anhtld commited on
Commit
8316510
·
verified ·
1 Parent(s): f8219f8

Auto-sync: 2026-06-27 10:13:44

Browse files
results/h16_lattice_rollout_summary.json CHANGED
@@ -1,225 +1,224 @@
1
  {
2
  "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs",
3
- "out_name": "lattice_rollout.json",
4
  "num_completed": 3,
5
  "baseline_h4_policy_success": 0.2967,
6
  "baseline_h16_policy_success": 0.29739130434782607,
7
- "mean_success": 0.6933333333333334,
8
- "std_success": 0.035655708555783816,
9
- "mean_progress": 0.8108624100247803,
10
  "mean_oracle_success": 0.8678260869565217,
11
- "mean_action_mse_to_best": 0.43754452051773696,
12
- "gain_vs_h4": 0.39663333333333334,
13
- "gain_vs_h16_policy": 0.3959420289855073,
14
  "selected_candidate_type_counts": {
15
- "lattice_expert": 977,
16
- "lattice_near_miss": 348,
17
- "lattice_no_op": 177,
18
- "lattice_random_negative": 138,
19
- "lattice_wrong_direction": 30,
20
- "lattice_wrong_gripper": 55
21
  },
22
  "rows": [
23
  {
24
  "seed": 0,
25
- "path": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs/seed_0/lattice_rollout.json",
26
  "selection_mode": "lattice",
27
  "num_candidates": 16,
28
  "num_groups": 575,
29
- "policy_rollout_success_rate": 0.6521739130434783,
30
- "policy_rollout_progress": 0.792444871255885,
31
  "oracle_success_rate": 0.8573913043478261,
32
- "action_mse_to_best": 0.414993303792207,
33
  "per_task": {
34
  "LiftPegUpright-v1": {
35
- "action_mse_to_best": 0.28245431445983055,
36
  "expert_success_rate": 0.8865979381443299,
37
  "num_groups": 97,
38
  "oracle_success_rate": 0.9278350515463918,
39
- "policy_expert_regret": 0.2610538084175169,
40
- "policy_oracle_regret": 0.27425835227843415,
41
- "policy_rollout_progress": 0.9071177396270418,
42
- "policy_rollout_success_rate": 0.7216494845360825,
43
  "restore_max_error": 4.76837158203125e-07
44
  },
45
  "PickCube-v1": {
46
- "action_mse_to_best": 0.41056775828250325,
47
  "expert_success_rate": 0.9375,
48
  "num_groups": 208,
49
  "oracle_success_rate": 0.9471153846153846,
50
- "policy_expert_regret": 0.26980068733414203,
51
- "policy_oracle_regret": 0.2764672613881815,
52
- "policy_rollout_progress": 0.8699935431269785,
53
- "policy_rollout_success_rate": 0.7788461538461539,
54
  "restore_max_error": 4.76837158203125e-07
55
  },
56
  "PullCube-v1": {
57
- "action_mse_to_best": 0.5239523940465667,
58
  "expert_success_rate": 0.19480519480519481,
59
  "num_groups": 77,
60
  "oracle_success_rate": 0.36363636363636365,
61
- "policy_expert_regret": 0.15346811585999154,
62
- "policy_oracle_regret": 0.37490944806244464,
63
- "policy_rollout_progress": 0.3196387503828321,
64
  "policy_rollout_success_rate": 0.16883116883116883,
65
  "restore_max_error": 4.76837158203125e-07
66
  },
67
  "PushCube-v1": {
68
- "action_mse_to_best": 0.38503194246143896,
69
  "expert_success_rate": 0.8617021276595744,
70
  "num_groups": 94,
71
  "oracle_success_rate": 0.9893617021276596,
72
- "policy_expert_regret": 0.28232911355952,
73
- "policy_oracle_regret": 0.37923398661486646,
74
- "policy_rollout_progress": 0.8140388716091501,
75
- "policy_rollout_success_rate": 0.7872340425531915,
76
  "restore_max_error": 4.76837158203125e-07
77
  },
78
  "StackCube-v1": {
79
- "action_mse_to_best": 0.4978550557797361,
80
  "expert_success_rate": 0.6767676767676768,
81
  "num_groups": 99,
82
  "oracle_success_rate": 0.8585858585858586,
83
- "policy_expert_regret": 0.2738224623179195,
84
- "policy_oracle_regret": 0.4027370100069528,
85
- "policy_rollout_progress": 0.8643927833046576,
86
- "policy_rollout_success_rate": 0.5656565656565656,
87
  "restore_max_error": 3.948807716369629e-07
88
  }
89
  }
90
  },
91
  {
92
  "seed": 1,
93
- "path": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs/seed_1/lattice_rollout.json",
94
  "selection_mode": "lattice",
95
  "num_candidates": 16,
96
  "num_groups": 575,
97
- "policy_rollout_success_rate": 0.7147826086956521,
98
- "policy_rollout_progress": 0.8205086112848442,
99
  "oracle_success_rate": 0.8695652173913043,
100
- "action_mse_to_best": 0.41892785357189893,
101
  "per_task": {
102
  "LiftPegUpright-v1": {
103
- "action_mse_to_best": 0.31486138409473396,
104
  "expert_success_rate": 0.8584070796460177,
105
  "num_groups": 113,
106
  "oracle_success_rate": 0.9380530973451328,
107
- "policy_expert_regret": 0.2672024345503444,
108
- "policy_oracle_regret": 0.2922932648025783,
109
- "policy_rollout_progress": 0.8941378234762006,
110
- "policy_rollout_success_rate": 0.7345132743362832,
111
  "restore_max_error": 4.76837158203125e-07
112
  },
113
