anhtld commited on
Commit
5994bfa
·
verified ·
1 Parent(s): de5b6bd

Auto-sync: 2026-06-27 18:32:13

Browse files
results/h16_field_optim_near_miss_policy_bc5_bestpt_s4_trust05_afterany_summary.json ADDED
@@ -0,0 +1,746 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05",
3
+ "baseline_h4_policy_success": 0.2967,
4
+ "baseline_h16_policy_success": 0.29739130434782607,
5
+ "num_result_files": 9,
6
+ "configs": [
7
+ {
8
+ "config": "k32_sigma0.50",
9
+ "completed_seeds": [
10
+ 0
11
+ ],
12
+ "num_completed": 1,
13
+ "mean_success": 0.2539130434782609,
14
+ "std_success": 0.0,
15
+ "mean_progress": 0.5365319993136369,
16
+ "mean_action_mse_to_best": 0.3908262099843958,
17
+ "absolute_gain_vs_h4_baseline": -0.04278695652173914,
18
+ "absolute_gain_vs_h16_policy": -0.04347826086956519
19
+ },
20
+ {
21
+ "config": "k1_sigma0.00",
22
+ "completed_seeds": [
23
+ 0,
24
+ 2
25
+ ],
26
+ "num_completed": 2,
27
+ "mean_success": 0.24608695652173912,
28
+ "std_success": 0.00614875461901348,
29
+ "mean_progress": 0.5327426943620261,
30
+ "mean_action_mse_to_best": 0.39773258084512275,
31
+ "absolute_gain_vs_h4_baseline": -0.0506130434782609,
32
+ "absolute_gain_vs_h16_policy": -0.05130434782608695
33
+ },
34
+ {
35
+ "config": "k16_sigma0.35",
36
+ "completed_seeds": [
37
+ 0,
38
+ 1,
39
+ 2
40
+ ],
41
+ "num_completed": 3,
42
+ "mean_success": 0.24231884057971015,
43
+ "std_success": 0.002008174849355231,
44
+ "mean_progress": 0.5237070613686257,
45
+ "mean_action_mse_to_best": 0.4064106829307865,
46
+ "absolute_gain_vs_h4_baseline": -0.05438115942028987,
47
+ "absolute_gain_vs_h16_policy": -0.05507246376811592
48
+ },
49
+ {
50
+ "config": "k8_sigma0.20",
51
+ "completed_seeds": [
52
+ 0,
53
+ 1,
54
+ 2
55
+ ],
56
+ "num_completed": 3,
57
+ "mean_success": 0.24115942028985507,
58
+ "std_success": 0.011315490606343844,
59
+ "mean_progress": 0.5236840639907219,
60
+ "mean_action_mse_to_best": 0.4043724872630791,
61
+ "absolute_gain_vs_h4_baseline": -0.05554057971014495,
62
+ "absolute_gain_vs_h16_policy": -0.056231884057971
63
+ }
64
+ ],
65
+ "best": {
66
+ "config": "k32_sigma0.50",
67
+ "completed_seeds": [
68
+ 0
69
+ ],
70
+ "num_completed": 1,
71
+ "mean_success": 0.2539130434782609,
72
+ "std_success": 0.0,
73
+ "mean_progress": 0.5365319993136369,
74
+ "mean_action_mse_to_best": 0.3908262099843958,
75
+ "absolute_gain_vs_h4_baseline": -0.04278695652173914,
76
+ "absolute_gain_vs_h16_policy": -0.04347826086956519
77
+ },
78
+ "rows": [
79
+ {
80
+ "config": "k16_sigma0.35",
81
+ "seed": 0,
82
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k16_sigma0.35/seed_0/online_rollout.json",
83
+ "selection_mode": "field_optim",
84
+ "num_candidates": 16,
85
+ "candidate_sigma": 0.35,
86
+ "field_optim_steps": 4,
87
+ "field_optim_step_size": 0.05,
88
+ "field_optim_trust_radius": 0.5,
89
+ "field_optim_l2_penalty": 0.02,
90
+ "num_groups": 575,
91
+ "policy_rollout_success_rate": 0.24347826086956523,
92
+ "policy_rollout_progress": 0.5289620748977515,
93
+ "oracle_success_rate": 0.8573913043478261,
94
+ "action_mse_to_best": 0.390246681454389,
95
+ "per_task": {
96
+ "LiftPegUpright-v1": {
97
+ "action_mse_to_best": 0.31632087363371836,
98
+ "expert_success_rate": 0.8865979381443299,
99
+ "num_groups": 97,
100
+ "oracle_success_rate": 0.9278350515463918,
101
+ "policy_expert_regret": 1.0068789176719706,
102
+ "policy_oracle_regret": 1.033257112982347,
103
+ "policy_rollout_progress": 0.6514236106700504,
104
+ "policy_rollout_success_rate": 0.25773195876288657,
105
+ "restore_max_error": 4.76837158203125e-07
106
+ },
107
+ "PickCube-v1": {
108
+ "action_mse_to_best": 0.3099510838779119,
109
+ "expert_success_rate": 0.9375,
110
+ "num_groups": 208,
111
+ "oracle_success_rate": 0.9471153846153846,
112
+ "policy_expert_regret": 1.3923178799987699,
113
+ "policy_oracle_regret": 1.4057407474252754,
114
+ "policy_rollout_progress": 0.47146958250624055,
115
+ "policy_rollout_success_rate": 0.04807692307692308,
116
+ "restore_max_error": 4.76837158203125e-07
117
+ },
118
+ "PullCube-v1": {
119
+ "action_mse_to_best": 0.6181663733411145,
