adventurous-ability-21671
08/06/2026, 12:29 PMtraining, processing, batch transform, hyperparameter tuning, and Inference Recommender. You supply the boto3 config and the connector handles submission, polling, and cancellation.
github.com/flyteorg/flytekit/pull/3456
• #3459: Pythonic mode for training and processing, so you can write a plain @task and have it run inside the SageMaker container instead of authoring a config blob. Modelled on the EMR Serverless plugin's dual-mode approach. Stacked on #3456, currently draft.
github.com/flyteorg/flytekit/pull/3459
Both are covered by unit tests and have been exercised against real SageMaker jobs. Rationale and design notes are in the PR descriptions.
Happy to split, rework, or rebase these however suits the maintainers.adventurous-ability-21671
08/06/2026, 12:31 PMglamorous-carpet-83516
08/06/2026, 4:31 PMadventurous-ability-21671
08/06/2026, 4:48 PMglamorous-carpet-83516
08/06/2026, 5:22 PMadventurous-ability-21671
08/06/2026, 5:30 PMadventurous-ability-21671
08/10/2026, 5:28 PMglamorous-carpet-83516
08/10/2026, 5:29 PMglamorous-carpet-83516
08/10/2026, 5:54 PMadventurous-ability-21671
08/10/2026, 6:10 PMflytekit-greatexpectations, which fails on an external 404 unrelated to these changes.
github.com/flyteorg/flytekit/pull/3459glamorous-carpet-83516
08/10/2026, 7:05 PMadventurous-ability-21671
08/10/2026, 7:31 PMglamorous-carpet-83516
08/10/2026, 7:33 PMadventurous-ability-21671
08/10/2026, 7:37 PMadventurous-ability-21671
08/11/2026, 10:41 AMglamorous-carpet-83516
08/11/2026, 4:46 PM