magicml 0.0.0

☕️ About

MagicML is an open source software with a Graphical User Interface (GUI) to simplify Machine Learning (ML) models usage following the MLOps paradigm. It provides a collection of (scientific) machine learning algorithms and the necessary tooling for model management.

MagicML take some inspiration from the Weka software but in a lightweight manner, leverage the python ML ecosystem, the modern web stack (python Dash) and try to follow the MLOps guidelines.

🎯 Goals

• Make ML algorithms more acessible
• Create an open source ML workbench for researchers and enginners

🚀 Features

• Model architecture -> model manager for using pre-implemented models and allow the user to add this owns. In addition, by supporting the Open Neural Network Exchange format (ONNX), MagicML is framework agnostic.
• Model training -> train from scratch the model using the built-in tooling. In addition pre-trained models could also be used for specific tasks.
• Model evaluation -> for a selected model, the corresponding State Of The Art (SOTA) metrics are provided in order to assess the model performances.
• Model versioning -> a tool for versioning ML training runs and experiments
• Model deployment -> utilities for model deployment in production
• Summary -> automatic ML experiments reports after each run
• Stand on the shoulders of giants -> MagicML is built on top of Dash, Pandas, Keras, scikit-learn, mlflow, and seaborn.

⚠️ Warnings

For the moment:

• The development of the project takes place in a private GitHub repository
• The project is at a very early stage -> Nothing is implemented in this published version
• The documentation is missing

The GitHub repository as well as a usable Python package will be available in the upcoming months when the project will be more advanced.

🤝 Community-driven

MagicML is foremost a community-driven project ! The project will be highly collaborative and everyone is welcome to the project ! 🤗

Don't hesitate to contact me if you want to know more or are interested in ! 😃

Stay tuned ! 🗓️