120
+ "expert_success_rate": 0.19480519480519481,
121
+ "num_groups": 77,
122
+ "oracle_success_rate": 0.36363636363636365,
123
+ "policy_expert_regret": 0.25833353278876253,
124
+ "policy_oracle_regret": 0.3980089887791188,
125
+ "policy_rollout_progress": 0.3531065471752575,
126
+ "policy_rollout_success_rate": 0.23376623376623376,
127
+ "restore_max_error": 4.76837158203125e-07
128
+ },
129
+ "PushCube-v1": {
130
+ "action_mse_to_best": 0.40011150191756006,
131
+ "expert_success_rate": 0.8617021276595744,
132
+ "num_groups": 94,
133
+ "oracle_success_rate": 0.9893617021276596,
134
+ "policy_expert_regret": 0.3669963699389011,
135
+ "policy_oracle_regret": 0.4460776372158781,
136
+ "policy_rollout_progress": 0.7879649159756112,
137
+ "policy_rollout_success_rate": 0.7659574468085106,
138
+ "restore_max_error": 4.76837158203125e-07
139
+ },
140
+ "StackCube-v1": {
141
+ "action_mse_to_best": 0.44474343151192774,
142
+ "expert_success_rate": 0.6767676767676768,
143
+ "num_groups": 99,
144
+ "oracle_success_rate": 0.8585858585858586,
145
+ "policy_expert_regret": 1.0384967644708325,
146
+ "policy_oracle_regret": 1.249219367449934,
147
+ "policy_rollout_progress": 0.42062144884557434,
148
+ "policy_rollout_success_rate": 0.15151515151515152,
149
+ "restore_max_error": 3.948807716369629e-07
150
+ }
151
+ }
152
+ },
153
+ {
154
+ "config": "k16_sigma0.35",
155
+ "seed": 1,
156
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k16_sigma0.35/seed_1/online_rollout.json",
157
+ "selection_mode": "field_optim",
158
+ "num_candidates": 16,
159
+ "candidate_sigma": 0.35,
160
+ "field_optim_steps": 4,
161
+ "field_optim_step_size": 0.05,
162
+ "field_optim_trust_radius": 0.5,
163
+ "field_optim_l2_penalty": 0.02,
164
+ "num_groups": 575,
165
+ "policy_rollout_success_rate": 0.24347826086956523,
166
+ "policy_rollout_progress": 0.5165389964462298,
167
+ "oracle_success_rate": 0.8695652173913043,
168
+ "action_mse_to_best": 0.4029614827927688,
169
+ "per_task": {
170
+ "LiftPegUpright-v1": {
171
+ "action_mse_to_best": 0.3265482701146893,
172
+ "expert_success_rate": 0.8584070796460177,
173
+ "num_groups": 113,
174
+ "oracle_success_rate": 0.9380530973451328,
175
+ "policy_expert_regret": 0.925960123538971,
176
+ "policy_oracle_regret": 1.0071366368669323,
177
+ "policy_rollout_progress": 0.6748347817796522,
178
+ "policy_rollout_success_rate": 0.23893805309734514,
179
+ "restore_max_error": 4.76837158203125e-07
180
+ },
181
+ "PickCube-v1": {
182
+ "action_mse_to_best": 0.3257596911660031,
183
+ "expert_success_rate": 0.9402173913043478,
184
+ "num_groups": 184,
185
+ "oracle_success_rate": 0.9456521739130435,
186
+ "policy_expert_regret": 1.4697572176835663,
187
+ "policy_oracle_regret": 1.4772941226509158,
188
+ "policy_rollout_progress": 0.43528266355101747,
189
+ "policy_rollout_success_rate": 0.010869565217391304,
190
+ "restore_max_error": 4.76837158203125e-07
191
+ },
192
+ "PullCube-v1": {
193
+ "action_mse_to_best": 0.5878818454711061,
194
+ "expert_success_rate": 0.25,
195
+ "num_groups": 76,
196
+ "oracle_success_rate": 0.40789473684210525,
197
+ "policy_expert_regret": 0.28736334068602637,
198
+ "policy_oracle_regret": 0.5531476327267132,
199
+ "policy_rollout_progress": 0.28927135144017246,
200
+ "policy_rollout_success_rate": 0.15789473684210525,
201
+ "restore_max_error": 4.76837158203125e-07
202
+ },
203
+ "PushCube-v1": {
204
+ "action_mse_to_best": 0.40456463876474014,
205
+ "expert_success_rate": 0.8198198198198198,
206
+ "num_groups": 111,
207
+ "oracle_success_rate": 1.0,
208
+ "policy_expert_regret": 0.36372163905216764,
209
+ "policy_oracle_regret": 0.4141334221438245,
210
+ "policy_rollout_progress": 0.8020827940723918,
211
+ "policy_rollout_success_rate": 0.7837837837837838,
212
+ "restore_max_error": 4.76837158203125e-07
213
+ },
214
+ "StackCube-v1": {
215
+ "action_mse_to_best": 0.4975540631829382,
216
+ "expert_success_rate": 0.6923076923076923,
217
+ "num_groups": 91,
218
+ "oracle_success_rate": 0.8571428571428571,
219
+ "policy_expert_regret": 1.1893479218194773,
220
