jgrusewski
eceaf45b11
feat: --full mode for 3-phase hyperopt pipeline (BC → RL → Refinement)
- CampaignMode enum: Quick, Standard, Full
- dqn_full() constructor: 20 trials × 100 epochs (covers all 3 phases)
- fxt tune start --full flag: auto-sets 20 trials, 100 epochs
- Phase 1 (BC): MSE warmup + expert demos + DT pretrain
- Phase 2 (RL): all 25 features, C51 ramp, HER
- Phase 3 (Refine): pure C51, shrink-and-perturb
Usage: fxt tune start --model dqn --full --gpu
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-25 08:44:41 +01:00
..
2026-03-03 22:16:35 +01:00
2026-03-04 01:34:34 +01:00
2026-03-03 22:16:35 +01:00
2026-03-03 22:16:35 +01:00
2026-03-03 22:29:13 +01:00
2026-03-04 23:46:36 +01:00
2026-03-03 22:29:13 +01:00
2026-03-04 14:54:38 +01:00
2026-03-03 22:29:13 +01:00
2026-03-03 22:29:13 +01:00
2026-03-03 22:29:13 +01:00
2026-03-03 22:29:13 +01:00
2026-03-03 22:16:35 +01:00
2026-03-04 22:58:40 +01:00
2026-03-25 08:44:41 +01:00
2026-03-03 21:48:59 +01:00