  "PickCube-v1": {
114
- "action_mse_to_best": 0.44310908647173125,
115
  "expert_success_rate": 0.9402173913043478,
116
  "num_groups": 184,
117
  "oracle_success_rate": 0.9456521739130435,
118
- "policy_expert_regret": 0.17421010318819596,
119
- "policy_oracle_regret": 0.1742341076631261,
120
- "policy_rollout_progress": 0.9089716651436427,
121
- "policy_rollout_success_rate": 0.8478260869565217,
122
  "restore_max_error": 4.76837158203125e-07
123
  },
124
  "PullCube-v1": {
125
- "action_mse_to_best": 0.5217870954473159,
126
  "expert_success_rate": 0.25,
127
  "num_groups": 76,
128
  "oracle_success_rate": 0.40789473684210525,
129
- "policy_expert_regret": 0.12901650541952173,
130
- "policy_oracle_regret": 0.3323121542037514,
131
- "policy_rollout_progress": 0.379051765659824,
132
- "policy_rollout_success_rate": 0.23684210526315788,
133
  "restore_max_error": 4.76837158203125e-07
134
  },
135
  "PushCube-v1": {
136
- "action_mse_to_best": 0.34412760804365344,
137
  "expert_success_rate": 0.8198198198198198,
138
  "num_groups": 111,
139
  "oracle_success_rate": 1.0,
140
- "policy_expert_regret": 0.1992848778630162,
141
- "policy_oracle_regret": 0.3170433482071301,
142
- "policy_rollout_progress": 0.8541278229640411,
143
- "policy_rollout_success_rate": 0.8288288288288288,
144
  "restore_max_error": 4.76837158203125e-07
145
  },
146
  "StackCube-v1": {
147
- "action_mse_to_best": 0.5045947664120531,
148
  "expert_success_rate": 0.6923076923076923,
149
  "num_groups": 91,
150
  "oracle_success_rate": 0.8571428571428571,
151
- "policy_expert_regret": 0.1715098601448667,
152
- "policy_oracle_regret": 0.2727490429367338,
153
- "policy_rollout_progress": 0.8778897638504322,
154
- "policy_rollout_success_rate": 0.6813186813186813,
155
  "restore_max_error": 3.948807716369629e-07
156
  }
157
  }
158
  },
159
  {
160
  "seed": 2,
161
- "path": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs/seed_2/lattice_rollout.json",
162
  "selection_mode": "lattice",
163
  "num_candidates": 16,
164
  "num_groups": 575,
165
- "policy_rollout_success_rate": 0.7130434782608696,
166
- "policy_rollout_progress": 0.8196337475336116,
167
  "oracle_success_rate": 0.8765217391304347,
168
- "action_mse_to_best": 0.47871240418910493,
169
  "per_task": {
170
  "LiftPegUpright-v1": {
171
- "action_mse_to_best": 0.33342877044075675,
172
  "expert_success_rate": 0.8229166666666666,
173
  "num_groups": 96,
174
  "oracle_success_rate": 0.9270833333333334,
175
- "policy_expert_regret": 0.22194465808570385,
176
- "policy_oracle_regret": 0.28883908068140346,
177
- "policy_rollout_progress": 0.9184654237081608,
178
- "policy_rollout_success_rate": 0.7395833333333334,
179
  "restore_max_error": 3.5762786865234375e-07
180
  },
181
  "PickCube-v1": {
182
- "action_mse_to_best": 0.4627937324570887,
183
  "expert_success_rate": 0.9444444444444444,
184
  "num_groups": 198,
185
  "oracle_success_rate": 0.9595959595959596,
186
- "policy_expert_regret": 0.10305700688199564,
187
- "policy_oracle_regret": 0.11014045741070401,
188
- "policy_rollout_progress": 0.9385366825518584,
189
- "policy_rollout_success_rate": 0.898989898989899,
190
  "restore_max_error": 4.76837158203125e-07
191
  },
192
  "PullCube-v1": {
193
- "action_mse_to_best": 0.6827966811549332,
194
  "expert_success_rate": 0.24444444444444444,
195
  "num_groups": 90,
196
  "oracle_success_rate": 0.4666666666666667,
197
- "policy_expert_regret": 0.15796189763479762,
198
- "policy_oracle_regret": 0.40946051337652734,
199
- "policy_rollout_progress": 0.39318432178762225,
200
- "policy_rollout_success_rate": 0.26666666666666666,
201
  "restore_max_error": 4.0978193283081055e-07
202
  },
203
  "PushCube-v1": {
204
- "action_mse_to_best": 0.34167051138264115,
205
  "expert_success_rate": 0.8514851485148515,
206
  "num_groups": 101,
207
  "oracle_success_rate": 1.0,
208
- "policy_expert_regret": 0.24242213811024582,
209
- "policy_oracle_regret": 0.29742342971339086,
210
- "policy_rollout_progress": 0.8609924118707676,
211
- "policy_rollout_success_rate": 0.8415841584158416,
212
  "restore_max_error": 4.76837158203125e-07
213
  },
214
  "StackCube-v1": {
215
- "action_mse_to_best": 0.6184098718480932,
216
  "expert_success_rate": 0.7666666666666667,
217
  "num_groups": 90,
218
  "oracle_success_rate": 0.9111111111111111,
219
- "policy_expert_regret": 0.3941795117325253,
220
- "policy_oracle_regret": 0.4843028916252984,
221
- "policy_rollout_progress": 0.8326626494526863,
222
- "policy_rollout_success_rate": 0.5777777777777777,
223
  "restore_max_error": 4.76837158203125e-07
224
  }
225
  }
 