+ "policy_oracle_regret": 1.3616850798601632,
221
+ "policy_rollout_progress": 0.3257776886879743,
222
+ "policy_rollout_success_rate": 0.13186813186813187,
223
+ "restore_max_error": 3.948807716369629e-07
224
+ }
225
+ }
226
+ },
227
+ {
228
+ "config": "k16_sigma0.35",
229
+ "seed": 2,
230
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k16_sigma0.35/seed_2/online_rollout.json",
231
+ "selection_mode": "field_optim",
232
+ "num_candidates": 16,
233
+ "candidate_sigma": 0.35,
234
+ "field_optim_steps": 4,
235
+ "field_optim_step_size": 0.05,
236
+ "field_optim_trust_radius": 0.5,
237
+ "field_optim_l2_penalty": 0.02,
238
+ "num_groups": 575,
239
+ "policy_rollout_success_rate": 0.24,
240
+ "policy_rollout_progress": 0.5256201127618956,
241
+ "oracle_success_rate": 0.8765217391304347,
242
+ "action_mse_to_best": 0.42602388454520185,
243
+ "per_task": {
244
+ "LiftPegUpright-v1": {
245
+ "action_mse_to_best": 0.3519886921470364,
246
+ "expert_success_rate": 0.8229166666666666,
247
+ "num_groups": 96,
248
+ "oracle_success_rate": 0.9270833333333334,
249
+ "policy_expert_regret": 0.9503137821642061,
250
+ "policy_oracle_regret": 1.0841905994651218,
251
+ "policy_rollout_progress": 0.6382713741622865,
252
+ "policy_rollout_success_rate": 0.19791666666666666,
253
+ "restore_max_error": 3.5762786865234375e-07
254
+ },
255
+ "PickCube-v1": {
256
+ "action_mse_to_best": 0.34274348870597104,
257
+ "expert_success_rate": 0.9444444444444444,
258
+ "num_groups": 198,
259
+ "oracle_success_rate": 0.9595959595959596,
260
+ "policy_expert_regret": 1.3842022730375056,
261
+ "policy_oracle_regret": 1.4055294888770218,
262
+ "policy_rollout_progress": 0.4946433123032976,
263
+ "policy_rollout_success_rate": 0.04040404040404041,
264
+ "restore_max_error": 4.76837158203125e-07
265
+ },
266
+ "PullCube-v1": {
267
+ "action_mse_to_best": 0.6682874644796054,
268
+ "expert_success_rate": 0.24444444444444444,
269
+ "num_groups": 90,
270
+ "oracle_success_rate": 0.4666666666666667,
271
+ "policy_expert_regret": 0.3146628489717841,
272
+ "policy_oracle_regret": 0.5527156790097555,
273
+ "policy_rollout_progress": 0.3432616694105996,
274
+ "policy_rollout_success_rate": 0.2222222222222222,
275
+ "restore_max_error": 4.0978193283081055e-07
276
+ },
277
+ "PushCube-v1": {
278
+ "action_mse_to_best": 0.3917583738370697,
279
+ "expert_success_rate": 0.8514851485148515,
280
+ "num_groups": 101,
281
+ "oracle_success_rate": 1.0,
282
+ "policy_expert_regret": 0.32143961734110765,
283
+ "policy_oracle_regret": 0.3768879201742682,
284
+ "policy_rollout_progress": 0.8211318818059298,
285
+ "policy_rollout_success_rate": 0.801980198019802,
286
+ "restore_max_error": 4.76837158203125e-07
287
+ },
288
+ "StackCube-v1": {
289
+ "action_mse_to_best": 0.484401564920942,
290
+ "expert_success_rate": 0.7666666666666667,
291
+ "num_groups": 90,
292
+ "oracle_success_rate": 0.9111111111111111,
293
+ "policy_expert_regret": 1.3108843771947756,
294
+ "policy_oracle_regret": 1.4598074724276862,
295
+ "policy_rollout_progress": 0.32433629747894077,
296
+ "policy_rollout_success_rate": 0.1111111111111111,
297
+ "restore_max_error": 4.76837158203125e-07
298
+ }
299
+ }
300
+ },
301
+ {
302
+ "config": "k1_sigma0.00",
303
+ "seed": 0,
304
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k1_sigma0.00/seed_0/online_rollout.json",
305
+ "selection_mode": "field_optim",
306
+ "num_candidates": 1,
307
+ "candidate_sigma": 0.0,
308
+ "field_optim_steps": 4,
309
+ "field_optim_step_size": 0.05,
310
+ "field_optim_trust_radius": 0.5,
311
+ "field_optim_l2_penalty": 0.02,
312
+ "num_groups": 575,
313
+ "policy_rollout_success_rate": 0.25043478260869567,
314
+ "policy_rollout_progress": 0.5385008239519337,
315
+ "oracle_success_rate": 0.8573913043478261,
316
+ "action_mse_to_best": 0.38061382845044134,
317
+ "per_task": {
318
+ "LiftPegUpright-v1": {
319
+ "action_mse_to_best": 0.31332231308189556,
320
+ "expert_success_rate": 0.8865979381443299,
321
+ "num_groups": 97,
322
+ "oracle_success_rate": 0.9278350515463918,
323
+ "policy_expert_regret": 0.9673091362124866,
