v0.4.0

Version 0.4.0

We are excited to present the AutoGluon-Cloud 0.4.0 release. This release brings up compatibility with AutoGluon 1.1.0, and minor fixes in TimeSeriesCloudPredictor API and documentation. As always, we encourage you to try this new release and share your feedback!

Updates and Improvements:

• Simplify TimeSeriesCloudPredictor API by @shchur in #105
• Update dependencies to align with AutoGluon 1.1.0 by @suzhoum in #110

Tutorial Enhancements:

• Minor fixes to the documentation by @shchur in #104

Explore more in our updated tutorial.

As always, we thank our community for their ongoing support and contributions. We look forward to your feedback on AutoGluon-Cloud 0.4.0!

v0.3.1

Version 0.3.1

We are excited to present the AutoGluon-Cloud 0.3.1 release. This is a critical bug fix release to include sagemaker and ray scripts for PyPI distribution

Explore more in our updated tutorial.

As always, we thank our community for their ongoing support and contributions. We look forward to your feedback on AutoGluon-Cloud 0.3.1!

v0.3.0

Version 0.3.0

We are excited to present the AutoGluon-Cloud 0.3.0 release. This version brings significant enhancements and features, including support for leaderboard API, batch inference improvements, and compatibility with AutoGluon 1.0.0. As always, we encourage you to try this new release and share your feedback!

AutoGluon-Cloud now aligns with AutoGluon version 1.0.0 with Python 3.11 support, ensuring a seamless and efficient experience for users.

Explore more in our updated tutorial.

This release includes 17 commits from 4 contributors.

Special

See the full commit change-log here: 0.2.0...0.3.0

Full Contributor List (ordered by # of commits):

• @yinweisu, @tonyhoo, @YiruMu, @Jwely

This release continues to support Python versions 3.8, 3.9, 3.10 and 3.11.

NEW: leaderboard API

• Support for Leaderboard API allowing more insights into model performance. @YiruMu (#94)

Updates and Improvements:

• Updated dependency versions to match AutoGluon 1.0.0 requirements. @tonyhoo (#97)
• Enhanced batch inference capabilities, including support for no-header scenarios. @yinweisu (#91)
• Support for extra arguments for real-time prediction, enhancing flexibility. @yinweisu (#78)
• Addition of predictor_path to to_local_predictor for better model management. @yinweisu (#88)

Tutorial Enhancements:

• Distributed training tutorial improvements, providing comprehensive guidance. @yinweisu (#87)
• General tutorial updates for better user guidance. @yinweisu (#133848f)

Infrastructure and Permissions:

• Improved permissions handling for enhanced security. @yinweisu (#86)
• Optional specification of cloud output paths for more control over data storage. @yinweisu (#85)
• Deployment options now include volume deployment for additional flexibility. @yinweisu (#84)

Continuous Integration and Model Persistence:

• Fixes and updates to continuous integration processes. @yinweisu (#81)
• Enabling model persistence for long-term utility. @yinweisu (#76)
• Support for pickle to facilitate model serialization. @yinweisu (#75)

Cluster Management and Distributed Training:

• Using latest AMI for cluster management, ensuring up-to-date infrastructure. @yinweisu (#73)
• Introduction of Tabular Distributed Training, paving the way for scalable model training. @yinweisu (#72)

Miscellaneous:

• Nightly release process improvements. @yinweisu (#74)
• Minor version update to 0.2.1 as part of ongoing maintenance. @yinweisu (#66)

As always, we thank our community for their ongoing support and contributions. We look forward to your feedback on AutoGluon-Cloud 0.3.0!

v0.2.0

Version 0.2.0

We're happy to announce the AutoGluon-Cloud 0.2.0 release. This release added support for TimeSeriesCloudPredictor.
Please give it a try and provide us feedback!

Currently, AutoGluon-Cloud supports training and deploying TabularPredictor, MultiModalPredictor and TimeSeriesPredictor with AutoGluon Deep Learning Containers version 0.6.0 and newer.
It is always recommended to use the latest version of the container as it has more features and are updated with security patches.

To learn more, check out our tutorial

This release contains 27 commits from 2 contributors.

See the full commit change-log here: 0.1.0...0.2.0

Full Contributor List (ordered by # of commits):

• @yinweisu, @gidler

This release supports Python versions 3.8, 3.9, and 3.10. Python 3.7 is no longer supported as of this release.

NEW: TimeSeriesCloudPredictor

We are happy to announce that you can train TimeSeriesPredictor with SageMaker backend in the cloud just like TabularPredictor and MultiModalPredictor now.
Checkout the quick example here!

• @yinweisu (#37, #38)

NEW Doc Site Style

We have updated the doc site to match the new style being used by the main AutoGluon site.
Check it out here

• @yinweisu (#62)
• Special thanks to @gidler for guiding the update.

General:

• Added Python 3.10 support and dropped python 3.7 support. @yinweisu (#29)
• Updated dependency ranges. @yinweisu (#59)

Improvements and Refactoring:

• Concurrency and space controls for large models. @yinweisu (#42)
• Extract Sagemaker backend. @yinweisu (#43)
• IAM role and policy utils. @yinweisu (#53)
• Add better warning message for to_local_predictor() call. @yinweisu (#64)

Bug Fixes/Doc Improvements:

• Tutorial fixes. @yinweisu @gidler (#33 #41 #55 #56 #57 #63)

CI:

• @yinweisu (#42, #45, #48)

Distributed Training:

Added components to support distributed training in the future. This feature is not available yet, but is being actively working on

• Ray interface. @yinweisu (#43)
• Ray AWS Cluster Config Generator. @yinweius (#47)
• Ray AWS Cluster Manager. @yinweisu (#49)
• Ray Job. @yinweisu (#51)

Miscellaneous

• Miscellaneous. @yinweisu (#31, #34, #36, #58)

v0.1.0

Version 0.1.0

We're happy to announce the AutoGluon-Cloud 0.1 release. 0.1 is the initial release of the Cloud module and enables you to train and deploy AutoGluon backed models on AWS SageMaker.
Don't wait to try it out and we are happy to any feedback!

Currently, AutoGluon-Cloud supports training and deploying TabularPredictor and MultiModalPredictor with AutoGluon Deep Learning Containers version 0.6.0 and newer.

To learn more, check out our tutorial

This release contains 50 commits from 7 contributors.

See the full commit change-log here: https://github.com/autogluon/autogluon/commits/0.6.1/cloud and here: https://github.com/autogluon/autogluon-cloud/commits/0.1.0

Full Contributor List (ordered by # of commits):

• @yinweisu, @gradientsky, @tonyhoo, @gidler, @giswaq, @sxjscience, @suzhoum

This version supports Python versions 3.7 to 3.9.