1st collaboration workshop on Reinforcement Learning for Autonomous Accelerators (RL4AA'23)

This repository contains the material for the second day of the RL4AA'23 event.

Homepage for RL4AA Collaboration: https://rl4aa.github.io/

Workshop organizing committee

• Andrea Santamaria Garcia (KIT)
• Simon Hirländer (University of Salzburg)
• Jan Kaiser (DESY)
• Chenran Xu (KIT)

Slides

• Welcome and basics of RL, Andrea Santamaria Garcia
• Advanced concepts in RL, Simon Hirländer
• RL in particle accelerator control: are we there yet?, Simon Hirländer

Python tutorial: reinforcement learning implementation example

• Github repository containing the material: https://github.com/RL4AA/RL4AA23
• Tutorial in slide form: here

Getting started

• First, download the material to your local disk by cloning the repository: git clone https://github.com/ansantam/RL4AA23
• If you don't have git installed, you can click on the green button that says "Code", and choose to download it as a .zip file.

Install ffmpeg

• OS X: brew install ffmpeg
• Ubuntu: sudo apt-get install ffmpeg
• Ubuntu 14.04: sudo apt-get install libav-tools
• With pip: pip install imageio-ffmpeg

Setup the environment locally

• Open terminal app
• (Suggested) Create a virtual envrionment using conda or venv.

venv

python3 -m venv rl4aa
source rl4aa/bin/activate
pip3 install -r requirements.txt
jupyter notebook

• Open the tutorial notebook tutorial.ipynb in the jupyter server in browser
• When you are done type deactivate

conda only

Instructions to install conda here

conda env create -f environment.yml
conda activate rl4aa
jupyter notebook

• Open the tutorial notebook tutorial.ipynb in the jupyter server in browser
• When you are done type conda deactivate to deactivate the virtual environment

conda + pip

cd path_to_your_folder/RL4AA23

conda create -n rl4aa python=3.10
conda activate rl4aa
pip3 install -r requirements.txt
jupyter notebook

• Open the tutorial notebook tutorial.ipynb in the jupyter server in browser
• When you are done type conda deactivate to deactivate the virtual environment