1
  {
2
  "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs",
3
+ "out_name": "lattice_no_expert_rollout.json",
4
  "num_completed": 3,
5
  "baseline_h4_policy_success": 0.2967,
6
  "baseline_h16_policy_success": 0.29739130434782607,
7
+ "mean_success": 0.5698550724637681,
8
+ "std_success": 0.046188021535170105,
9
+ "mean_progress": 0.7501174603224449,
10
  "mean_oracle_success": 0.8678260869565217,
11
+ "mean_action_mse_to_best": 0.4590102900127523,
12
+ "gain_vs_h4": 0.2731550724637681,
13
+ "gain_vs_h16_policy": 0.27246376811594203,
14
  "selected_candidate_type_counts": {
15
+ "lattice_near_miss": 1263,
16
+ "lattice_no_op": 222,
17
+ "lattice_random_negative": 144,
18
+ "lattice_wrong_direction": 34,
19
+ "lattice_wrong_gripper": 62
 
20
  },
21
  "rows": [
22
  {
23
  "seed": 0,
24
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs/seed_0/lattice_no_expert_rollout.json",
25
  "selection_mode": "lattice",
26
  "num_candidates": 16,
27
  "num_groups": 575,
28
+ "policy_rollout_success_rate": 0.5165217391304348,
29
+ "policy_rollout_progress": 0.7308024795159049,
30
  "oracle_success_rate": 0.8573913043478261,
31
+ "action_mse_to_best": 0.4291479373977864,
32
  "per_task": {
33
  "LiftPegUpright-v1": {
34
+ "action_mse_to_best": 0.2888141949958115,
35
  "expert_success_rate": 0.8865979381443299,
36
  "num_groups": 97,
37
  "oracle_success_rate": 0.9278350515463918,
38
+ "policy_expert_regret": 0.45705366580142187,
39
+ "policy_oracle_regret": 0.47029784183526774,
40
+ "policy_rollout_progress": 0.8553990655645882,
41
+ "policy_rollout_success_rate": 0.5773195876288659,
42
  "restore_max_error": 4.76837158203125e-07
43
  },
44
  "PickCube-v1": {
45
+ "action_mse_to_best": 0.4340394471700375,
46
  "expert_success_rate": 0.9375,
47
  "num_groups": 208,
48
  "oracle_success_rate": 0.9471153846153846,
49
+ "policy_expert_regret": 0.6624748194769311,
50
+ "policy_oracle_regret": 0.6691413935309706,
51
+ "policy_rollout_progress": 0.7513381671697761,
52
+ "policy_rollout_success_rate": 0.5048076923076923,
53
  "restore_max_error": 4.76837158203125e-07
54
  },
55
  "PullCube-v1": {
56
+ "action_mse_to_best": 0.5183976920587676,
57
  "expert_success_rate": 0.19480519480519481,
58
  "num_groups": 77,
59
  "oracle_success_rate": 0.36363636363636365,
60
+ "policy_expert_regret": 0.1787171581058533,
61
+ "policy_oracle_regret": 0.37732240485100005,
62
+ "policy_rollout_progress": 0.3172257935942768,
63
  "policy_rollout_success_rate": 0.16883116883116883,
64
  "restore_max_error": 4.76837158203125e-07
65
  },
66
  "PushCube-v1": {
67
+ "action_mse_to_best": 0.39197504613300826,
68
  "expert_success_rate": 0.8617021276595744,
69
  "num_groups": 94,
70
  "oracle_success_rate": 0.9893617021276596,
71
+ "policy_expert_regret": 0.2812512713543912,
72
+ "policy_oracle_regret": 0.3391457902307206,
73
+ "policy_rollout_progress": 0.8417052735990667,
74
+ "policy_rollout_success_rate": 0.8191489361702128,
75
  "restore_max_error": 4.76837158203125e-07
76
  },
77
  "StackCube-v1": {
78
