Ruinong Tian

``` =========================== short test summary info ============================ FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[keras.test.Dockerfile-required_packages0] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[scipy.test.Dockerfile-required_packages4] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[sagemaker-studio-analytics-extension.test.Dockerfile-required_packages20] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[pandas.test.Dockerfile-required_packages7] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[sm-python-sdk.test.Dockerfile-required_packages8] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[pytorch.examples.Dockerfile-required_packages9] ====== 6 failed, 7 passed, 13 skipped, 6 warnings in 18020.18s...

``` =========================== short test summary info ============================ FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[keras.test.Dockerfile-required_packages0] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[scipy.test.Dockerfile-required_packages4] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[sagemaker-studio-analytics-extension.test.Dockerfile-required_packages20] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[pandas.test.Dockerfile-required_packages7] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[sm-python-sdk.test.Dockerfile-required_packages8] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[pytorch.examples.Dockerfile-required_packages9] ====== 6 failed, 7 passed, 13 skipped, 6 warnings in 18038.08s...

From 1.12.0, seeing error ``` Failed: DID NOT WARN. No warnings of type (, , ) were emitted. ``` Seems to be test case issue

How much is image size increased comparing to 1.10.0 or 1.9.1?

``` FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[keras.test.Dockerfile-required_packages0] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[pandas.test.Dockerfile-required_packages8] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[sagemaker-studio-analytics-extension.test.Dockerfile-required_packages21] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[gpu-dependencies.test.Dockerfile-required_packages31] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[sm-python-sdk.test.Dockerfile-required_packages9] FAILED test/test_dockerfile_based_harness.py::test_dockerfiles_for_gpu[pytorch.examples.Dockerfile-required_packages10] ``` (Rerun locally) * keras - Single test fails with `AttributeError: 'CustomLayer' object has no attribute '_build_input_shape'`....

This issue should be reported to AWS SageMaker Unified Studio service, instead of SMD image.

Hi, may I ask why is this change required? What's the motivation?