LunarLander DQN

This is my personal DQN project for LunarLander (via gymnasium) — TensorFlow-based, notebook-driven, and meant for tinkering.

What's here

• main.ipynb: training + eval + video recording
• q_network.h5, target_q_network.h5: saved networks
• videos/: evaluation clips (mp4)

Quick setup

Create a venv and install the essentials:

python -m venv .venv; .\.venv\Scripts\Activate.ps1
pip install -U pip
pip install gymnasium[box2d] tensorflow numpy

Run main.ipynb to train or eval; the notebook uses RecordVideo and drops MP4s into videos/.

Learning-progress clips

Here are a few representative eval clips from the training run — early → mid → later. Playable in the README (if your browser supports MP4):

Episode 0 — very early

Episode 10 — mid training

Episode 29 — later / best-ish