gitricko / rljs-service

RL JS MicroService

rljs-service

RL JS MicroService

Prerequisites of setting up Development Environment

• VSCode as IDE
• Miniconda

Dev Setup

• Clone github and change into the github directory
• Create conda environment: conda env create -f environment.yml
• Enter conda environment: conda activate rljs-service
• Setting up npm packages: npm run build

Running

• Enter conda environment: conda activate rljs-service if you have not.
• Run Server with debugging & auto-reload/restart: npm run dev
  • In VSCode's debugging - use Attach by Process ID to start debugging
• Run Server in production mode: npm run prod

Testing

• Only waterworld example is integrated with server-side RL.DQNAgent (nodejs)
  • Go to: http://localhost:3000/waterworld.html
  • Let it run for a few hours to see agent improves; OR
  • Click on Load a Pretrained Agent button to see a train RL model
  • Check the scores on apples vs poison