Skip to content

[UPDATED] PyTorch Implementation of DQN and training Super Mario Bros

Notifications You must be signed in to change notification settings

Rochan-A/dqn-mario

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

45 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Reinforcement Learning for Super Mario Bros using DQN

Modified to work with gym-super-mario-bros

NOTE: This is an unofficial fork of the original code published as dqn-mario.

DQN

使用PyTorch實作DQN演算法,並訓練super-mario-bros,整體架構參考openai/baselines。

Warning:訓練DQN請開足夠的記憶體,Replay Buffer以預設值1000000為例至少會使用約8G的記憶體.

Dependencies

  • Python 3.6
  • PyTorch
  • gym
  • gym-super-mario-bros

Result

  • Super-Mario-Bros

使用8顆cpu在GCP上跑16個小時,RAM開24G非常足夠,但很難收斂,無法穩定過關。 訓練的影像預設位置在/video/mario/。

References

Playing Atari with Deep Reinforcement Learning
openai/baselines
transedward/pytorch-dqn
openai/gym
Kautenja/gym-super-mario-bros

About

[UPDATED] PyTorch Implementation of DQN and training Super Mario Bros

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%