Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures

This repository contains an PyTorch implementation of "Importance Weighted Actor-Learner Architectures".

For a detailed description of the architecture please read paper.

Requirements

• Python >=3.7
• PyTorch >=1.2.0
• atari_py >=0.1.7
• cv2 >= 3.4.2

TODO

• Experiment's results.
• Try to remove the ParameterServer.
• Distributed Training.
• Other games such as dmlab.