324
+ "policy_oracle_regret": 0.9936873315228629,
325
+ "policy_rollout_progress": 0.6600650664150101,
326
+ "policy_rollout_success_rate": 0.28865979381443296,
327
+ "restore_max_error": 4.76837158203125e-07
328
+ },
329
+ "PickCube-v1": {
330
+ "action_mse_to_best": 0.3059184052981436,
331
+ "expert_success_rate": 0.9375,
332
+ "num_groups": 208,
333
+ "oracle_success_rate": 0.9471153846153846,
334
+ "policy_expert_regret": 1.3921011273587982,
335
+ "policy_oracle_regret": 1.4055239947853038,
336
+ "policy_rollout_progress": 0.4716863351462122,
337
+ "policy_rollout_success_rate": 0.04807692307692308,
338
+ "restore_max_error": 4.76837158203125e-07
339
+ },
340
+ "PullCube-v1": {
341
+ "action_mse_to_best": 0.6041168522138101,
342
+ "expert_success_rate": 0.19480519480519481,
343
+ "num_groups": 77,
344
+ "oracle_success_rate": 0.36363636363636365,
345
+ "policy_expert_regret": 0.2339665520597588,
346
+ "policy_oracle_regret": 0.38729304597749337,
347
+ "policy_rollout_progress": 0.3639409321469146,
348
+ "policy_rollout_success_rate": 0.23376623376623376,
349
+ "restore_max_error": 4.76837158203125e-07
350
+ },
351
+ "PushCube-v1": {
352
+ "action_mse_to_best": 0.36802108233120845,
353
+ "expert_success_rate": 0.8617021276595744,
354
+ "num_groups": 94,
355
+ "oracle_success_rate": 0.9893617021276596,
356
+ "policy_expert_regret": 0.3255824437166782,
357
+ "policy_oracle_regret": 0.3837234648935338,
358
+ "policy_rollout_progress": 0.8184041946809343,
359
+ "policy_rollout_success_rate": 0.7978723404255319,
360
+ "restore_max_error": 4.76837158203125e-07
361
+ },
362
+ "StackCube-v1": {
363
+ "action_mse_to_best": 0.44160282149948554,
364
+ "expert_success_rate": 0.6767676767676768,
365
+ "num_groups": 99,
366
+ "oracle_success_rate": 0.8585858585858586,
367
+ "policy_expert_regret": 1.0494662336628846,
368
+ "policy_oracle_regret": 1.2601833939552307,
369
+ "policy_rollout_progress": 0.42977230852902537,
370
+ "policy_rollout_success_rate": 0.13131313131313133,
371
+ "restore_max_error": 3.948807716369629e-07
372
+ }
373
+ }
374
+ },
375
+ {
376
+ "config": "k1_sigma0.00",
377
+ "seed": 2,
378
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k1_sigma0.00/seed_2/online_rollout.json",
379
+ "selection_mode": "field_optim",
380
+ "num_candidates": 1,
381
+ "candidate_sigma": 0.0,
382
+ "field_optim_steps": 4,
383
+ "field_optim_step_size": 0.05,
384
+ "field_optim_trust_radius": 0.5,
385
+ "field_optim_l2_penalty": 0.02,
386
+ "num_groups": 575,
387
+ "policy_rollout_success_rate": 0.2417391304347826,
388
+ "policy_rollout_progress": 0.5269845647721187,
389
+ "oracle_success_rate": 0.8765217391304347,
390
+ "action_mse_to_best": 0.41485133323980417,
391
+ "per_task": {
392
+ "LiftPegUpright-v1": {
393
+ "action_mse_to_best": 0.3479427481070161,
394
+ "expert_success_rate": 0.8229166666666666,
395
+ "num_groups": 96,
396
+ "oracle_success_rate": 0.9270833333333334,
397
+ "policy_expert_regret": 0.9770944616757333,
398
+ "policy_oracle_regret": 1.1107273153029382,
399
+ "policy_rollout_progress": 0.6354146927284697,
400
+ "policy_rollout_success_rate": 0.1875,
401
+ "restore_max_error": 3.5762786865234375e-07
402
+ },
403
+ "PickCube-v1": {
404
+ "action_mse_to_best": 0.34223038080440027,
405
+ "expert_success_rate": 0.9444444444444444,
406
+ "num_groups": 198,
407
+ "oracle_success_rate": 0.9595959595959596,
408
+ "policy_expert_regret": 1.384397162143329,
409
+ "policy_oracle_regret": 1.4057243779828452,
410
+ "policy_rollout_progress": 0.4944484231974741,
411
+ "policy_rollout_success_rate": 0.04040404040404041,
412
+ "restore_max_error": 4.76837158203125e-07
413
+ },
414
+ "PullCube-v1": {
415
+ "action_mse_to_best": 0.6251726786295573,
416
+ "expert_success_rate": 0.24444444444444444,
417
+ "num_groups": 90,
418
+ "oracle_success_rate": 0.4666666666666667,
419
+ "policy_expert_regret": 0.3070303386905127,
420
+ "policy_oracle_regret": 0.5933954045590427,
421
+ "policy_rollout_progress": 0.311150934588578,
422
+ "policy_rollout_success_rate": 0.17777777777777778,
423
+ "restore_max_error": 4.0978193283081055e-07
