r2seedo

R2 See, Do - a journey in deep reinforcement learning.

Table of contents

• Notebooks
• Project Setup
• RL Environments
• Running Tests

Project Setup

0. Install hatch & hatch-conda

r2seedo uses hatch for project management. You'll need it installed (ideally in an isolated environment) before setting up r2seedo.

(optional) Install hatch-conda

r2seedo depends on gymnasium, which is best installed with conda / mamba, thus hatch-conda is also recommended.

# Update mamba
mamba update -n base mamba

# Update base environment packages 
mamba update -n base --all

# Install hatch & hatchling
mamba install -n base -c conda-forge hatch hatchling

# Install hatch-conda
mamba activate base && python -m pip install hatch-conda

1. Clone the r2seedo repository

git clone [email protected]:libertininick/r2seedo.git

2. Create the default (virtual) environment

hatch will install r2seedo in development mode along with its development dependencies inside of a virtual environment managed by hatch.

# Navigate to root project directory
cd r2seedo

# (optional) if using conda / mamba envs activate the base environment
mamba activate base

# Create default environment
hatch env create

Table of Contents

Create separate virtual environments for specific RL environments

• RL environment packages are best installed in independent virtual environment
• Each RL environment used in r2seedo has its own virtual environment configuration defined in pyproject.toml
• An environment can be created using hatch env create <env name> :

NOTE: Make sure the base environment is activated before creating a new environment: mamba activate base

Environment	Name	Create
Atari	gym-atari	hatch env create gym-atari
Box2D	gym-box2d	hatch env create gym-box2d
Toy Text	gym-toy_text	hatch env create gym-toy_text

Table of Contents

Running Tests

Run tests and coverage report using hatch

hatch run default:test-cov

Table of Contents