Reinforcement Learning Commentary Labelling Platform

This repository contains the code of a website to label the commentary for Reinforcement Learning decisions in Pong and Maze environment.

Web Interface

Pre-install dependencies

• Install requirements

pip install -r requirements

Setup Database

• Create Database
  • Change mode='local' if you want to run on the local database

python db/create_db.py

• Customize the database schema in db/create_db.py
• Change the name of db to not interfere with other database: DB_NAME = 'xxx'
• Look for function save_obs_to_db() in agent/test.py to see how to insert an observation into the database.
• Create more API functions in db/api.py.
• Remember to call API.close_connection() after finishing queries to database.
• Run web server

python app/routes/routes.py

Access to Web server

• Set up SSH if not have SSH key pair yet: https://pastebin.pl/view/62678967.
• After setting up SSH, input local public key to the IRender server (Password: Khang112@).

cat ~/.ssh/id_rsa.pub | ssh -p 2056 -t administrator@58.186.80.21 "mkdir -p ~/.ssh && touch ~/.ssh/authorized_keys && chmod -R go= ~/.ssh && cat >> ~/.ssh/authorized_keys"

• From now, connect directly to the IRender server without inputting the password.
• Access to the web server, run this command in the terminal:

ssh -L 8080:192.168.20.56:2402 -p 2056 -t administrator@58.186.80.21 -N

• Access http://localhost:8080/ on the web browser.

Architecture

.
├── ...
├── agent                           # Rainbow Agent  
├── app              
│   ├── routes         
│   │   ├── forms.py                # Create form to fill from front-end form
│   │   ├── routes.py               # Run main from here (contains controller)
│   ├── templates            
│   │   ├── _formhelpers.html       # Render forms
│   └────── index.html              # Index page
├── db                              # Database
│   ├── api.py                      # API query to insert/read data from database
│   ├── connect_db.py               # Connect to database through a SSHTunnelForwarder
│   ├── create_db.py                # Create models on database      
├── ref_app                         # Some references 
├── requirements.txt                # Some requirements 
└── README.md

Database Architecture:

Run Rainbow agent

python main.py --game pong --model "agent/results/pong/model.pth" --architecture data-efficient --hidden-size 256 --insert_obs --evaluate