issues
search
kengz
/
SLM-Lab
Modular Deep Reinforcement Learning framework in PyTorch. Companion library of the book "Foundations of Deep Reinforcement Learning".
https://slm-lab.gitbook.io/slm-lab/
MIT License
1.25k
stars
264
forks
source link
Tune benchmark specs
#377
Closed
kengz
closed
5 years ago
kengz
commented
5 years ago
Feature / Fix
fix PPO training frequency to be time horizon directly due to batched storage for parallel env
fix
GlobalRMSprop
ref
tune Pong specs to be runnable within reasonable time
tune BipedalWalker specs: A2C ok, PPO better but not 300
tune misc specs for benchmark
Feature / Fix
GlobalRMSprop
ref