easy-rl/codes

Eng|中文

Introduction

This repo is used to learn basic RL algorithms, we will make it detailed comment and clear structure as much as possible:

The code structure mainly contains several scripts as following：

• model.py basic network model of RL, like MLP, CNN
• memory.py Replay Buffer
• plot.py use seaborn to plot rewards curve，saved in folder result.
• env.py to custom or normalize environments
• agent.py core algorithms, include a python Class with functions(choose action, update)
• main.py main function

Note that model.py,memory.py,plot.py shall be utilized in different algorithms，thus they are put into common folder。

Runnig Environment

python 3.7、pytorch 1.6.0-1.7.1、gym 0.17.0-0.18.0

Usage

运行带有train的py文件或ipynb文件进行训练，如果前面带有task如task0_train.py，表示对task0任务训练 类似的带有eval即为测试。

run python scripts or jupyter notebook file with train to train the agent, if there is a task like task0_train.py, it means to train with task 0.

similar to file with eval, which means to evaluate the agent.

Schedule

Name	Related materials	Used Envs	Notes
On-Policy First-Visit MC		Racetrack
Q-Learning		CliffWalking-v0
Sarsa		Racetrack
DQN	DQN-paper,Nature DQN Paper	CartPole-v0
DQN-cnn	DQN-paper	CartPole-v0
DoubleDQN		CartPole-v0
Hierarchical DQN	Hierarchical DQN	CartPole-v0
PolicyGradient		CartPole-v0
A2C	A3C Paper	CartPole-v0
SAC	SAC Paper
PPO	PPO paper	CartPole-v0
DDPG	DDPG Paper	Pendulum-v0
TD3	TD3 Paper	HalfCheetah-v2

Refs

RL-Adventure-2

RL-Adventure