dqn-scratch is an implementation of Deep Q-Learning from scratch with Python and PyTorch. The goal of this project is to provide a simple, well-documented implementation of DQN that is easy to understand and modify.
To train the agent, run the following command:
python main.py
Use the main.py file as a template for how to train and evaluate the agent. You can modify the hyperparameters, the environment, and the neural network architecture to suit your needs.
conda create -n env_name
conda activate env_name
conda install pytorch::pytorch torchvision torchaudio -c pytorch
conda install pygame
conda install gymnasium
pip install box2d
- Define the Q-Network class [DONE]
- Define the Replay Buffer class [DONE]
- Define the Agent class [DONE]
- Define the training loop [DONE]
- Training and evaluation [DONE]
- Define the main function [DONE]