Add weighting function, change the reward logic

This commit is contained in:
2026-03-22 21:11:46 -04:00
parent a642274fa6
commit 7f7ec781c5
4 changed files with 84 additions and 104 deletions

View File

@@ -39,7 +39,7 @@ params:
reward_shaper:
scale_value: 1.0
normalize_advantage: True
gamma: 0.96
gamma: 0.98
tau: 0.95
learning_rate: 3e-4
lr_schedule: adaptive