
67
+ rkd_student_reconstruction_head_dim: 512
68
+ rkd_student_reconstruction_head_hidden_dim: 512
69
+ rkd_dtw_mix_mode: rowwise_control_gated_adaptive_prob_mix
70
+ rkd_dtw_pos_weight: 1.0
71
+ rkd_dtw_rot_weight: 1.0
72
+ rkd_dtw_gripper_weight: 0.5
73
+ rkd_dtw_base_weight: 1.0
74
+ rkd_dtw_normalize_components: true
75
+ rkd_dtw_adaptive_mix_lambda: 1.0
76
+ rkd_dtw_adaptive_floor: 0.0
77
+ rkd_dtw_timing_log_interval: 1000
78
+ rkd_exclude_diagonal: true
79
+ resolved_sweep: {}
RKDv2.1/KL_DTW_APHead_RowWiseControlGatedAdaptiveProbMix/launch_logs/train_gpu3_bs128_20260625_152846.log ADDED
The diff for this file is too large to render. See raw diff
 
clvla/1/default/clvla.yaml ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: clvla
2
+ policy_type: groot_rkd_v2
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: /home/ext_minje/groot_robocasa/robocasa_v2/clvla/1
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model: null
12
+ phases:
13
+ - name: phase2_rkd_da_only
14
+ max_steps: 600000
15
+ save_steps: 0
16
+ trainable:
17
+ preset: processing_line_only
18
+ tune_llm: false
19
+ tune_visual: false
20
+ tune_projector: false
21
+ tune_diffusion_model: false
22
+ losses:
23
+ rkd_enabled: true
24
+ rkd_fm_loss_weight: 0.0
25
+ rkd_loss_weight: 1.0
26
+ rkd_relation_mode: token_pair_mean
27
+ rkd_loss_type: distance_angle
28
+ rkd_distance_loss_weight: 1.0
29
+ rkd_angle_loss_weight: 2.0
30
+ rkd_exclude_diagonal: true
31
+ - name: phase3_fm_rkd_da_fixed_0p5
32
+ max_steps: 200000
33
+ save_steps: 0
34
+ trainable:
35
+ tune_llm: false
36
+ tune_visual: false
37
+ tune_projector: true
38
+ tune_diffusion_model: true
39
+ losses:
40
+ rkd_enabled: true
41
+ rkd_fm_loss_weight: 1.0
42
+ rkd_loss_weight: 0.5
43
+ rkd_relation_mode: token_pair_mean
44
+ rkd_loss_type: distance_angle
45
+ rkd_distance_loss_weight: 1.0
46
+ rkd_angle_loss_weight: 2.0
47
+ rkd_exclude_diagonal: true
clvla/1/default/phase2/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.3139495849609375,
8
+ 0.4876587688922882,
9
+ 0.7196521759033203
10
+ ],
11
+ "min": [
12
+ -4.94293737411499,
13
+ -6.890198230743408,
14
+ 0.6996827125549316
15
+ ],
16
+ "mean": [
17
+ 2.749288365724937,
18
+ -2.0455216239909983,
19
+ 0.700532551020533
20
+ ],
21
+ "std": [
22
+ 1.6786295897046262,
23
+ 1.312897969434244,
24
+ 0.00133751758906258
25
+ ],
26
+ "q01": [
27
+ -1.4241907881148037,
28
+ -5.1716547935950885,
29
+ 0.6999670898768614
30
+ ],
31
+ "q99": [
32
+ 6.198111545831555,
33
+ -0.5873531334104489,
34
+ 0.7038833014242587
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.23084999587450247,
54
+ 0.6238431088581847
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6556404371644047,
60
+ 0.3572252105732042
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 1.7482354593273349e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9014528393745422,
78
+ 0.8003877401351929,
79
+ 0.9829942584037781
80
+ ],
81
+ "min": [
82
+ -0.37471598386764526,
83
+ -0.8472502827644348,
84
+ -0.25070279836654663
85
+ ],
86
+ "mean": [
87
+ 0.28667332601863704,
88
+ -0.038315473170876704,
89
+ 0.45974658906777444
90
+ ],
91
+ "std": [
92
+ 0.17067058828305154,
93
+ 0.23569952037784195,
94
+ 0.21950702354181487
95
+ ],
96
+ "q01": [
97
+ 0.015172948963784228,
98
+ -0.44450502063220615,
99
+ 0.23314444103329338
100
+ ],
101
+ "q99": [
102
+ 0.5609069689572412,
103
+ 0.38454179128009547,
104
+ 0.7028102736473852
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9984455108642578,
111
+ 0.9434727430343628,
112
+ 0.9062229990959167
113
+ ],
114
+ "min": [
115
+ -0.9999930262565613,
116
+ -0.9988337755203247,
117
+ -0.9618149995803833,
118
+ 2.1872274658107926e-07
119
+ ],
120
+ "mean": [
121
+ -0.25672297401336,
122
