bruno/homemaker-layout
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
homemaker-layout/examples/maple-court/coldstart-500000-s1.log

148 lines
9 KiB
Text
Raw Normal View History

coldstart harbor-house seed 0 @ 500000: 33 fails (8h/25s) Cold-start re-baseline after the DESIGN.md 38.10/38.11 objective change. Single worker (avoids homemaker-py-b8g), scored by the shipped scorer from the programme directory. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MJ84Feep79Hhm3E4zZJmnB
2026-08-31 09:46:46 +01:00
WARNING: 338 m2 per storey needs ~70 m of daylit wall; the plot's non-private perimeter gives 55 m. Roughly 22 m2 of courtyard closes the gap. (DESIGN.md §38.3)
seed : /home/bruno/src/homemaker-layout/examples/maple-court/init.dom
programme : maple-court
budget : 500000
pop : 16
child_budget : 80
workers : 1
rng seed : 1
leaf sharing : True (factor=3)
superpose : False
multi_use : False
conn grade : False
use tiers : False
bridge circulation : False
ruin recreate : False
collapse in-search : True
shapecurve warmstart : False
shapecurve prune : False
output : /home/bruno/src/homemaker-layout/examples/maple-court/coldstart-500000-s1.dom
[ 80 evals] best 7.48546e-46 (fails 141) via construct/0
[ 160 evals] best 4.82423e-42 (fails 127) via construct/1
[ 480 evals] best 8.67067e-42 (fails 124) via construct/5
[ 800 evals] best 4.80398e-39 (fails 122) via construct/9
[ 1440 evals] best 6.36848e-40 (fails 118) via place_missing r -> rrrr
[ 2160 evals] best 1.32118e-39 (fails 117) via level_fix noop
[ 2400 evals] best 1.60106e-38 (fails 114) via place_missing em -> rlll
[ 3040 evals] best 3.2838e-38 (fails 113) via rotate 0/rrll
[ 3760 evals] best 3.65129e-38 (fails 113) via core_undivide noop
[ 4320 evals] best 5.79431e-37 (fails 109) via place_missing r -> rrrl
[ 4400 evals] best 5.8784e-37 (fails 109) via core_undivide noop
[ 4640 evals] best 2.78722e-36 (fails 107) via retype 1/rlllr->py
[ 5760 evals] best 4.90785e-34 (fails 100) via crossover lrrl<->rrrl
[ 8480 evals] best 5.06412e-34 (fails 100) via level_retype 2/rrlr<->0/rrrl
[ 8960 evals] best 1.00937e-33 (fails 99) via retype 1/rrlr->O
[ 9840 evals] best 2.11427e-33 (fails 98) via swap 0/lrr
[ 10480 evals] best 2.17495e-33 (fails 98) via swap 0/rlrl
[ 11040 evals] best 2.25388e-33 (fails 98) via core_divide noop
[ 12480 evals] best 4.78668e-33 (fails 98) via undivide 2/llr
[ 12720 evals] best 4.80876e-33 (fails 98) via core_divide noop
[ 13760 evals] best 9.1709e-33 (fails 96) via divide 2/llrr
[ 14560 evals] best 1.84586e-32 (fails 95) via core_undivide noop
[ 16880 evals] best 1.85284e-32 (fails 95) via crossover llrlr<->rlllr
[ 19840 evals] best 7.42595e-32 (fails 93) via core_undivide noop
[ 20800 evals] best 1.48504e-31 (fails 92) via core_undivide noop
[ 23680 evals] best 1.48739e-31 (fails 92) via core_undivide noop
[ 24560 evals] best 1.55172e-31 (fails 92) via core_divide noop
[ 25680 evals] best 1.56264e-31 (fails 92) via core_undivide noop
[ 26080 evals] best 6.36171e-31 (fails 90) via swap 2/rrr
[ 27040 evals] best 6.41788e-31 (fails 90) via level_retype 0/rrrl<->2/rllr
[ 28240 evals] best 1.02644e-30 (fails 89) via crossover lrl<->lrr
[ 29200 evals] best 1.04874e-30 (fails 89) via crossover llrrr<->llrrr
[ 30320 evals] best 3.66451e-30 (fails 88) via level_compound_fix k1: lvl2 → lvl0/rllll + insert lr1 into lrlr/lvl0
[ 30960 evals] best 1.09242e-29 (fails 86) via crossover rrlrr<->llrrl
[ 33120 evals] best 3.5356e-29 (fails 84) via crossover llrll<->lrll
[ 36160 evals] best 3.56494e-29 (fails 84) via level_retype 0/lrrr<->2/rllr
