Porygon2

Successor from meloetta, hence porygon 2

Porygon2 provides:

• A UNIX server written in Node.js wrapped around the sim and client packages from pkmn.
• A reinforcement learning framework for interacting with this server. Currently supports R-NaD from DeepMind.

Overview

Porygon2 is a platform that simulates Pokémon battles and provides an environment for training reinforcement learning agents. It leverages the pkmn library for accurate game mechanics and offers a server-client architecture to facilitate interactions between agents and the simulation environment.

Installation

To set up the project, you can use the provided setup.sh script, which automates the installation process.

Prerequisites

• Python 3 installed on your system.
• Node.js and npm installed for running the server components.

Steps

1. Clone the Repository

git clone https://github.com/yourusername/porygon2.git
cd porygon2

2. Run the Setup Script

Make sure the setup.sh script is executable:

chmod +x setup.sh

Then run the script:

./setup.sh

This script will:

• Create a Python virtual environment in the venv directory.
• Install all Python dependencies from requirements.txt files located in the root directory and immediate subdirectories, excluding any directories named env.
• Install all Node.js dependencies by running npm install in directories (one layer deep) containing a package.json file, excluding any directories named env or node_modules.

3. Activate the Python Virtual Environment

After running the script, your Python virtual environment is activated. If you open a new terminal session, reactivate it using:

source venv/bin/activate

Training

Server Setup

1. Navigate to the Server Directory

cd service

2. Install Dependencies

npm install

3. Run Tests (Optional)

npm run test

4. Start the Training Server

npm run start

Client Setup

Open a new terminal and navigate to the root directory of the repository.

1. Activate the Virtual Environment

source venv/bin/activate

2. Run the Training Client

python ml/rl.py

Evaluation

Server Setup

1. Activate the Virtual Environment

source venv/bin/activate

2. Start the Inference Server

python inference/server.py

Client Setup

1. Navigate to the Server Directory

cd server

2. Install Dependencies

npm install

3. Start the Evaluation Client

npm run start-evaluation-client

Scripts

The scripts/ directory contains helper scripts for various tasks:

• compile_protos.sh: Compiles protocol buffer definitions.
• generate_requirements.sh: Generates requirements.txt files.
• lint.sh: Runs code linters to ensure code quality.
• make_data.sh: Generates necessary data for the project.

Development Notes

• Directory Structure: The project contains several subdirectories with their own npm installs. The setup.sh script handles these installations automatically.
• Python Virtual Environment: A virtual environment named env is created in the root directory. Always activate it before running Python scripts.

Additional Information

• License: This project is licensed under the MIT License.