PyTorch implementation of Advantage async actor-critic Algorithms (A3C) in PyTorch
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
assets
models/lstm
.gitignore
A3C.py
README.md
__init__.py
environment.py
my_optim.py
test_lstm.sh
train.py
train_lstm.sh

README.md

Advantage async actor-critic Algorithms (A3C) in PyTorch

@inproceedings{mnih2016asynchronous,
  title={Asynchronous methods for deep reinforcement learning},
  author={Mnih, Volodymyr and Badia, Adria Puigdomenech and Mirza, Mehdi and Graves, Alex and Lillicrap, Timothy P and Harley, Tim and Silver, David and Kavukcuoglu, Koray},
  booktitle={International Conference on Machine Learning},
  year={2016}}

This repository contains an implementation of Adavantage async Actor-Critic (A3C) in PyTorch based on the original paper by the authors and the PyTorch implementation by Ilya Kostrikov.

A3C is the state-of-art Deep Reinforcement Learning method.

Dependencies

  • Python 2.7
  • PyTorch
  • gym (OpenAI)
  • universe (OpenAI)
  • opencv (for env state processing)
  • visdom (for visualization)

Training

./train_lstm.sh

Test wigh trained weight after 169000 updates for PongDeterminisitc-v3.

./test_lstm.sh 169000

A test result video is available.

Check the loss curves of all threads in http://localhost:8097

loss_png

References