[ 40320 evals] best 4.93101e-29 (fails 84) via crossover lr<->lr
[ 40960 evals] best 4.96131e-29 (fails 84) via crossover rlrrr<->rlrrr
[ 42240 evals] best 8.67773e-29 (fails 83) via retype 2/rrrrr->t
[ 43920 evals] best 1.52112e-28 (fails 81) via crossover lllr<->rlrr
[ 46480 evals] best 1.52992e-28 (fails 81) via crossover lrr<->lrr
[ 48720 evals] best 3.06138e-28 (fails 80) via level_retype 2/lllr<->0/rlrlr
[ 51120 evals] best 3.07068e-28 (fails 80) via swap 2/llrr
[ 53440 evals] best 3.07434e-28 (fails 80) via swap 0/rlrr
[ 54560 evals] best 3.44534e-28 (fails 80) via crossover lrrl<->lllrl
[ 57600 evals] best 3.45695e-28 (fails 80) via core_divide noop
[ 58240 evals] best 3.46141e-28 (fails 80) via core_undivide noop
[ 61440 evals] best 3.46205e-28 (fails 80) via core_divide noop
[ 61840 evals] best 3.4624e-28 (fails 80) via core_divide noop
[ 62720 evals] best 3.46265e-28 (fails 80) via core_divide noop
[ 62800 evals] best 3.46265e-28 (fails 80) via core_undivide noop
[ 64880 evals] best 3.51879e-28 (fails 79) via level_fix k1: lvl2/rrlr → lvl0/rllll
[ 66240 evals] best 3.53264e-28 (fails 79) via retype 1/lrll->gh1
[ 66480 evals] best 3.55549e-28 (fails 79) via core_divide noop
[ 66800 evals] best 3.57445e-28 (fails 79) via crossover llrll<->llrll
[ 68000 evals] best 2.06412e-27 (fails 77) via level_fix k1: lvl2/rrlr → lvl0/rllll
[ 69120 evals] best 2.06883e-27 (fails 77) via core_divide noop
[ 71520 evals] best 2.76174e-27 (fails 76) via rotate 0/lr
[ 72640 evals] best 2.76227e-27 (fails 76) via core_undivide noop
[ 73520 evals] best 2.76276e-27 (fails 76) via core_divide noop
[ 75600 evals] best 2.7632e-27 (fails 76) via core_divide noop
[ 76320 evals] best 2.76358e-27 (fails 76) via core_undivide noop
[ 76960 evals] best 2.76389e-27 (fails 76) via core_undivide noop
[ 78880 evals] best 1.83174e-26 (fails 75) via level_retype 0/rrlrlr<->2/rlll
[ 78960 evals] best 1.87802e-26 (fails 75) via core_divide noop
[ 79360 evals] best 3.90146e-26 (fails 74) via swap 0/rlrl
[ 79520 evals] best 3.95859e-26 (fails 74) via rotate 0/llrr
[ 79600 evals] best 3.96056e-26 (fails 74) via crossover lrrl<->lllrl
[ 82720 evals] best 3.98751e-26 (fails 74) via core_undivide noop
[ 84240 evals] best 8.37192e-26 (fails 73) via crossover llrl<->lrr
[ 84640 evals] best 8.42852e-26 (fails 73) via rotate 2/rrrr
[ 84800 evals] best 3.5634e-25 (fails 71) via crossover rllr<->lrrl
[ 87040 evals] best 3.59072e-25 (fails 71) via rotate 0/rrll
[ 89120 evals] best 3.61957e-25 (fails 71) via retype 1/rlllr->lr1
[ 90800 evals] best 1.19078e-24 (fails 70) via crossover rlrrl<->rrlrr
[ 97680 evals] best 5.2661e-24 (fails 68) via rotate 0/lrl
[ 98560 evals] best 5.54831e-24 (fails 68) via retype 0/rlrlr->ur3
[103360 evals] best 5.61335e-24 (fails 68) via level_retype 2/llll<->1/rllll
[104400 evals] best 5.65917e-24 (fails 68) via core_divide noop
[106240 evals] best 5.68365e-24 (fails 68) via core_undivide noop
[108800 evals] best 1.16147e-23 (fails 67) via core_divide noop
[109760 evals] best 2.26647e-23 (fails 66) via swap 0/lrl
[112160 evals] best 2.27596e-23 (fails 66) via core_undivide noop
[112800 evals] best 9.05097e-23 (fails 64) via level_retype 0/llrll<->2/rrll
[115920 evals] best 9.55098e-23 (fails 64) via core_undivide noop
[117200 evals] best 9.59197e-23 (fails 64) via rotate 0/rlrl
coldstart harbor-house seed 1 @ 500000: 43 fails (7h/36s) Cold-start re-baseline after the DESIGN.md 38.10/38.11 objective change. Single worker (avoids homemaker-py-b8g), scored by the shipped scorer from the programme directory. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MJ84Feep79Hhm3E4zZJmnB
