anhtld commited on
Commit
e05d6e6
·
verified ·
1 Parent(s): 14a6fff

Auto-sync: 2026-06-27 10:32:18

Browse files
results/h16_lattice_no_near_miss_no_expert_summary.json ADDED
@@ -0,0 +1,227 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs",
3
+ "out_name": "lattice_no_near_miss_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_near_miss_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_near_miss_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_near_miss_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
+ }
225
+ }
226
+ ]
227
+ }
results/h16_lattice_no_near_miss_no_expert_summary.md ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # h=16 Lattice-Selected Rollout
2
+
3
+ Run root: `/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs`
4
+ Completed seeds: 3
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
results/h16_retrieval_lattice_no_expert_summary.json ADDED
@@ -0,0 +1,227 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs",
3
+ "out_name": "retrieval_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": 0.271304347826087,
8
+ "std_success": 0.018157054798105306,
9
+ "mean_progress": 0.5005514273950088,
10
+ "mean_oracle_success": 0.8678260869565217,
11
+ "mean_action_mse_to_best": 0.5836967330339594,
12
+ "gain_vs_h4": -0.02539565217391304,
13
+ "gain_vs_h16_policy": -0.02608695652173909,
14
+ "selected_candidate_type_counts": {
15
+ "lattice_near_miss": 1276,
16
+ "lattice_no_op": 212,
17
+ "lattice_random_negative": 143,
18
+ "lattice_wrong_direction": 27,
19
+ "lattice_wrong_gripper": 67
20
+ },
21
+ "rows": [
22
+ {
23
+ "seed": 0,
24
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs/seed_0/retrieval_lattice_no_expert_rollout.json",
25
+ "selection_mode": "retrieval_lattice",
26
+ "num_candidates": 16,
27
+ "num_groups": 575,
28
+ "policy_rollout_success_rate": 0.25043478260869567,
29
+ "policy_rollout_progress": 0.49867984499296414,
30
+ "oracle_success_rate": 0.8573913043478261,
31
+ "action_mse_to_best": 0.5693303860076095,
32
+ "per_task": {
33
+ "LiftPegUpright-v1": {
34
+ "action_mse_to_best": 0.35727074150879357,
35
+ "expert_success_rate": 0.8865979381443299,
36
+ "num_groups": 97,
37
+ "oracle_success_rate": 0.9278350515463918,
38
+ "policy_expert_regret": 1.1045561598748277,
39
+ "policy_oracle_regret": 1.1440839441781192,
40
+ "policy_rollout_progress": 0.5836748188918399,
41
+ "policy_rollout_success_rate": 0.17525773195876287,
42
+ "restore_max_error": 4.76837158203125e-07
43
+ },
44
+ "PickCube-v1": {
45
+ "action_mse_to_best": 0.4945910416650944,
46
+ "expert_success_rate": 0.9375,
47
+ "num_groups": 208,
48
+ "oracle_success_rate": 0.9471153846153846,
49
+ "policy_expert_regret": 1.1998865354958421,
50
+ "policy_oracle_regret": 1.2065531095498814,
51
+ "policy_rollout_progress": 0.5120033742277883,
52
+ "policy_rollout_success_rate": 0.20673076923076922,
53
+ "restore_max_error": 4.76837158203125e-07
54
+ },
55
+ "PullCube-v1": {
56
+ "action_mse_to_best": 0.8415415789206306,
57
+ "expert_success_rate": 0.19480519480519481,
58
+ "num_groups": 77,
59
+ "oracle_success_rate": 0.36363636363636365,
60
+ "policy_expert_regret": 0.35076726408215697,
61
+ "policy_oracle_regret": 0.5829326225589816,
62
+ "policy_rollout_progress": 0.23609928142085865,
63
+ "policy_rollout_success_rate": 0.09090909090909091,
64
+ "restore_max_error": 4.76837158203125e-07
65
+ },
66
