[prateekdesai04]

Issue #, if available:

Description of changes:

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[mohamedzeidan2021]

The CI failures on the latest run (37870479052, head 5f45e1f6) are not caused by this change.

integ-tests — 4 failures, all in unrelated workflow tests

FAILED tests/integ/sagemaker/workflow/test_step_decorator.py::test_step_level_serialization
FAILED tests/integ/sagemaker/workflow/test_step_decorator.py::test_step_decorator_with_execution_dependencies
FAILED tests/integ/sagemaker/workflow/test_step_decorator.py::test_step_decorator_with_data_dependencies
FAILED tests/integ/sagemaker/workflow/test_selective_execution.py::test_selective_execution_among_pure_function_steps
= 4 failed, 531 passed, 162 skipped, 3 xfailed, 14 rerun in 3776.24s (1:02:56) =

All four exercise @step-decorator pipeline serialization and execution against a per-run dummy remote-function container, failing inside create_and_execute_pipeline. This PR changes exactly one file — src/sagemaker/image_uri_config/autogluon.json, adding the 1.4.0 and 1.5.0 version keys and their aliases — so there is no code path from it to step-decorator serialization.

The full integ suite runs here only because the CodeBuild filter matches src/sagemaker/image_uri_config/*.json, so any image-config PR picks up whatever is currently flaky in the suite. These tests run under -n 120 with --reruns 1 --reruns-delay 15, and the summary shows 14 reruns occurred.

codestyle-doc-tests — pre-existing lint drift on master-v2

13 files are flagged with F401 ... imported but unused and one E501 (121 > 120), in files such as src/sagemaker/amazon/amazon_estimator.py and src/sagemaker/apiutils/_base_types.py. The file touched by this PR does not appear in that output. The same job fails on other current master-v2 PRs (e.g. #6396) with unrelated diffs.

What does pass: unit-tests on all four interpreters (py39, py310, py311, py312), codecov/patch at 97.41%, collab-check, and wait-for-approval. The earlier py39-only unit-test failure on the previous head is gone. tests/unit/sagemaker/image_uris/test_autogluon.py is data-driven — it iterates whatever versions the JSON contains — so the two new versions are covered without a test change.

A rerun of integ-tests should be enough to confirm the four workflow failures are environmental.

---

Edited to correct the file list: an earlier version of this comment also referenced a KNOWN_MISSING_COMBINATIONS entry in tests/unit/sagemaker/image_uris/test_djl.py. That belongs to a different PR; this one touches only autogluon.json.