2026-08-31 16:50:54 +01:00
[120080 evals] best 9.62542e-23 (fails 64) via core_undivide noop
[120160 evals] best 3.45118e-22 (fails 62) via crossover rrlrl<->rlrrr
[120400 evals] best 3.4623e-22 (fails 62) via core_divide noop
[121280 evals] best 3.47395e-22 (fails 62) via core_divide noop
[127040 evals] best 3.48617e-22 (fails 62) via crossover llrrr<->lrlr
[130400 evals] best 3.49897e-22 (fails 62) via crossover ll<->ll
[141760 evals] best 6.80652e-22 (fails 61) via swap 0/lrl
[143680 evals] best 9.54701e-22 (fails 61) via crossover rlrrl<->lrll
[151440 evals] best 9.57869e-22 (fails 61) via level_retype 2/rllr<->1/lrll
[153120 evals] best 9.61188e-22 (fails 61) via retype 0/llrll->n
[157360 evals] best 9.64666e-22 (fails 61) via retype 0/rlrlr->tt
[159680 evals] best 1.85943e-21 (fails 60) via undivide 0/rlrr
[161040 evals] best 1.86564e-21 (fails 60) via core_undivide noop
[161440 evals] best 1.87142e-21 (fails 60) via retype 1/lrrr->gh1
[170240 evals] best 2.15038e-21 (fails 60) via swap 0/llrr
[172160 evals] best 2.16303e-21 (fails 60) via core_undivide noop
[180800 evals] best 2.17437e-21 (fails 60) via swap 0/rlrl
[188640 evals] best 2.18483e-21 (fails 60) via core_undivide noop
[192560 evals] best 2.1924e-21 (fails 60) via core_undivide noop
[197280 evals] best 2.19779e-21 (fails 60) via crossover rlrlr<->rlrlr
[199040 evals] best 2.20487e-21 (fails 60) via swap 0/lrl
[205040 evals] best 2.21102e-21 (fails 60) via crossover rrr<->rrr
[206480 evals] best 2.21103e-21 (fails 60) via retype 0/rrlrll->gh1
[212000 evals] best 2.21402e-21 (fails 60) via core_undivide noop
[215760 evals] best 2.21403e-21 (fails 60) via retype 2/llrrl->tt
[215920 evals] best 2.21422e-21 (fails 60) via retype 0/rrlrll->gh1
coldstart programme-house seed 1 @ 500000: 1 fails (1h/0s) Cold-start re-baseline after the DESIGN.md 38.10/38.11 objective change. Single worker (avoids homemaker-py-b8g), scored by the shipped scorer from the programme directory. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MJ84Feep79Hhm3E4zZJmnB
2026-08-31 21:50:30 +01:00
[225600 evals] best 2.21422e-21 (fails 60) via rotate 2/llrr
[227200 evals] best 2.21441e-21 (fails 60) via rotate 2/llrr
[230720 evals] best 2.21445e-21 (fails 60) via core_undivide noop
[234720 evals] best 2.21445e-21 (fails 60) via swap 0/rrll
[240240 evals] best 2.7509e-20 (fails 56) via level_fix ws1: lvl1/rllll → lvl0/rllll
[268800 evals] best 6.17587e-20 (fails 55) via swap 0/lr
[271440 evals] best 6.21537e-20 (fails 55) via rotate 2/llrr
[272480 evals] best 6.21597e-20 (fails 55) via swap 0/lrr
[275920 evals] best 6.23221e-20 (fails 55) via core_undivide noop
[279920 evals] best 6.23222e-20 (fails 55) via core_undivide noop
coldstart maple-court seed 0 @ 500000: 54 fails (18h/36s) Cold-start re-baseline after the DESIGN.md 38.10/38.11 objective change. Single worker (avoids homemaker-py-b8g), scored by the shipped scorer from the programme directory. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MJ84Feep79Hhm3E4zZJmnB
2026-09-01 03:21:09 +01:00
[290320 evals] best 2.47209e-19 (fails 53) via undivide 0/lrr
[290640 evals] best 2.47209e-19 (fails 53) via level_retype 2/rrrrl<->0/llrlr
[338880 evals] best 2.54027e-19 (fails 53) via undivide 0/llrll
[344000 evals] best 2.54987e-19 (fails 53) via core_undivide noop
[345120 evals] best 2.54991e-19 (fails 53) via core_undivide noop
[365760 evals] best 2.54992e-19 (fails 53) via rotate 0/rrll
Reference in a new issue Copy permalink