Reinforcement Learning

started: 2/23/2022

last edited: 1/10/2023

By: James Nesfield

PURPOSE:

The purpose of this repo is to store and publicly display my exploratory work in reinforcement learning

use docker in ubuntu:

• install ubuntu 18.04 lts as an operating system :

  • https://ubuntu.com/download/desktop
    

• update ubuntu and install proper nvidia drivers using apt-get:

  • https://www.cyberciti.biz/faq/ubuntu-linux-install-nvidia-driver-latest-proprietary-driver/
    

• install docker and if needed nvidia packages so you may access gpu resources:

  • https://docs.nvidia.com/ai-enterprise/deployment-guide/dg-docker.html
    

• build a new image using the dockerfile in this repo:

  • https://raw.githubusercontent.com/jnesfield/Reinforcement-Learning/main/DockerFile
    

• download the docker file and run this changing the user name and path to match your own set up:

sudo docker build  -f /home/james/docker_aigym/DockerFile /home/james/docker_aigym -t james_n_aigym:0.1

• /home/james/docker_aigym/DockerFile is where the docker file is

• /home/james/docker_aigym is the context

• -t james_n_aigym:0.1 specifies the name you will use to run the image later thus:

docker run james_n_aigym:0.1

• ideally use something like this to run

docker run -p 8888:8888 -it --gpus all -v <local directory to mount>:/tf/mystuff/ james_n_aigym:0.1

some learning resources

http://incompleteideas.net/

https://stable-baselines.readthedocs.io/en/master/

https://stable-baselines.readthedocs.io/en/master/guide/rl.html

https://spinningup.openai.com/en/latest/

good luck!