+ "PushCube-v1": {
67
+ "action_mse_to_best": 0.5355885234919,
68
+ "expert_success_rate": 0.8617021276595744,
69
+ "num_groups": 94,
70
+ "oracle_success_rate": 0.9893617021276596,
71
+ "policy_expert_regret": 0.5391762771822036,
72
+ "policy_oracle_regret": 0.6366645524159391,
73
+ "policy_rollout_progress": 0.6841333678111117,
74
+ "policy_rollout_success_rate": 0.6595744680851063,
75
+ "restore_max_error": 4.76837158203125e-07
76
+ },
77
+ "StackCube-v1": {
78
+ "action_mse_to_best": 0.7544520260258154,
79
+ "expert_success_rate": 0.6767676767676768,
80
+ "num_groups": 99,
81
+ "oracle_success_rate": 0.8585858585858586,
82
+ "policy_expert_regret": 1.0733418645280781,
83
+ "policy_oracle_regret": 1.2771704949513831,
84
+ "policy_rollout_progress": 0.41555121571126613,
85
+ "policy_rollout_success_rate": 0.15151515151515152,
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/retrieval_lattice_no_expert_rollout.json",
93
+ "selection_mode": "retrieval_lattice",
94
+ "num_candidates": 16,
95
+ "num_groups": 575,
96
+ "policy_rollout_success_rate": 0.28,
97
+ "policy_rollout_progress": 0.49805013884861105,
98
+ "oracle_success_rate": 0.8695652173913043,
99
+ "action_mse_to_best": 0.5685278348916251,
100
+ "per_task": {
101
+ "LiftPegUpright-v1": {
102
+ "action_mse_to_best": 0.3830389907435242,
103
+ "expert_success_rate": 0.8584070796460177,
104
+ "num_groups": 113,
105
+ "oracle_success_rate": 0.9380530973451328,
106
+ "policy_expert_regret": 0.9959914846494134,
107
+ "policy_oracle_regret": 1.0887631516804737,
108
+ "policy_rollout_progress": 0.6071000562041207,
109
+ "policy_rollout_success_rate": 0.24778761061946902,
110
+ "restore_max_error": 4.76837158203125e-07
111
+ },
112
+ "PickCube-v1": {
113
+ "action_mse_to_best": 0.5331694724888582,
114
+ "expert_success_rate": 0.9402173913043478,
115
+ "num_groups": 184,
116
+ "oracle_success_rate": 0.9456521739130435,
117
+ "policy_expert_regret": 1.203628732283782,
118
+ "policy_oracle_regret": 1.2111656372511315,
119
+ "policy_rollout_progress": 0.5057589750377582,
120
+ "policy_rollout_success_rate": 0.20652173913043478,
121
+ "restore_max_error": 4.76837158203125e-07
122
+ },
123
+ "PullCube-v1": {
124
+ "action_mse_to_best": 0.8114424096665492,
125
+ "expert_success_rate": 0.25,
126
+ "num_groups": 76,
127
+ "oracle_success_rate": 0.40789473684210525,
128
+ "policy_expert_regret": 0.3376943706940725,
129
+ "policy_oracle_regret": 0.577079668853242,
130
+ "policy_rollout_progress": 0.2529444615931042,
131
+ "policy_rollout_success_rate": 0.11842105263157894,
132
+ "restore_max_error": 4.76837158203125e-07
133
+ },
134
+ "PushCube-v1": {
135
+ "action_mse_to_best": 0.4605464878248739,
136
+ "expert_success_rate": 0.8198198198198198,
137
+ "num_groups": 111,
138
+ "oracle_success_rate": 1.0,
139
+ "policy_expert_regret": 0.5753093846477904,
140
+ "policy_oracle_regret": 0.6765443254430015,
141
+ "policy_rollout_progress": 0.6748070259083498,
142
+ "policy_rollout_success_rate": 0.6486486486486487,
143
+ "restore_max_error": 4.76837158203125e-07
144
+ },
145
+ "StackCube-v1": {
146
+ "action_mse_to_best": 0.7991937679944785,
147
+ "expert_success_rate": 0.6923076923076923,
148
+ "num_groups": 91,
149
+ "oracle_success_rate": 0.8571428571428571,
150
+ "policy_expert_regret": 1.1693862085486506,
151
+ "policy_oracle_regret": 1.3293362507125834,
152
+ "policy_rollout_progress": 0.3361484958575322,
