SirStone
ca3e3d2272
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)
19 lines
477 B
Bash
19 lines
477 B
Bash
# TRAINING_OPPONENT=Fire
|
|
TRAINING_ROUNDS=30000
|
|
PPOB_LOG_FILE=/home/davide/Projects/SirRoboGarage/tools/training_runner/logs/fire_training.jsonl
|
|
PPOB_LR=1e-4
|
|
PPOB_UPDATE_INTERVAL=10
|
|
PPOB_CLIP_EPSILON=0.2
|
|
PPOB_ENTROPY_COEFF=0.0
|
|
PPOB_VALUE_LOSS_COEFF=0.5
|
|
PPOB_MAX_GRAD_NORM=0.5
|
|
PPOB_GAMMA=0.99
|
|
PPOB_LAM=0.95
|
|
PPOB_EPOCHS=4
|
|
PPOB_MINI_BATCH_SIZE=64
|
|
PPOB_LOG_STD_FLOOR=-3.0
|
|
PPOB_LOG_STD_CEILING=-1.0
|
|
PPOB_INITIAL_LOG_STD=-0.5
|
|
|
|
# PPOB_EVAL_ONLY=1 → freeze training (pure evaluation)
|