+ "action_mse_to_best": 0.5222485399266441,
79
  "expert_success_rate": 0.6767676767676768,
80
  "num_groups": 99,
81
  "oracle_success_rate": 0.8585858585858586,
82
+ "policy_expert_regret": 0.5384464679342328,
83
+ "policy_oracle_regret": 0.574341721606977,
84
+ "policy_rollout_progress": 0.7819464218736899,
85
+ "policy_rollout_success_rate": 0.46464646464646464,
86
  "restore_max_error": 3.948807716369629e-07
87
  }
88
  }
89
  },
90
  {
91
  "seed": 1,
92
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs/seed_1/lattice_no_expert_rollout.json",
93
  "selection_mode": "lattice",
94
  "num_candidates": 16,
95
  "num_groups": 575,
96
+ "policy_rollout_success_rate": 0.5965217391304348,
97
+ "policy_rollout_progress": 0.756582982868928,
98
  "oracle_success_rate": 0.8695652173913043,
99
+ "action_mse_to_best": 0.44271026729079693,
100
  "per_task": {
101
  "LiftPegUpright-v1": {
102
+ "action_mse_to_best": 0.3169037180301626,
103
  "expert_success_rate": 0.8584070796460177,
104
  "num_groups": 113,
105
  "oracle_success_rate": 0.9380530973451328,
106
+ "policy_expert_regret": 0.39746501211044005,
107
+ "policy_oracle_regret": 0.43380374942733124,
108
+ "policy_rollout_progress": 0.858787138245802,
109
+ "policy_rollout_success_rate": 0.6283185840707964,
110
  "restore_max_error": 4.76837158203125e-07
111
  },
112
  "PickCube-v1": {
113
+ "action_mse_to_best": 0.4862241648218554,
114
  "expert_success_rate": 0.9402173913043478,
115
  "num_groups": 184,
116
  "oracle_success_rate": 0.9456521739130435,
117
+ "policy_expert_regret": 0.4408608768540232,
118
+ "policy_oracle_regret": 0.44088488132895337,
119
+ "policy_rollout_progress": 0.8249527744381971,
120
+ "policy_rollout_success_rate": 0.657608695652174,
121
  "restore_max_error": 4.76837158203125e-07
122
  },
123
  "PullCube-v1": {
124
+ "action_mse_to_best": 0.5220103718977618,
125
  "expert_success_rate": 0.25,
126
  "num_groups": 76,
127
  "oracle_success_rate": 0.40789473684210525,
128
+ "policy_expert_regret": 0.2234261535524734,
129
+ "policy_oracle_regret": 0.44941299718706623,
130
+ "policy_rollout_progress": 0.32774039636071967,
131
+ "policy_rollout_success_rate": 0.17105263157894737,
132
  "restore_max_error": 4.76837158203125e-07
133
  },
134
  "PushCube-v1": {
135
+ "action_mse_to_best": 0.35074781935225735,
136
  "expert_success_rate": 0.8198198198198198,
137
  "num_groups": 111,
138
  "oracle_success_rate": 1.0,
139
+ "policy_expert_regret": 0.24994636481409674,
140
+ "policy_oracle_regret": 0.3667086133280316,
141
+ "policy_rollout_progress": 0.8314895848701666,
142
+ "policy_rollout_success_rate": 0.8018018018018018,
143
  "restore_max_error": 4.76837158203125e-07
144
  },
145
  "StackCube-v1": {
146
+ "action_mse_to_best": 0.5568927584093187,
147
  "expert_success_rate": 0.6923076923076923,
148
  "num_groups": 91,
149
  "oracle_success_rate": 0.8571428571428571,
150
+ "policy_expert_regret": 0.47163032846791403,
151
+ "policy_oracle_regret": 0.5352666910205569,
152
+ "policy_rollout_progress": 0.7582125712882032,
153
