microsoft/nni ? reverse-engineered prompt

Build me an open source AutoML toolkit for Python that helps people tune models without doing all the manual trial and error.

I want it to support hyperparameter tuning, neural architecture search, feature engineering, and model compression for deep learning models. It should work with common ML frameworks like PyTorch and TensorFlow, and also be useful for regular machine learning workflows. Please include a simple way to define an experiment, run lots of trials, compare results, and stop bad runs early so it saves time.

Also make sure there’s a clean command line and a basic web style interface or dashboard for tracking experiments, plus a few practical examples and docs so someone can get started quickly. Use the existing project structure if it helps, and look up current docs online if you need to.