A custom PyTorch neural network engine and course notebooks for CSCI 357 - AI and Neural Networks.

This repo has two main components:

• src/lantern — A custom PyTorch neural network engine that provides reusable building blocks for training and evaluating models. Includes support for configurable model architectures (MLP, CNN, NLP), training loops, optimizers, learning rate schedulers, checkpointing, early stopping, metrics, and W&B sweep utilities.

• notebooks/ — Jupyter notebooks containing class notes, labs, and homework assignments for CSCI 357.

It supports Weights and Biases (wandb) logging for both regular training and sweep training.

Using VSCode with Google Colab extension (no Python environment/package management required)

Name(s): Chang Min
Course: CSCI 357 - AI and Neural Networks
Section: 01 - 1:00pm
Semester: Spring 2026
Instructor: Prof. King

Some of the code may have been generated by LLMs (will be indicated in comment)

Some of the Jupyter notebooks may not run successfully as the neural network training framework/engine has evolved continuously

I use the following command to make uv point to a local .venv:

export UV_PROJECT_ENVIRONMENT=/Users/changminbark/Desktop/Bucknell/2026SS/CSCI357/.venv