+ "policy_rollout_success_rate": 0.5384615384615384,
154
  "restore_max_error": 3.948807716369629e-07
155
  }
156
  }
157
  },
158
  {
159
  "seed": 2,
160
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs/seed_2/lattice_no_expert_rollout.json",
161
  "selection_mode": "lattice",
162
  "num_candidates": 16,
163
  "num_groups": 575,
164
+ "policy_rollout_success_rate": 0.5965217391304348,
165
+ "policy_rollout_progress": 0.7629669185825017,
166
  "oracle_success_rate": 0.8765217391304347,
167
+ "action_mse_to_best": 0.5051726653496735,
168
  "per_task": {
169
  "LiftPegUpright-v1": {
170
+ "action_mse_to_best": 0.3387383111627666,
171
  "expert_success_rate": 0.8229166666666666,
172
  "num_groups": 96,
173
  "oracle_success_rate": 0.9270833333333334,
174
+ "policy_expert_regret": 0.349092789615194,
175
+ "policy_oracle_regret": 0.38970588706433773,
176
+ "policy_rollout_progress": 0.8847468371192614,
177
+ "policy_rollout_success_rate": 0.6458333333333334,
178
  "restore_max_error": 3.5762786865234375e-07
179
  },
180
  "PickCube-v1": {
181
+ "action_mse_to_best": 0.5211711752068515,
182
  "expert_success_rate": 0.9444444444444444,
183
  "num_groups": 198,
184
  "oracle_success_rate": 0.9595959595959596,
185
+ "policy_expert_regret": 0.38791667779136185,
186
+ "policy_oracle_regret": 0.38796894105546403,
187
+ "policy_rollout_progress": 0.8505876985086938,
188
+ "policy_rollout_success_rate": 0.702020202020202,
189
  "restore_max_error": 4.76837158203125e-07
190
  },
191
  "PullCube-v1": {
192
+ "action_mse_to_best": 0.682747141416702,
193
  "expert_success_rate": 0.24444444444444444,
194
  "num_groups": 90,
195
  "oracle_success_rate": 0.4666666666666667,
196
+ "policy_expert_regret": 0.2159867504404651,
197
+ "policy_oracle_regret": 0.4278881938921081,
198
+ "policy_rollout_progress": 0.38586775238315263,
199
+ "policy_rollout_success_rate": 0.25555555555555554,
200
  "restore_max_error": 4.0978193283081055e-07
201
  },
202
  "PushCube-v1": {
203
+ "action_mse_to_best": 0.3426748497024326,
204
  "expert_success_rate": 0.8514851485148515,
205
  "num_groups": 101,
206
  "oracle_success_rate": 1.0,
207
+ "policy_expert_regret": 0.33127955694009764,
208
+ "policy_oracle_regret": 0.40476624254543003,
209
+ "policy_rollout_progress": 0.8130555396327878,
210
+ "policy_rollout_success_rate": 0.7821782178217822,
211
  "restore_max_error": 4.76837158203125e-07
212
  },
213
  "StackCube-v1": {
214
+ "action_mse_to_best": 0.6522901051781244,
215
  "expert_success_rate": 0.7666666666666667,
216
  "num_groups": 90,
217
  "oracle_success_rate": 0.9111111111111111,
218
+ "policy_expert_regret": 0.6501921941836675,
219
+ "policy_oracle_regret": 0.6890770451890098,
220
+ "policy_rollout_progress": 0.7611912255485852,
221
+ "policy_rollout_success_rate": 0.4444444444444444,
222
  "restore_max_error": 4.76837158203125e-07
223
  }
224
  }
results/h16_lattice_rollout_summary.md CHANGED
@@ -5,21 +5,20 @@ Completed seeds: 3
5
  Baseline h=4 policy success: 29.67%
6
  Baseline h=16 policy success: 29.74%
7
 