153
+ "policy_rollout_success_rate": 0.15384615384615385,
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/retrieval_lattice_no_expert_rollout.json",
161
+ "selection_mode": "retrieval_lattice",
162
+ "num_candidates": 16,
163
+ "num_groups": 575,
164
+ "policy_rollout_success_rate": 0.28347826086956524,
165
+ "policy_rollout_progress": 0.5049242983434511,
166
+ "oracle_success_rate": 0.8765217391304347,
167
+ "action_mse_to_best": 0.6132319782026436,
168
+ "per_task": {
169
+ "LiftPegUpright-v1": {
170
+ "action_mse_to_best": 0.3866134468504849,
171
+ "expert_success_rate": 0.8229166666666666,
172
+ "num_groups": 96,
173
+ "oracle_success_rate": 0.9270833333333334,
174
+ "policy_expert_regret": 0.9153858826806148,
175
+ "policy_oracle_regret": 1.0359393774221342,
176
+ "policy_rollout_progress": 0.6372524791707596,
177
+ "policy_rollout_success_rate": 0.2604166666666667,
178
+ "restore_max_error": 3.5762786865234375e-07
179
+ },
180
+ "PickCube-v1": {
181
+ "action_mse_to_best": 0.5512155196582428,
182
+ "expert_success_rate": 0.9444444444444444,
183
+ "num_groups": 198,
184
+ "oracle_success_rate": 0.9595959595959596,
185
+ "policy_expert_regret": 1.1322207739341514,
186
+ "policy_oracle_regret": 1.13930422446286,
187
+ "policy_rollout_progress": 0.5335958494447323,
188
+ "policy_rollout_success_rate": 0.2676767676767677,
189
+ "restore_max_error": 4.76837158203125e-07
190
+ },
191
+ "PullCube-v1": {
192
+ "action_mse_to_best": 0.8858442465671235,
193
+ "expert_success_rate": 0.24444444444444444,
194
+ "num_groups": 90,
195
+ "oracle_success_rate": 0.4666666666666667,
196
+ "policy_expert_regret": 0.30565778343006966,
197
+ "policy_oracle_regret": 0.5796548407421344,
198
+ "policy_rollout_progress": 0.3154497137189739,
199
+ "policy_rollout_success_rate": 0.17777777777777778,
200
+ "restore_max_error": 4.0978193283081055e-07
201
+ },
202
+ "PushCube-v1": {
203
+ "action_mse_to_best": 0.46022101115472247,
204
+ "expert_success_rate": 0.8514851485148515,
205
+ "num_groups": 101,
206
+ "oracle_success_rate": 1.0,
207
+ "policy_expert_regret": 0.6615248527562264,
208
+ "policy_oracle_regret": 0.7336106011182955,
209
+ "policy_rollout_progress": 0.6525280127430907,
210
+ "policy_rollout_success_rate": 0.6138613861386139,
211
+ "restore_max_error": 4.76837158203125e-07
212
+ },
213
+ "StackCube-v1": {
214
+ "action_mse_to_best": 0.8904946595430374,
215
+ "expert_success_rate": 0.7666666666666667,
216
+ "num_groups": 90,
217
+ "oracle_success_rate": 0.9111111111111111,
218
+ "policy_expert_regret": 1.3177169235216246,
219
+ "policy_oracle_regret": 1.4797099113464356,
220
+ "policy_rollout_progress": 0.32452724261416327,
221
+ "policy_rollout_success_rate": 0.07777777777777778,
222
+ "restore_max_error": 4.76837158203125e-07
223
+ }
224
+ }
225
+ }
226
+ ]
227
+ }
results/h16_retrieval_lattice_no_expert_summary.md ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # h=16 Lattice-Selected Rollout
2
+
3
+ Run root: `/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs`
4
+ Completed seeds: 3
5
+ Baseline h=4 policy success: 29.67%
6
+ Baseline h=16 policy success: 29.74%
7
+
8
+ Mean success: 27.13% +/- 1.82%
9
+ Gain vs h=16 policy: -2.61%
10
+ Mean oracle success: 86.78%
11
+ Mean progress: 50.06%
12
+
13
+ | seed | success | progress | oracle | candidates | action MSE |
14
+ |---:|---:|---:|---:|---:|---:|
15