+ 0.024425335806687216,
123
+ -0.08740977164483847,
124
+ 0.16608739644018397
125
+ ],
126
+ "std": [
127
+ 0.7932585208392383,
128
+ 0.3102260195580416,
129
+ 0.3772472576315148,
130
+ 0.17451959300158773
131
+ ],
132
+ "q01": [
133
+ -0.99379006987882,
134
+ -0.5478102050038828,
135
+ -0.6101777952971636,
136
+ 0.002274085014134748
137
+ ],
138
+ "q99": [
139
+ 0.8804856038896288,
140
+ 0.5910611092873037,
141
+ 0.5216058042993562,
142
+ 0.5231965974012255
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.055869169533252716,
148
+ 0.010916369967162609
149
+ ],
150
+ "min": [
151
+ -0.011436971835792065,
152
+ -0.05664053186774254
153
+ ],
154
+ "mean": [
155
+ 0.031651551516052555,
156
+ -0.03161481387920421
157
+ ],
158
+ "std": [
159
+ 0.013119769222217526,
160
+ 0.01306078225192402
161
+ ],
162
+ "q01": [
163
+ 0.0060999246446777336,
164
+ -0.04062601653964198
165
+ ],
166
+ "q99": [
167
+ 0.04054891621517283,
168
+ -0.0059163567113456345
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.008144896753718373,
188
+ -0.00018893135146078263,
189
+ -0.0008739062727221845,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11030045424364411,
194
+ 0.10148082570313594,
195
+ 0.08975698373467506,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.07287373067581518,
200
+ -0.09446994199569948,
201
+ -0.07702818259249572,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.04485485414407898,
206
+ 0.09626941605965177,
207
+ 0.07610665906122742,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9221974313425939
220
+ ],
221
+ "std": [
222
+ 0.38670194342612674
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.384693946883099
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ 0.01976129923017491,
244
+ -0.020870631344490034,
245
+ -0.05993476517349181
246
+ ],
247
+ "std": [
248
+ 0.4347024990804053,
249
+ 0.4221662976569997,
250
+ 0.3845506843044066
251
+ ],
252
+ "q01": [
253
+ -0.8835792939767807,
254
+ -0.9280919251713778,
255
+ -0.783361071470956
256
+ ],
257
+ "q99": [
258
+ 0.7622084438449307,
259
+ 0.8625125252609077,
260
+ 0.7470974753250706
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008089673068544335,
276
+ -0.026498254627927348,
277
+ 0.0029083551743059005
278
+ ],
279
+ "std": [
280
+ 0.11216734503733773,
281
+ 0.13206002613798629,
282
+ 0.12615870412692864
283
+ ],
284
+ "q01": [
285
+ -0.2814627972846672,
286
+ -0.4342386516115439,
287
+ -0.34489755628340574
288
+ ],
289
+ "q99": [
290
+ 0.31879353266939386,
291
+ 0.28411797683220563,
292
+ 0.3597718671669894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3824228874769045
304
+ ],
305
+ "std": [
306
+ 0.9240015096469786
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.5597272305813592
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
clvla/1/default/resolved_config.yaml ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: clvla
2
+ policy_type: groot_rkd_v2
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: /home/ext_minje/groot_robocasa/robocasa_v2/clvla/1
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model: null
12
+ phases:
13
+ - name: phase2_rkd_da_only
14
+ max_steps: 600000
15
+ save_steps: 0
16
+ trainable:
17
+ preset: processing_line_only
18
+ tune_llm: false
19
+ tune_visual: false
20
+ tune_projector: false
21
+ tune_diffusion_model: false
22
+ losses:
23
+ rkd_enabled: true
24
+ rkd_fm_loss_weight: 0.0
25
+ rkd_loss_weight: 1.0
26
+ rkd_relation_mode: token_pair_mean
27
+ rkd_loss_type: distance_angle
28
+ rkd_distance_loss_weight: 1.0
29
+ rkd_angle_loss_weight: 2.0
30
+ rkd_exclude_diagonal: true
31
+ - name: phase3_fm_rkd_da_fixed_0p5
32
+ max_steps: 200000
33
+ save_steps: 0
34
+ trainable:
35
+ tune_llm: false
36
+ tune_visual: false
37