8
- Mean success: 69.33% +/- 3.57%
9
- Gain vs h=16 policy: +39.59%
10
  Mean oracle success: 86.78%
11
- Mean progress: 81.09%
12
 
13
  | seed | success | progress | oracle | candidates | action MSE |
14
  |---:|---:|---:|---:|---:|---:|
15
- | 0 | 65.22% | 79.24% | 85.74% | 16 | 0.415 |
16
- | 1 | 71.48% | 82.05% | 86.96% | 16 | 0.419 |
17
- | 2 | 71.30% | 81.96% | 87.65% | 16 | 0.479 |
18
 
19
  Selected candidate types:
20
- - lattice_expert: 977
21
- - lattice_near_miss: 348
22
- - lattice_no_op: 177
23
- - lattice_random_negative: 138
24
- - lattice_wrong_direction: 30
25
- - lattice_wrong_gripper: 55
 
5
  Baseline h=4 policy success: 29.67%
6
  Baseline h=16 policy success: 29.74%
7
 
8
+ Mean success: 56.99% +/- 4.62%
9
+ Gain vs h=16 policy: +27.25%
10
  Mean oracle success: 86.78%
11
+ Mean progress: 75.01%
12
 
13
  | seed | success | progress | oracle | candidates | action MSE |
14
  |---:|---:|---:|---:|---:|---:|
15
+ | 0 | 51.65% | 73.08% | 85.74% | 16 | 0.429 |
16
+ | 1 | 59.65% | 75.66% | 86.96% | 16 | 0.443 |
17
+ | 2 | 59.65% | 76.30% | 87.65% | 16 | 0.505 |
18
 
19
  Selected candidate types:
20
+ - lattice_near_miss: 1263
21
+ - lattice_no_op: 222
22
+ - lattice_random_negative: 144
23
+ - lattice_wrong_direction: 34
24
+ - lattice_wrong_gripper: 62