Tune Custom Models

Tutorials to hyperparameter-tune any custom models or Python code.

Search Space and Decoratorcore.html

Using AutoGluon’s Core APIs to hyperparameter-tune any model/code by making existing objects/training-functions searchable.

Search Algorithmsalgorithm.html

How to use AutoGluon’s built-in hyperparameter search algorithms, including early-stopping strategies.

Searchable Objectsobject.html

Tune the hyperparameters of custom objects such as your own: neural network, optimizer, dataset, etc.

Tune Training Scriptsscript.html

Tune the argument values (hyperparameters) of arbitrary Python scripts using AutoGluon.

Distributed Searchdistributed.html

Easily distribute the hyperparameter search across multiple machines to improve efficiency.

Example: Tune a Multi-Layer Perceptronmlp.html

Complete example of using AutoGluon’s state-of-the-art hyperparameter optimization to tune a basic MLP model.

Fair Bayesian Optimizationfairbo.html

Tune models under fairness constraints using AutoGluon’s constrained Bayesian optimization.