homemaker-layout

bruno/homemaker-layout

Fork 0

Commit graph

Author	SHA1	Message	Date
Bruno Postle	bb9b355f14	x3b/§13.10: productionise leaf-sharing — per-code share grain + CLI wiring Make the §13.3 lever a first-class feature, not experiment-only. - programme.py: SpaceReq.share (default 1) + has_share, parsed from patterns.config 'share: N'. - operators._share_grain: resolve per-code grain from leaf_share_factor selector — 0 = per-code opt-in (share iff share:N>=2), >=2 = global with per-code override (share:1 opts OUT, share:N sets grain). _share_rooms groups per resolved grain. - End-to-end conf injection without monkeypatch: load_config(overrides=) merges run-level keys last; driver.search / innerloop.optimise / NativeEvaluator / _fitness_for thread conf_overrides={leaf_sharing:True} through both inner-loop and off-tree scorers when sharing is on. - homemaker-evolve: --leaf-sharing/--no-leaf-sharing + --leaf-share-factor (env HOMEMAKER_LEAF_SHARING / HOMEMAKER_LEAF_SHARE_FACTOR). - Example programmes untouched (§13.3/§13.9 stay reproducible). Experiment load_config monkeypatches updated to accept overrides=. Tests: grain modes, opt-out, default-OFF parity, load_config overrides, programme parse, CLI parse. 233 pass. Smoke: harbor 37 vs 95 fails on/off. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-06-28 22:04:35 +01:00
Bruno Postle	4aaa295dc1	erc.4: depth-balanced construction mechanism + floor probe (§13.4) _grow_leaves grew a random caterpillar, so equal-target rooms landed at wildly different binary-tree depths — the depth-driven size maldistribution Diagnostic B (§13.2) localized (same code at 0.05x and 14.7x target). The depth_balanced flag always splits a shallowest leaf instead, growing a near-complete tree so the proportion-aware sizing pass hits each target with cut fractions near their proportional value. Floor probe (diag_depth_balance.py): depth spread collapses 7->1, the giant ratio falls (maxR 12->8 harbor / 16->6 maple), %undersize 54->25 / 42->22, and the achievable floor drops -12% harbor / -11% maple at EQUAL leaf count. Additive with leaf-sharing (bal+sh3 beats §13.3 share3-alone). Default OFF, 214 tests pass; threaded through driver.search/search_staged and exposed via DEPTHBAL in run_staged_search.py. End-to-end 20k A/B running. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-06-25 22:36:24 +01:00

Author

SHA1

Message

Date

Bruno Postle

bb9b355f14

x3b/§13.10: productionise leaf-sharing — per-code share grain + CLI wiring

Make the §13.3 lever a first-class feature, not experiment-only.

- programme.py: SpaceReq.share (default 1) + has_share, parsed from
  patterns.config 'share: N'.
- operators._share_grain: resolve per-code grain from leaf_share_factor
  selector — 0 = per-code opt-in (share iff share:N>=2), >=2 = global with
  per-code override (share:1 opts OUT, share:N sets grain). _share_rooms
  groups per resolved grain.
- End-to-end conf injection without monkeypatch: load_config(overrides=)
  merges run-level keys last; driver.search / innerloop.optimise /
  NativeEvaluator / _fitness_for thread conf_overrides={leaf_sharing:True}
  through both inner-loop and off-tree scorers when sharing is on.
- homemaker-evolve: --leaf-sharing/--no-leaf-sharing + --leaf-share-factor
  (env HOMEMAKER_LEAF_SHARING / HOMEMAKER_LEAF_SHARE_FACTOR).
- Example programmes untouched (§13.3/§13.9 stay reproducible). Experiment
  load_config monkeypatches updated to accept overrides=.

Tests: grain modes, opt-out, default-OFF parity, load_config overrides,
programme parse, CLI parse. 233 pass. Smoke: harbor 37 vs 95 fails on/off.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2026-06-28 22:04:35 +01:00

Bruno Postle

4aaa295dc1

erc.4: depth-balanced construction mechanism + floor probe (§13.4)

_grow_leaves grew a random caterpillar, so equal-target rooms landed at
wildly different binary-tree depths — the depth-driven size maldistribution
Diagnostic B (§13.2) localized (same code at 0.05x and 14.7x target). The
depth_balanced flag always splits a shallowest leaf instead, growing a
near-complete tree so the proportion-aware sizing pass hits each target with
cut fractions near their proportional value.

Floor probe (diag_depth_balance.py): depth spread collapses 7->1, the giant
ratio falls (maxR 12->8 harbor / 16->6 maple), %undersize 54->25 / 42->22,
and the achievable floor drops -12% harbor / -11% maple at EQUAL leaf count.
Additive with leaf-sharing (bal+sh3 beats §13.3 share3-alone). Default OFF,
214 tests pass; threaded through driver.search/search_staged and exposed via
DEPTHBAL in run_staged_search.py. End-to-end 20k A/B running.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2026-06-25 22:36:24 +01:00

2 commits