KoalaTrapLord 11990c2fd3 feat(ml): add --data-source forge flag to PPO trainer (issue #83 phase 1.3) ...

Wires forge_episode_generator into the PPO training loop:
- New --data-source flag: synthetic (default) | forge | mixed
- New --forge-results-dir, --forge-batch-id, --forge-weight flags
- ForgeDataSource class: runs run_forge_pipeline() in background thread,
  drains episodes into RolloutBuffer via episodes_to_buffer()
- Mixed mode: blends Forge episodes + synthetic self-play per iteration
  using --forge-weight ratio (0.0–1.0)
- Synthetic path unchanged — no regression for existing training runs

2026-03-29 03:02:30 -04:00

..

actions

fix(ml): replace undefined 'log' with 'logger' in labeler.py

2026-03-21 15:47:47 -04:00

config

feat: Phase 5 architecture follow-up — zone boundary, WAL store, Forge-only defaults (#87)

2026-03-22 14:41:33 -04:00

data

feat: Phase 5 architecture follow-up — zone boundary, WAL store, Forge-only defaults (#87)

2026-03-22 14:41:33 -04:00

encoder

feat: Phase 5 architecture follow-up — zone boundary, WAL store, Forge-only defaults (#87)

2026-03-22 14:41:33 -04:00

eval

fix: add top_n cutoff to ELO tournament (default top 10 generations)

2026-03-27 01:48:10 -04:00

models

v23: React + TypeScript SPA migration

2026-03-12 01:12:02 +00:00

scripts

feat(ml): add update_weights.py — outcome-weighted RL learning from decision logs

2026-03-14 02:58:32 -04:00

serving

feat: Phase 5 architecture follow-up — zone boundary, WAL store, Forge-only defaults (#87)

2026-03-22 14:41:33 -04:00

training

feat(ml): add --data-source forge flag to PPO trainer (issue #83 phase 1.3)

2026-03-29 03:02:30 -04:00

__init__.py

Initial commit

2026-03-10 12:13:56 -04:00