Skip to article frontmatterSkip to article content
Site not loading correctly?

This may be due to an incorrect BASE_URL configuration. See the MyST Documentation for reference.

Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!

https://github.com/awslabs/autogluon

https://autogluon.mxnet.io/

AutoGluon automatically inferences the problem type. However, you are still able to specify the probelm type if AutoGluon fails to infer the right problem type.

  1. AutoGluon auto saves all models into the specified output directory during training. And the save models can be load back using the load method of the corresponding predictor.

  2. GPU support in AutoGluon is for image/text but not Tabular data currently. For more details, please refer to issue 262.

Questions

Can I choose a model to save and choose a model to load?

Customization

Search Space and Decorator

Search Algorithms

Customize User Objects

Customize Training Script

Hyperparameter Tuning

Use AutoGluon for Hyperparameter Optimization for MNIST Training in PyTorch

https://autogluon.mxnet.io/tutorials/nas/enas_mnist.html

References

https://futurumresearch.com/aws-releases-autogluon-an-innovative-open-source-tooling-for-automated-machine-learning/

https://www.amazon.science/amazons-autogluon-helps-developers-get-up-and-running-with-state-of-the-art-deep-learning-models-with-just-a-few-lines-of-code

https://towardsdatascience.com/autogluon-deep-learning-automl-5cdb4e2388ec

https://venturebeat.com/2020/01/09/amazons-autogluon-produces-ai-models-with-as-little-as-three-lines-of-code/

https://autogluon.mxnet.io/tutorials/course/distributed.html

Jan
Tips on Apache Ray
Jan
Tips on Feature Engineering for Machine Learning