Add weighting function, change the reward logic
This commit is contained in:
@@ -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
|
||||
|
||||
Reference in New Issue
Block a user