Supermariobros-PPO-pytorch

基于超级玛丽游戏的pytorch版本强化学习实践教程

rl(ppo) course with super-mario-bros

你可以直接在jupyter notebook中开始学习（course.ipynb、course2.ipynb）

run the code with docker (推荐)

play with docker (ON your local computer with display),just run:

推荐使用docker直接运行，可以无需关注软件环境

docker run --gpus all -v /tmp/.X11-unix:/tmp/.X11-unix registry.cn-shanghai.aliyuncs.com/tcc-public/super-mario-ppo:localdisplay 

if you want debuge the code and exec into container ,command like this:

docker run --gpus all -it -v /tmp/.X11-unix:/tmp/.X11-unix registry.cn-shanghai.aliyuncs.com/tcc-public/super-mario-ppo:localdisplay  /bin/bash

train the model:

python ppo_lstm.py

test on super-mario-bros(see the video of agent):

python test_lstm.py

run the code witch conda

conda create -n ppo python=3.7
conda activate ppo

python request:

torch torchvision
gym_super_mario_bros
spinup(要用源码安装：https://spinningup.openai.com/en/latest/user/installation.html)
opencv-python

train:

python ppo_lstm.py

test:

python test_lstm.py

learn the course in jupyter notebook:

the notebook can be find at course.ipynb、course2.ipynb

jion the rl Communication group,contact us:

remarks（添加请备注）：github rl

learn more in our DRL Training camp （aliyun tianchi）

you can find some ppo info on https://tianchi.aliyun.com/specials/promotion/aicamprl