+ | 0 | 25.04% | 49.87% | 85.74% | 16 | 0.569 |
16
+ | 1 | 28.00% | 49.81% | 86.96% | 16 | 0.569 |
17
+ | 2 | 28.35% | 50.49% | 87.65% | 16 | 0.613 |
18
+
19
+ Selected candidate types:
20
+ - lattice_near_miss: 1276
21
+ - lattice_no_op: 212
22
+ - lattice_random_negative: 143
23
+ - lattice_wrong_direction: 27
24
+ - lattice_wrong_gripper: 67
results/h16_retrieval_lattice_summary.json ADDED
@@ -0,0 +1,228 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "run_root": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs",
3
+ "out_name": "retrieval_lattice_rollout.json",
4
+ "num_completed": 3,
5
+ "baseline_h4_policy_success": 0.2967,
6
+ "baseline_h16_policy_success": 0.29739130434782607,
7
+ "mean_success": 0.28927536231884055,
8
+ "std_success": 0.007842173483170244,
9
+ "mean_progress": 0.5070036883118725,
10
+ "mean_oracle_success": 0.8678260869565217,
11
+ "mean_action_mse_to_best": 0.5642791039512857,
12
+ "gain_vs_h4": -0.007424637681159474,
13
+ "gain_vs_h16_policy": -0.008115942028985523,
14
+ "selected_candidate_type_counts": {
15
+ "lattice_expert": 970,
16
+ "lattice_near_miss": 361,
17
+ "lattice_no_op": 172,
18
+ "lattice_random_negative": 133,
19
+ "lattice_wrong_direction": 25,
20
+ "lattice_wrong_gripper": 64
21
+ },
22
+ "rows": [
23
+ {
24
+ "seed": 0,
25
+ "path": "/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs/seed_0/retrieval_lattice_rollout.json",
26
+ "selection_mode": "retrieval_lattice",
27
+ "num_candidates": 16,
28
+ "num_groups": 575,
29
+ "policy_rollout_success_rate": 0.2817391304347826,
30
+ "policy_rollout_progress": 0.5097964013108741,
31
+ "oracle_success_rate": 0.8573913043478261,
32
+ "action_mse_to_best": 0.5467841655761003,
33
+ "per_task": {
34
+ "LiftPegUpright-v1": {
35
+ "action_mse_to_best": 0.35563730830614715,
36
+ "expert_success_rate": 0.8865979381443299,
37
+ "num_groups": 97,
38
+ "oracle_success_rate": 0.9278350515463918,
39
+ "policy_expert_regret": 1.0326172715609836,
40
+ "policy_oracle_regret": 1.0722317271625872,
41
+ "policy_rollout_progress": 0.6039806441547945,
42
+ "policy_rollout_success_rate": 0.2268041237113402,
43
+ "restore_max_error": 4.76837158203125e-07
44
+ },
45
+ "PickCube-v1": {
46
+ "action_mse_to_best": 0.4484794621821493,
47
+ "expert_success_rate": 0.9375,
48
+ "num_groups": 208,
49
+ "oracle_success_rate": 0.9471153846153846,
50
+ "policy_expert_regret": 1.1537728436658374,
51
+ "policy_oracle_regret": 1.1537940783936602,
52
+ "policy_rollout_progress": 0.5311086406134284,
53
+ "policy_rollout_success_rate": 0.2403846153846154,
54
+ "restore_max_error": 4.76837158203125e-07
55
+ },
56
+ "PullCube-v1": {
57
+ "action_mse_to_best": 0.84151822512413,
58
+ "expert_success_rate": 0.19480519480519481,
59
+ "num_groups": 77,
60
+ "oracle_success_rate": 0.36363636363636365,
61
+ "policy_expert_regret": 0.3055240137067127,
62
+ "policy_oracle_regret": 0.5390667356671645,
63
+ "policy_rollout_progress": 0.244317766750691,
64
+ "policy_rollout_success_rate": 0.1038961038961039,
65
+ "restore_max_error": 4.76837158203125e-07
66
+ },
67
+ "PushCube-v1": {
68
+ "action_mse_to_best": 0.5283121121056537,
69
+ "expert_success_rate": 0.8617021276595744,
70
+ "num_groups": 94,
71
+ "oracle_success_rate": 0.9893617021276596,
72
+ "policy_expert_regret": 0.5605980893398853,
73
+ "policy_oracle_regret": 0.6775499712279502,
74
+ "policy_rollout_progress": 0.6735138585592838,
