项目作者: jbecke

项目描述 :
Simple, well-commented Pytorch implementations of REINFORCE and Actor Critic RL methods.
高级语言: Python
项目地址: git://github.com/jbecke/Open-AI-Gym-ABCs.git
创建时间: 2018-07-04T17:52:48Z
项目社区:https://github.com/jbecke/Open-AI-Gym-ABCs

开源协议:MIT License

下载


Open AI Gym ABCs

Many RL tutorials contain superfluous code or use a Jupyter notebook with lots of text. This repo aims to be a no-frills implementation (i.e. maximally shortened) of REINFORCE and Actor Critic RL methods.

Suggestions

Start by reading over reinforce.py as it is the simplest method. After that, read about actor critic methods (see this comic), then look at the code. The comments assume basic knowledge of RL. It helps to have an understanding of traditional RL (i.e. pre-NNs, Bellman equations) before trying your hand with Pytorch or OpenAI Gym.

Prereqs

  • OpenAI gym with the cartpole environment installed
  • Python 3.6 (but older versions including Python 2 may work)
  • numpy

Training time

The models shold train within several minutes on a modern laptop CPU.