SirRoboGarage

SirStone/SirRoboGarage

Fork 0

Commit Graph

Author	SHA1	Message	Date
SirStone	ca3e3d2272	tune(PPO_Bot): logStd=-2.0 (std≈0.135), entropy=0, ceiling=-1.0 Stochastic eval at std≈0.37 was 0/10 vs Corners (deterministic: 10/10). Warm-start policy is correct but brittle — any noise breaks it. - log_std initialized to -2.0 (std≈0.135) for moderate exploration - entropy_coeff=0.0 (no push toward exploration during fine-tuning) - logStd ceiling=-1.0 (cap at std≈0.37)	2026-08-20 15:28:15 +02:00
SirStone	0d35646dc9	feat(PPO_Bot): deterministic eval + fix logStd warm-start - actorForward: deterministic param, uses mean-only when PPOB_EVAL_ONLY=1 (eval was adding unit Gaussian noise to every action — unreliable scores) - warm_start.py: log_std initialized to -1.0 (std≈0.37) instead of copying snapshot values (were 2.27-4.68 → std 9-108, completely drowning signal) - training.env: LOG_STD_CEILING 0.0→-0.5 (cap exploration at std≈0.6)	2026-08-20 15:18:07 +02:00
SirStone	12624d3069	feat(PPO_Bot): bot-relative bullets + scan staleness (STATE_DIM=57) - Bullet state (indices 44-55): enemy-relative → bot-relative frame (bot needs threat vectors to itself for dodging, not to enemy) - New index 56: scan staleness = min(ticksSinceLastScan / 30, 1.0) (gives policy a confidence signal for enemy data freshness) - warm_start.py updated: 44→57 dim expansion, TARGET_DIM variable - Tests updated for new state layout Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>	2026-08-20 14:27:28 +02:00

Author

SHA1

Message

Date

SirStone

ca3e3d2272

tune(PPO_Bot): logStd=-2.0 (std≈0.135), entropy=0, ceiling=-1.0

Stochastic eval at std≈0.37 was 0/10 vs Corners (deterministic: 10/10).
Warm-start policy is correct but brittle — any noise breaks it.
- log_std initialized to -2.0 (std≈0.135) for moderate exploration
- entropy_coeff=0.0 (no push toward exploration during fine-tuning)
- logStd ceiling=-1.0 (cap at std≈0.37)

2026-08-20 15:28:15 +02:00

SirStone

0d35646dc9

feat(PPO_Bot): deterministic eval + fix logStd warm-start

- actorForward: deterministic param, uses mean-only when PPOB_EVAL_ONLY=1
  (eval was adding unit Gaussian noise to every action — unreliable scores)
- warm_start.py: log_std initialized to -1.0 (std≈0.37) instead of copying
  snapshot values (were 2.27-4.68 → std 9-108, completely drowning signal)
- training.env: LOG_STD_CEILING 0.0→-0.5 (cap exploration at std≈0.6)

2026-08-20 15:18:07 +02:00

SirStone

12624d3069

feat(PPO_Bot): bot-relative bullets + scan staleness (STATE_DIM=57)

- Bullet state (indices 44-55): enemy-relative → bot-relative frame
  (bot needs threat vectors to itself for dodging, not to enemy)
- New index 56: scan staleness = min(ticksSinceLastScan / 30, 1.0)
  (gives policy a confidence signal for enemy data freshness)
- warm_start.py updated: 44→57 dim expansion, TARGET_DIM variable
- Tests updated for new state layout

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

2026-08-20 14:27:28 +02:00

3 Commits