+ tune_projector: true
38
+ tune_diffusion_model: true
39
+ losses:
40
+ rkd_enabled: true
41
+ rkd_fm_loss_weight: 1.0
42
+ rkd_loss_weight: 0.5
43
+ rkd_relation_mode: token_pair_mean
44
+ rkd_loss_type: distance_angle
45
+ rkd_distance_loss_weight: 1.0
46
+ rkd_angle_loss_weight: 2.0
47
+ rkd_exclude_diagonal: true
48
+ resolved_sweep: {}
clvla/2/default/clvla.yaml ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: clvla
2
+ policy_type: groot_rkd_v2
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: /home/ext_minje/groot_robocasa/robocasa_v2/clvla/2
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model: null
12
+ phases:
13
+ - name: phase2_rkd_da_only
14
+ max_steps: 600000
15
+ save_steps: 0
16
+ trainable:
17
+ preset: processing_line_only
18
+ tune_llm: false
19
+ tune_visual: false
20
+ tune_projector: false
21
+ tune_diffusion_model: false
22
+ losses:
23
+ rkd_enabled: true
24
+ rkd_fm_loss_weight: 0.0
25
+ rkd_loss_weight: 1.0
26
+ rkd_relation_mode: token_pair_mean
27
+ rkd_loss_type: distance_angle
28
+ rkd_distance_loss_weight: 1.0
29
+ rkd_angle_loss_weight: 2.0
30
+ rkd_exclude_diagonal: true
31
+ - name: phase3_fm_rkd_da_fixed_0p5
32
+ max_steps: 200000
33
+ save_steps: 0
34
+ trainable:
35
+ tune_llm: false
36
+ tune_visual: false
37
+ tune_projector: true
38
+ tune_diffusion_model: true
39
+ losses:
40
+ rkd_enabled: true
41
+ rkd_fm_loss_weight: 1.0
42
+ rkd_loss_weight: 0.5
43
+ rkd_relation_mode: token_pair_mean
44
+ rkd_loss_type: distance_angle
45
+ rkd_distance_loss_weight: 1.0
46
+ rkd_angle_loss_weight: 2.0
47
+ rkd_exclude_diagonal: true
clvla/2/default/phase2/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.3139495849609375,
8
+ 0.4876587688922882,
9
+ 0.7196521759033203
10
+ ],
11
+ "min": [
12
+ -4.94293737411499,
13
+ -6.890198230743408,
14
+ 0.6996827125549316
15
+ ],
16
+ "mean": [
17
+ 2.749288365724937,
18
+ -2.0455216239909983,
19
+ 0.700532551020533
20
+ ],
21
+ "std": [
22
+ 1.6786295897046262,
23
+ 1.312897969434244,
24
+ 0.00133751758906258
25
+ ],
26
+ "q01": [
27
+ -1.4241907881148037,
28
+ -5.1716547935950885,
29
+ 0.6999670898768614
30
+ ],
31
+ "q99": [
32
+ 6.198111545831555,
33
+ -0.5873531334104489,
34
+ 0.7038833014242587
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.23084999587450247,
54
+ 0.6238431088581847
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6556404371644047,
60
+ 0.3572252105732042
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 1.7482354593273349e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9014528393745422,
78
+ 0.8003877401351929,
79
+ 0.9829942584037781
80
+ ],
81
+ "min": [
82
+ -0.37471598386764526,
83
+ -0.8472502827644348,
84
+ -0.25070279836654663
85
+ ],
86
+ "mean": [
87
+ 0.28667332601863704,
88
+ -0.038315473170876704,
89
+ 0.45974658906777444
90
+ ],
91
+ "std": [
92
+ 0.17067058828305154,
93
+ 0.23569952037784195,
94
+ 0.21950702354181487
95
+ ],
96
+ "q01": [
97
+ 0.015172948963784228,
98
+ -0.44450502063220615,
99
+ 0.23314444103329338
100
+ ],
101
+ "q99": [
102
+ 0.5609069689572412,
103
+ 0.38454179128009547,
104
+ 0.7028102736473852
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9984455108642578,
111
+ 0.9434727430343628,
112
+ 0.9062229990959167
113
+ ],
114
+ "min": [
115
+ -0.9999930262565613,
116
+ -0.9988337755203247,
117
+ -0.9618149995803833,
118
+ 2.1872274658107926e-07
119
+ ],
120
+ "mean": [
121
+ -0.25672297401336,
122
+ 0.024425335806687216,
123
+ -0.08740977164483847,
124
+ 0.16608739644018397
125
+ ],
126
+ "std": [
127
+ 0.7932585208392383,
128
+ 0.3102260195580416,
129
+ 0.3772472576315148,
130
+ 0.17451959300158773
131
+ ],
132
+ "q01": [
133
+ -0.99379006987882,
134
+ -0.5478102050038828,
135
+ -0.6101777952971636,
136
+ 0.002274085014134748