75
+ "policy_rollout_success_rate": 0.648936170212766,
76
+ "restore_max_error": 4.76837158203125e-07
77
+ },
78
+ "StackCube-v1": {
79
+ "action_mse_to_best": 0.7289101645877265,
80
+ "expert_success_rate": 0.6767676767676768,
81
+ "num_groups": 99,
82
+ "oracle_success_rate": 0.8585858585858586,
83
+ "policy_expert_regret": 1.0163410561855393,
84
+ "policy_oracle_regret": 1.2083435260286235,
85
+ "policy_rollout_progress": 0.42377212402796505,
86
+ "policy_rollout_success_rate": 0.21212121212121213,
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/retrieval_lattice_rollout.json",
94
+ "selection_mode": "retrieval_lattice",
95
+ "num_candidates": 16,
96
+ "num_groups": 575,
97
+ "policy_rollout_success_rate": 0.288695652173913,
98
+ "policy_rollout_progress": 0.500746654110414,
99
+ "oracle_success_rate": 0.8695652173913043,
100
+ "action_mse_to_best": 0.5489565526718355,
101
+ "per_task": {
102
+ "LiftPegUpright-v1": {
103
+ "action_mse_to_best": 0.3817376020133693,
104
+ "expert_success_rate": 0.8584070796460177,
105
+ "num_groups": 113,
106
+ "oracle_success_rate": 0.9380530973451328,
107
+ "policy_expert_regret": 0.9911989344168554,
108
+ "policy_oracle_regret": 1.0726828230957015,
109
+ "policy_rollout_progress": 0.614330827266769,
110
+ "policy_rollout_success_rate": 0.25663716814159293,
111
+ "restore_max_error": 4.76837158203125e-07
112
+ },
113
+ "PickCube-v1": {
114
+ "action_mse_to_best": 0.4809453334942784,
115
+ "expert_success_rate": 0.9402173913043478,
116
+ "num_groups": 184,
117
+ "oracle_success_rate": 0.9456521739130435,
118
+ "policy_expert_regret": 1.1634100335979145,
119
+ "policy_oracle_regret": 1.170946938565264,
120
+ "policy_rollout_progress": 0.5209976178264933,
121
+ "policy_rollout_success_rate": 0.2391304347826087,
122
+ "restore_max_error": 4.76837158203125e-07
123
+ },
124
+ "PullCube-v1": {
125
+ "action_mse_to_best": 0.8191917944258373,
126
+ "expert_success_rate": 0.25,
127
+ "num_groups": 76,
128
+ "oracle_success_rate": 0.40789473684210525,
129
+ "policy_expert_regret": 0.3627372664119129,
130
+ "policy_oracle_regret": 0.6016990902357257,
131
+ "policy_rollout_progress": 0.2507582266507119,
132
+ "policy_rollout_success_rate": 0.11842105263157894,
133
+ "restore_max_error": 4.76837158203125e-07
134
+ },
135
+ "PushCube-v1": {
136
+ "action_mse_to_best": 0.4591157559204746,
137
+ "expert_success_rate": 0.8198198198198198,
138
+ "num_groups": 111,
139
+ "oracle_success_rate": 1.0,
140
+ "policy_expert_regret": 0.5930133084724615,
141
+ "policy_oracle_regret": 0.7112691351422319,
142
+ "policy_rollout_progress": 0.6581002342271375,
143
+ "policy_rollout_success_rate": 0.6306306306306306,
144
+ "restore_max_error": 4.76837158203125e-07
145
+ },
146
+ "StackCube-v1": {
147
+ "action_mse_to_best": 0.7780143089264959,
148
+ "expert_success_rate": 0.6923076923076923,
149
+ "num_groups": 91,
150
+ "oracle_success_rate": 0.8571428571428571,
151
+ "policy_expert_regret": 1.1573464636291777,
152
+ "policy_oracle_regret": 1.3298844198931705,
153
+ "policy_rollout_progress": 0.3356003266769451,
154
+ "policy_rollout_success_rate": 0.15384615384615385,
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/retrieval_lattice_rollout.json",
162
+ "selection_mode": "retrieval_lattice",
163
+ "num_candidates": 16,
164
+ "num_groups": 575,
165
+ "policy_rollout_success_rate": 0.29739130434782607,
166
