Skip to content

Latest commit

 

History

History
20 lines (12 loc) · 621 Bytes

README.md

File metadata and controls

20 lines (12 loc) · 621 Bytes

RL_Atari

using DQN to learn atari games

Required Dependencies : gym, torch, torchvision, cv2, numpy

main: Training loop, just run this to get results

  • change env_name for different environment, use '{env_name}NoFrameskip-v4' format
  • change episode numbers in train-loop execution

environments: preprocessing

model: The DQN model

memory: Replay memory

results on 400 episodes of training, test wins all 21 points data

test