424
+ },
425
+ "PushCube-v1": {
426
+ "action_mse_to_best": 0.3742850503266448,
427
+ "expert_success_rate": 0.8514851485148515,
428
+ "num_groups": 101,
429
+ "oracle_success_rate": 1.0,
430
+ "policy_expert_regret": 0.28293753555505585,
431
+ "policy_oracle_regret": 0.31926996253504614,
432
+ "policy_rollout_progress": 0.8490468691481222,
433
+ "policy_rollout_success_rate": 0.8316831683168316,
434
+ "restore_max_error": 4.76837158203125e-07
435
+ },
436
+ "StackCube-v1": {
437
+ "action_mse_to_best": 0.4811896248410145,
438
+ "expert_success_rate": 0.7666666666666667,
439
+ "num_groups": 90,
440
+ "oracle_success_rate": 0.9111111111111111,
441
+ "policy_expert_regret": 1.2645780030224059,
442
+ "policy_oracle_regret": 1.41350101199415,
443
+ "policy_rollout_progress": 0.33731342835558786,
444
+ "policy_rollout_success_rate": 0.14444444444444443,
445
+ "restore_max_error": 4.76837158203125e-07
446
+ }
447
+ }
448
+ },
449
+ {
450
+ "config": "k32_sigma0.50",
451
+ "seed": 0,
452
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k32_sigma0.50/seed_0/online_rollout.json",
453
+ "selection_mode": "field_optim",
454
+ "num_candidates": 32,
455
+ "candidate_sigma": 0.5,
456
+ "field_optim_steps": 4,
457
+ "field_optim_step_size": 0.05,
458
+ "field_optim_trust_radius": 0.5,
459
+ "field_optim_l2_penalty": 0.02,
460
+ "num_groups": 575,
461
+ "policy_rollout_success_rate": 0.2539130434782609,
462
+ "policy_rollout_progress": 0.5365319993136369,
463
+ "oracle_success_rate": 0.8573913043478261,
464
+ "action_mse_to_best": 0.3908262099843958,
465
+ "per_task": {
466
+ "LiftPegUpright-v1": {
467
+ "action_mse_to_best": 0.3133223102687253,
468
+ "expert_success_rate": 0.8865979381443299,
469
+ "num_groups": 97,
470
+ "oracle_success_rate": 0.9278350515463918,
471
+ "policy_expert_regret": 0.9540133671355002,
472
+ "policy_oracle_regret": 0.9803915624458766,
473
+ "policy_rollout_progress": 0.6630521756164807,
474
+ "policy_rollout_success_rate": 0.29896907216494845,
475
+ "restore_max_error": 4.76837158203125e-07
476
+ },
477
+ "PickCube-v1": {
478
+ "action_mse_to_best": 0.3123190854496967,
479
+ "expert_success_rate": 0.9375,
480
+ "num_groups": 208,
481
+ "oracle_success_rate": 0.9471153846153846,
482
+ "policy_expert_regret": 1.3941655335634446,
483
+ "policy_oracle_regret": 1.40758840098995,
484
+ "policy_rollout_progress": 0.4696219289415659,
485
+ "policy_rollout_success_rate": 0.04807692307692308,
486
+ "restore_max_error": 4.76837158203125e-07
487
+ },
488
+ "PullCube-v1": {
489
+ "action_mse_to_best": 0.6222828338285545,
490
+ "expert_success_rate": 0.19480519480519481,
491
+ "num_groups": 77,
492
+ "oracle_success_rate": 0.36363636363636365,
493
+ "policy_expert_regret": 0.2396799704519572,
494
+ "policy_oracle_regret": 0.35152300004544973,
495
+ "policy_rollout_progress": 0.3750774884330375,
496
+ "policy_rollout_success_rate": 0.2597402597402597,
497
+ "restore_max_error": 4.76837158203125e-07
498
+ },
499
+ "PushCube-v1": {
500
+ "action_mse_to_best": 0.4014465888129904,
501
+ "expert_success_rate": 0.8617021276595744,
502
+ "num_groups": 94,
503
+ "oracle_success_rate": 0.9893617021276596,
504
+ "policy_expert_regret": 0.38544646381063663,
505
+ "policy_oracle_regret": 0.4258071031976253,
506
+ "policy_rollout_progress": 0.7975971521215236,
507
+ "policy_rollout_success_rate": 0.776595744680851,
508
+ "restore_max_error": 4.76837158203125e-07
509
+ },
510
+ "StackCube-v1": {
511
+ "action_mse_to_best": 0.44160282139600526,
512
+ "expert_success_rate": 0.6767676767676768,
513
+ "num_groups": 99,
514
+ "oracle_success_rate": 0.8585858585858586,
515
+ "policy_expert_regret": 1.0384830933327627,
516
+ "policy_oracle_regret": 1.249108613289968,
517
+ "policy_rollout_progress": 0.43084220648413957,
518
+ "policy_rollout_success_rate": 0.1414141414141414,
519
+ "restore_max_error": 3.948807716369629e-07
520
+ }
521
+ }
522
+ },
523
+ {
524
+ "config": "k8_sigma0.20",
525
+ "seed": 0,
526
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k8_sigma0.20/seed_0/online_rollout.json",
527