+ "policy_rollout_progress": 0.5104680095143292,
167
+ "oracle_success_rate": 0.8765217391304347,
168
+ "action_mse_to_best": 0.5970965936059213,
169
+ "per_task": {
170
+ "LiftPegUpright-v1": {
171
+ "action_mse_to_best": 0.3823804637795547,
172
+ "expert_success_rate": 0.8229166666666666,
173
+ "num_groups": 96,
174
+ "oracle_success_rate": 0.9270833333333334,
175
+ "policy_expert_regret": 0.8790868987950186,
176
+ "policy_oracle_regret": 0.99962346519654,
177
+ "policy_rollout_progress": 0.6527350580630203,
178
+ "policy_rollout_success_rate": 0.28125,
179
+ "restore_max_error": 3.5762786865234375e-07
180
+ },
181
+ "PickCube-v1": {
182
+ "action_mse_to_best": 0.5080512181639371,
183
+ "expert_success_rate": 0.9444444444444444,
184
+ "num_groups": 198,
185
+ "oracle_success_rate": 0.9595959595959596,
186
+ "policy_expert_regret": 1.084108995668816,
187
+ "policy_oracle_regret": 1.0911924461975242,
188
+ "policy_rollout_progress": 0.5463541408229355,
189
+ "policy_rollout_success_rate": 0.30303030303030304,
190
+ "restore_max_error": 4.76837158203125e-07
191
+ },
192
+ "PullCube-v1": {
193
+ "action_mse_to_best": 0.8852970499338375,
194
+ "expert_success_rate": 0.24444444444444444,
195
+ "num_groups": 90,
196
+ "oracle_success_rate": 0.4666666666666667,
197
+ "policy_expert_regret": 0.3057925239722762,
198
+ "policy_oracle_regret": 0.5990210206247866,
199
+ "policy_rollout_progress": 0.307098902337667,
200
+ "policy_rollout_success_rate": 0.16666666666666666,
201
+ "restore_max_error": 4.0978193283081055e-07
202
+ },
203
+ "PushCube-v1": {
204
+ "action_mse_to_best": 0.45949358190640366,
205
+ "expert_success_rate": 0.8514851485148515,
206
+ "num_groups": 101,
207
+ "oracle_success_rate": 1.0,
208
+ "policy_expert_regret": 0.6463685169963553,
209
+ "policy_oracle_regret": 0.7728047379762819,
210
+ "policy_rollout_progress": 0.633135856083124,
211
+ "policy_rollout_success_rate": 0.594059405940594,
212
+ "restore_max_error": 4.76837158203125e-07
213
+ },
214
+ "StackCube-v1": {
215
+ "action_mse_to_best": 0.8882476593057315,
216
+ "expert_success_rate": 0.7666666666666667,
217
+ "num_groups": 90,
218
+ "oracle_success_rate": 0.9111111111111111,
219
+ "policy_expert_regret": 1.287395055923197,
220
+ "policy_oracle_regret": 1.4365394057499037,
221
+ "policy_rollout_progress": 0.3454755259884728,
222
+ "policy_rollout_success_rate": 0.1,
223
+ "restore_max_error": 4.76837158203125e-07
224
+ }
225
+ }
226
+ }
227
+ ]
228
+ }
results/h16_retrieval_lattice_summary.md ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # h=16 Lattice-Selected Rollout
2
+
3
+ Run root: `/scratch/knguy52/dovla/experiments/dovla_h16_rollout_runs`
4
+ Completed seeds: 3
5
+ Baseline h=4 policy success: 29.67%
6
+ Baseline h=16 policy success: 29.74%
7
+
8
+ Mean success: 28.93% +/- 0.78%
9
+ Gain vs h=16 policy: -0.81%
10
+ Mean oracle success: 86.78%
11
+ Mean progress: 50.70%
12
+
13
+ | seed | success | progress | oracle | candidates | action MSE |
14
+ |---:|---:|---:|---:|---:|---:|
15
+ | 0 | 28.17% | 50.98% | 85.74% | 16 | 0.547 |
16
+ | 1 | 28.87% | 50.07% | 86.96% | 16 | 0.549 |
17
+ | 2 | 29.74% | 51.05% | 87.65% | 16 | 0.597 |
18
+
19
+ Selected candidate types:
20
+ - lattice_expert: 970
21
+ - lattice_near_miss: 361
22
+ - lattice_no_op: 172
23
+ - lattice_random_negative: 133
24
+ - lattice_wrong_direction: 25
25
+ - lattice_wrong_gripper: 64