AutoGluon is an AutoML library for text, image, and tabular data. The easy-to-use and easy-to-extend framework focuses on automated ensembling of stack, deep learning, and spanning image, text, and tabular data. 

It’s an ideal product ML experts and beginners. Engineers can use AutoGluon to quickly prototype ML and deep learning solutions for raw data with shortcodes.

AutoGluon automatically leverages state-of-the-art techniques without expert coding knowledge. It uses automatic hyperparameter tuning, data processing, model selection, and architecture search. The framework easily improves models and data pipelines while helping you customize AutoGluon for a variety of use cases. Also, users can reduce dependencies among different sub-modules of AutoGluon via python3 -m pip install <submodule>.

Project Background

• • Library: AutoGluon
  • Author: Nick Erickson, Jonas Mueller, Hang Zhang, and Balaji Kamakoti
  • Initial Release: May 26, 2014
  • Type:  Works with deep learning and classical ML
  • License: Apache License 2.0
  • Contains: Python functions, datasets, submodules for for tabular, text, or image data
  • Language: Python
  • GitHub: /awslabs/autogluon
  • Runs On: Linux, Mac
  • GitHub Discussions: /awslabs/autogluon/discussions
  • Stackflow: AutoGluon

Applications

• Tabular prediction
• Object detection
• Image prediction
• Text prediction 
• Multimodal prediction
• Computer vision
• Natural language processing