+ "selection_mode": "field_optim",
528
+ "num_candidates": 8,
529
+ "candidate_sigma": 0.2,
530
+ "field_optim_steps": 4,
531
+ "field_optim_step_size": 0.05,
532
+ "field_optim_trust_radius": 0.5,
533
+ "field_optim_l2_penalty": 0.02,
534
+ "num_groups": 575,
535
+ "policy_rollout_success_rate": 0.2417391304347826,
536
+ "policy_rollout_progress": 0.5251557968394912,
537
+ "oracle_success_rate": 0.8573913043478261,
538
+ "action_mse_to_best": 0.3893296971884759,
539
+ "per_task": {
540
+ "LiftPegUpright-v1": {
541
+ "action_mse_to_best": 0.3167373553859358,
542
+ "expert_success_rate": 0.8865979381443299,
543
+ "num_groups": 97,
544
+ "oracle_success_rate": 0.9278350515463918,
545
+ "policy_expert_regret": 1.002053538855818,
546
+ "policy_oracle_regret": 1.0284317341661944,
547
+ "policy_rollout_progress": 0.6562606996482181,
548
+ "policy_rollout_success_rate": 0.25773195876288657,
549
+ "restore_max_error": 4.76837158203125e-07
550
+ },
551
+ "PickCube-v1": {
552
+ "action_mse_to_best": 0.3084647181049849,
553
+ "expert_success_rate": 0.9375,
554
+ "num_groups": 208,
555
+ "oracle_success_rate": 0.9471153846153846,
556
+ "policy_expert_regret": 1.3943369292851107,
557
+ "policy_oracle_regret": 1.4077597967116162,
558
+ "policy_rollout_progress": 0.4694505332198997,
559
+ "policy_rollout_success_rate": 0.04807692307692308,
560
+ "restore_max_error": 4.76837158203125e-07
561
+ },
562
+ "PullCube-v1": {
563
+ "action_mse_to_best": 0.6139777099544351,
564
+ "expert_success_rate": 0.19480519480519481,
565
+ "num_groups": 77,
566
+ "oracle_success_rate": 0.36363636363636365,
567
+ "policy_expert_regret": 0.23649730296297508,
568
+ "policy_oracle_regret": 0.4383279656047945,
569
+ "policy_rollout_progress": 0.3380576056706441,
570
+ "policy_rollout_success_rate": 0.2077922077922078,
571
+ "restore_max_error": 4.76837158203125e-07
572
+ },
573
+ "PushCube-v1": {
574
+ "action_mse_to_best": 0.39689942821860313,
575
+ "expert_success_rate": 0.8617021276595744,
576
+ "num_groups": 94,
577
+ "oracle_success_rate": 0.9893617021276596,
578
+ "policy_expert_regret": 0.4052272028428443,
579
+ "policy_oracle_regret": 0.4834587272811443,
580
+ "policy_rollout_progress": 0.7718604216550259,
581
+ "policy_rollout_success_rate": 0.7446808510638298,
582
+ "restore_max_error": 4.76837158203125e-07
583
+ },
584
+ "StackCube-v1": {
585
+ "action_mse_to_best": 0.44844001137435135,
586
+ "expert_success_rate": 0.6767676767676768,
587
+ "num_groups": 99,
588
+ "oracle_success_rate": 0.8585858585858586,
589
+ "policy_expert_regret": 1.0507651723996558,
590
+ "policy_oracle_regret": 1.2260844364310757,
591
+ "policy_rollout_progress": 0.4250128195442335,
592
+ "policy_rollout_success_rate": 0.18181818181818182,
593
+ "restore_max_error": 3.948807716369629e-07
594
+ }
595
+ }
596
+ },
597
+ {
598
+ "config": "k8_sigma0.20",
599
+ "seed": 1,
600
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k8_sigma0.20/seed_1/online_rollout.json",
601
+ "selection_mode": "field_optim",
602
+ "num_candidates": 8,
603
+ "candidate_sigma": 0.2,
604
+ "field_optim_steps": 4,
605
+ "field_optim_step_size": 0.05,
606
+ "field_optim_trust_radius": 0.5,
607
+ "field_optim_l2_penalty": 0.02,
608
+ "num_groups": 575,
609
+ "policy_rollout_success_rate": 0.25217391304347825,
610
+ "policy_rollout_progress": 0.52408728260142,
611
+ "oracle_success_rate": 0.8695652173913043,
612
+ "action_mse_to_best": 0.3996870838591586,
613
+ "per_task": {
614
+ "LiftPegUpright-v1": {
615
+ "action_mse_to_best": 0.3307521107021423,
616
+ "expert_success_rate": 0.8584070796460177,
617
+ "num_groups": 113,
618
+ "oracle_success_rate": 0.9380530973451328,
619
+ "policy_expert_regret": 0.9565544418529072,
620
+ "policy_oracle_regret": 1.0475532074945162,
621
+ "policy_rollout_progress": 0.6609668837184399,
622
+ "policy_rollout_success_rate": 0.21238938053097345,
623
+ "restore_max_error": 4.76837158203125e-07
624
+ },
625
+ "PickCube-v1": {
626
+ "action_mse_to_best": 0.32586156714545644,
627
+ "expert_success_rate": 0.9402173913043478,
628
+ "num_groups": 184,
629
+ "oracle_success_rate": 0.9456521739130435,