137
+ ],
138
+ "q99": [
139
+ 0.8804856038896288,
140
+ 0.5910611092873037,
141
+ 0.5216058042993562,
142
+ 0.5231965974012255
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.055869169533252716,
148
+ 0.010916369967162609
149
+ ],
150
+ "min": [
151
+ -0.011436971835792065,
152
+ -0.05664053186774254
153
+ ],
154
+ "mean": [
155
+ 0.031651551516052555,
156
+ -0.03161481387920421
157
+ ],
158
+ "std": [
159
+ 0.013119769222217526,
160
+ 0.01306078225192402
161
+ ],
162
+ "q01": [
163
+ 0.0060999246446777336,
164
+ -0.04062601653964198
165
+ ],
166
+ "q99": [
167
+ 0.04054891621517283,
168
+ -0.0059163567113456345
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.008144896753718373,
188
+ -0.00018893135146078263,
189
+ -0.0008739062727221845,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11030045424364411,
194
+ 0.10148082570313594,
195
+ 0.08975698373467506,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.07287373067581518,
200
+ -0.09446994199569948,
201
+ -0.07702818259249572,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.04485485414407898,
206
+ 0.09626941605965177,
207
+ 0.07610665906122742,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9221974313425939
220
+ ],
221
+ "std": [
222
+ 0.38670194342612674
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.384693946883099
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ 0.01976129923017491,
244
+ -0.020870631344490034,
245
+ -0.05993476517349181
246
+ ],
247
+ "std": [
248
+ 0.4347024990804053,
249
+ 0.4221662976569997,
250
+ 0.3845506843044066
251
+ ],
252
+ "q01": [
253
+ -0.8835792939767807,
254
+ -0.9280919251713778,
255
+ -0.783361071470956
256
+ ],
257
+ "q99": [
258
+ 0.7622084438449307,
259
+ 0.8625125252609077,
260
+ 0.7470974753250706
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008089673068544335,
276
+ -0.026498254627927348,
277
+ 0.0029083551743059005
278
+ ],
279
+ "std": [
280
+ 0.11216734503733773,
281
+ 0.13206002613798629,
282
+ 0.12615870412692864
283
+ ],
284
+ "q01": [
285
+ -0.2814627972846672,
286
+ -0.4342386516115439,
287
+ -0.34489755628340574
288
+ ],
289
+ "q99": [
290
+ 0.31879353266939386,
291
+ 0.28411797683220563,
292
+ 0.3597718671669894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3824228874769045
304
+ ],
305
+ "std": [
306
+ 0.9240015096469786
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.5597272305813592
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
clvla/2/default/resolved_config.yaml ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: clvla
2
+ policy_type: groot_rkd_v2
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: /home/ext_minje/groot_robocasa/robocasa_v2/clvla/2
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model: null
12
+ phases:
13
+ - name: phase2_rkd_da_only
14
+ max_steps: 600000
15
+ save_steps: 0
16
+ trainable:
17
+ preset: processing_line_only
18
+ tune_llm: false
19
+ tune_visual: false
20
+ tune_projector: false
21
+ tune_diffusion_model: false
22
+ losses:
23
+ rkd_enabled: true
24
+ rkd_fm_loss_weight: 0.0
25
+ rkd_loss_weight: 1.0
26
+ rkd_relation_mode: token_pair_mean
27
+ rkd_loss_type: distance_angle
28
+ rkd_distance_loss_weight: 1.0
29
+ rkd_angle_loss_weight: 2.0
30
+ rkd_exclude_diagonal: true
31
+ - name: phase3_fm_rkd_da_fixed_0p5
32
+ max_steps: 200000
33
+ save_steps: 0
34
+ trainable:
35
+ tune_llm: false
36
+ tune_visual: false
37
+ tune_projector: true
38
+ tune_diffusion_model: true
39
+ losses:
40
+ rkd_enabled: true
41
+ rkd_fm_loss_weight: 1.0
42
+ rkd_loss_weight: 0.5
43
+ rkd_relation_mode: token_pair_mean
44
+ rkd_loss_type: distance_angle
45
+ rkd_distance_loss_weight: 1.0
46
+ rkd_angle_loss_weight: 2.0
47
+ rkd_exclude_diagonal: true
48
+ resolved_sweep: {}