modlee

Modlee is a machine learning tool that documents experiments for reproduciblity and recommends neural network models suited for a particular dataset. Modlee bypasses costly machine learning experimentation by recommending performant models based on prior experiments. Built on top of MLFlow, Modlee documents traditional experiment assets (model checkpoints, (hyper)parameters, performance metrics) and meta-features for meta-learning. Based on these meta-features from prior experiments, Modlee recommends a neural network model matched to a new task.

Getting started

Learn about modlee and how to get started.

API Reference

Reference the source code and how to use the API.

Examples

View starter examples for documentation or recommendation.

Indices and tables