630
+ "policy_expert_regret": 1.4694518956492413,
631
+ "policy_oracle_regret": 1.4769888006165908,
632
+ "policy_rollout_progress": 0.43558798558534245,
633
+ "policy_rollout_success_rate": 0.010869565217391304,
634
+ "restore_max_error": 4.76837158203125e-07
635
+ },
636
+ "PullCube-v1": {
637
+ "action_mse_to_best": 0.5786957227085766,
638
+ "expert_success_rate": 0.25,
639
+ "num_groups": 76,
640
+ "oracle_success_rate": 0.40789473684210525,
641
+ "policy_expert_regret": 0.3357601676528391,
642
+ "policy_oracle_regret": 0.6036569659941291,
643
+ "policy_rollout_progress": 0.26700859815862615,
644
+ "policy_rollout_success_rate": 0.13157894736842105,
645
+ "restore_max_error": 4.76837158203125e-07
646
+ },
647
+ "PushCube-v1": {
648
+ "action_mse_to_best": 0.3886206160123284,
649
+ "expert_success_rate": 0.8198198198198198,
650
+ "num_groups": 111,
651
+ "oracle_success_rate": 1.0,
652
+ "policy_expert_regret": 0.29417850627555503,
653
+ "policy_oracle_regret": 0.34459435147745116,
654
+ "policy_rollout_progress": 0.835585828702729,
655
+ "policy_rollout_success_rate": 0.8198198198198198,
656
+ "restore_max_error": 4.76837158203125e-07
657
+ },
658
+ "StackCube-v1": {
659
+ "action_mse_to_best": 0.4985581654031853,
660
+ "expert_success_rate": 0.6923076923076923,
661
+ "num_groups": 91,
662
+ "oracle_success_rate": 0.8571428571428571,
663
+ "policy_expert_regret": 1.106395381656322,
664
+ "policy_oracle_regret": 1.2537258472088928,
665
+ "policy_rollout_progress": 0.3678028554051787,
666
+ "policy_rollout_success_rate": 0.1978021978021978,
667
+ "restore_max_error": 3.948807716369629e-07
668
+ }
669
+ }
670
+ },
671
+ {
672
+ "config": "k8_sigma0.20",
673
+ "seed": 2,
674
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05/k8_sigma0.20/seed_2/online_rollout.json",
675
+ "selection_mode": "field_optim",
676
+ "num_candidates": 8,
677
+ "candidate_sigma": 0.2,
678
+ "field_optim_steps": 4,
679
+ "field_optim_step_size": 0.05,
680
+ "field_optim_trust_radius": 0.5,
681
+ "field_optim_l2_penalty": 0.02,
682
+ "num_groups": 575,
683
+ "policy_rollout_success_rate": 0.22956521739130434,
684
+ "policy_rollout_progress": 0.5218091125312545,
685
+ "oracle_success_rate": 0.8765217391304347,
686
+ "action_mse_to_best": 0.42410068074160295,
687
+ "per_task": {
688
+ "LiftPegUpright-v1": {
689
+ "action_mse_to_best": 0.3535835361981299,
690
+ "expert_success_rate": 0.8229166666666666,
691
+ "num_groups": 96,
692
+ "oracle_success_rate": 0.9270833333333334,
693
+ "policy_expert_regret": 0.969089585977296,
694
+ "policy_oracle_regret": 1.0896811264877517,
695
+ "policy_rollout_progress": 0.6431976957246661,
696
+ "policy_rollout_success_rate": 0.1875,
697
+ "restore_max_error": 3.5762786865234375e-07
698
+ },
699
+ "PickCube-v1": {
700
+ "action_mse_to_best": 0.344063895782738,
701
+ "expert_success_rate": 0.9444444444444444,
702
+ "num_groups": 198,
703
+ "oracle_success_rate": 0.9595959595959596,
704
+ "policy_expert_regret": 1.384132306433678,
705
+ "policy_oracle_regret": 1.405459522273194,
706
+ "policy_rollout_progress": 0.4947132789071254,
707
+ "policy_rollout_success_rate": 0.04040404040404041,
708
+ "restore_max_error": 4.76837158203125e-07
709
+ },
710
+ "PullCube-v1": {
711
+ "action_mse_to_best": 0.6478720838824908,
712
+ "expert_success_rate": 0.24444444444444444,
713
+ "num_groups": 90,
714
+ "oracle_success_rate": 0.4666666666666667,
715
+ "policy_expert_regret": 0.2917847439698461,
716
+ "policy_oracle_regret": 0.5349132894486602,
717
+ "policy_rollout_progress": 0.3336249701833973,
718
+ "policy_rollout_success_rate": 0.2111111111111111,
719
+ "restore_max_error": 4.0978193283081055e-07
720
+ },
721
+ "PushCube-v1": {
722
+ "action_mse_to_best": 0.386228519279768,
723
+ "expert_success_rate": 0.8514851485148515,
724
+ "num_groups": 101,
725
+ "oracle_success_rate": 1.0,
726
+ "policy_expert_regret": 0.3565023746525887,
727
+ "policy_oracle_regret": 0.4105553013263362,
728
+ "policy_rollout_progress": 0.8072664808518816,
729
+ "policy_rollout_success_rate": 0.7821782178217822,
730
