pianomania/paac-pytorch

PyTorch implementation of PAAC from "Efficient Parallel Methods for Deep Reinforcement Learning"

paac-pytorch

This is a PyTorch implementation of PAAC from "Efficient Parallel Methods for Deep Reinforcement Learning"

Environments

• Ubuntu 16.04
• python 3.5.2
• PyTorch 0.2.0
• NumPy 1.13.1
• gym 0.9.2
• matplotlib

Usage

• You can train the agent by:

python main.py --env-name BreakoutDeterministic-v4 --num-workers 4

• You can play the game by:

python play.py --env-name BreakoutDeterministic-v4

Results

Notes

References

• origin paper's open source
• openai's universe-starter-agent
• pytorch-a3c