search

takoika / PrioritizedExperienceReplay

Yet another prioritized experience replay buffer implementation.
MIT License
48 stars 12 forks source link

Weight normalization on list #4

Open YiTanJang opened 6 years ago

YiTanJang commented 6 years ago

on line 84 of prorportional.py 

 weights /= max(weights) # Normalize for stability

should be fixed since weights is a list 

weights = [ i/max(weights) for i in weights]