+ "restore_max_error": 4.76837158203125e-07
731
+ },
732
+ "StackCube-v1": {
733
+ "action_mse_to_best": 0.4941294732193152,
734
+ "expert_success_rate": 0.7666666666666667,
735
+ "num_groups": 90,
736
+ "oracle_success_rate": 0.9111111111111111,
737
+ "policy_expert_regret": 1.3245210293266507,
738
+ "policy_oracle_regret": 1.4733497116300796,
739
+ "policy_rollout_progress": 0.31977633121940824,
740
+ "policy_rollout_success_rate": 0.08888888888888889,
741
+ "restore_max_error": 4.76837158203125e-07
742
+ }
743
+ }
744
+ }
745
+ ]
746
+ }
results/h16_field_optim_near_miss_policy_bc5_bestpt_s4_trust05_afterany_summary.md ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # h=16 Field-Guided Rollout Sweep
2
+
3
+ Result root: `/scratch/knguy52/dovla/experiments/dovla_h16_field_optim_sweep/near_miss_policy_bc5_bestpt_s4_trust05`
4
+ Completed result files: 9
5
+ Baseline h=4 policy success: 29.67%
6
+ Baseline h=16 policy success: 29.74%
7
+
8
+ | config | seeds | mean success | gain vs h=16 | progress | action MSE |
9
+ |---|---:|---:|---:|---:|---:|
10
+ | k32_sigma0.50 | 1 | 25.39% | -4.35% | 53.65% | 0.391 |
11
+ | k1_sigma0.00 | 2 | 24.61% | -5.13% | 53.27% | 0.398 |
12
+ | k16_sigma0.35 | 3 | 24.23% | -5.51% | 52.37% | 0.406 |
13
+ | k8_sigma0.20 | 3 | 24.12% | -5.62% | 52.37% | 0.404 |
14
+
15
+ Best config:
16
+ - k32_sigma0.50
17
+ - mean success: 25.39%
18
+ - gain vs h=16 policy: -4.35%
results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_knn4_summary.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs",
3
+ "objective": "near_miss_policy_bc5",
4
+ "out_name": "retrieval_residual_knn4_rollout.json",
5
+ "num_completed": 0,
6
+ "baseline_h4_policy_success": 0.2967,
7
+ "baseline_h16_rank_checkpoint_success": 0.29739130434782607,
8
+ "mean_success": 0.0,
9
+ "std_success": 0.0,
10
+ "mean_progress": 0.0,
11
+ "mean_action_mse_to_best": 0.0,
12
+ "gain_vs_h4": 0.0,
13
+ "gain_vs_h16_rank_checkpoint": 0.0,
14
+ "rows": []
15
+ }
results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_knn4_summary.md ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # h=16 Best-Policy Checkpoint Rollout
2
+
3
+ Run root: `/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs`
4
+ Objective: `near_miss_policy_bc5`
5
+ Result file: `retrieval_residual_knn4_rollout.json`
6
+ Completed seeds: 0
7
+ Baseline h=4 policy success: 29.67%
8
+ Baseline h=16 rank-checkpoint success: 29.74%
9
+
10
+ Mean success: 0.00% +/- 0.00%
11
+ Gain vs h=16 rank checkpoint: +0.00%
12
+ Mean progress: 0.00%
13
+ Mean action MSE to best: 0.000
14
+
15
+ | seed | mode | k | retrieval K | sigma | opt steps | trust | success | progress | oracle | action MSE |
16
+ |---:|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|
results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_summary.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs",
3
+ "objective": "near_miss_policy_bc5",
4
+ "out_name": "retrieval_residual_rollout.json",
5
+ "num_completed": 0,
6
+ "baseline_h4_policy_success": 0.2967,
7
+ "baseline_h16_rank_checkpoint_success": 0.29739130434782607,
8
+ "mean_success": 0.0,
9
+ "std_success": 0.0,
10
+ "mean_progress": 0.0,
11
+ "mean_action_mse_to_best": 0.0,
12
+ "gain_vs_h4": 0.0,
13
+ "gain_vs_h16_rank_checkpoint": 0.0,
14
+ "rows": []
15
+ }
results/h16_policy_ckpt_near_miss_policy_bc5_bestpt_retrieval_residual_summary.md ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # h=16 Best-Policy Checkpoint Rollout
2
+
3
+ Run root: `/scratch/knguy52/dovla/experiments/dovla_h16_policy_ckpt_runs`
4
+ Objective: `near_miss_policy_bc5`
5
+ Result file: `retrieval_residual_rollout.json`
6
+ Completed seeds: 0
7
+ Baseline h=4 policy success: 29.67%
8
+ Baseline h=16 rank-checkpoint success: 29.74%
9
+
10
+ Mean success: 0.00% +/- 0.00%
11
+ Gain vs h=16 rank checkpoint: +0.00%
12
+ Mean progress: 0.00%
13
+ Mean action MSE to best: 0.000
14
+
15
+ | seed | mode | k | retrieval K | sigma | opt steps | trust | success | progress | oracle | action MSE |
16
+ |---:|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|