Highlights
- First-class foundation model support — run state-of-the-art forecasting models like Chronos-2 with real-time, serverless, or batch inference in a few lines of the familiar AutoGluon API. (tutorial)
Simplified setup — run a single command to provision every AWS resource needed for AutoGluon-Cloud. (tutorial) - Brand-new docs — new improved tutorials for tabular and time series.
fit_predictforTimeSeriesCloudPredictor— single-job train + batch inference, with optional direct S3 output.- Works with the latest AutoGluon 1.5.0 DLC images and Python 3.10–3.13.
Note
Predictors trained and endpoints deployed with earlier versions of autogluon.cloud are not compatible with v0.5.0. Please retrain your predictors and redeploy your endpoints from scratch.
New features
- Foundation models for time series forecasting — new
TimeSeriesFoundationModelwith real-time, batch, and serverless inference; JumpStart-compatible payload format. (#217, #220, #221, #228, #230, #234, #235) by @shchur, @melopeo agcloud setupCLI and CloudFormation templates — provision IAM roles, S3 buckets, and execution roles in one command; templates shipped in-package. (#173, #213, #222, #225, #226, #229, #233, #236) by @AnirudhDagar, @shchurfit_predictforTimeSeriesCloudPredictor— single-job train + batch inference with optional direct S3 output. (#214, #224) by @melopeo- Serverless inference endpoints — deploy with
instance_type="serverless". (#242) by @shchur - Create endpoint from existing endpoint name — reuse a deployed endpoint without re-deploying. (#246) by @shchur
cache_model_artifact— speed up repeated deploys. (#241) by @shchur- Support for
known_covariatesinTimeSeriesCloudPredictor.predict_real_time(#181) by @shchur - Tag resources created by AG-Cloud. (#249) by @shchur
Improvements & bugfixes
- Python 3.12 and 3.13 support; AutoGluon 1.5.0 DLC images by default. (#184, #185, #246) by @suzhoum, @shchur
- Made
rayan optional dependency. (#229) by @shchur - Bumped dependency version ranges. (#209) by @shchur
- Rewritten tutorials and API documentation. (#230, #237, #243, #244, #248) by @shchur, @melopeo
- Numerous bugfixes and polish for v0.5.0 features (#238, #247, #250) by @shchur
- CI, tooling, and infrastructure improvements: ruff + pre-commit,
build_docson PRs, latest AMIs, dependency bumps. (#175, #180, #210, #211, #212, #215, #216, #231, #232, #194, #198, #199, #200, #202) by @shchur, @suzhoum, @prateekdesai04, @melopeo
New Contributors
- @melopeo made their first contribution in #214
- @AnirudhDagar made their first contribution in #213
Full Changelog: v0.4.2...v0.5.0