Open mathcass opened 3 weeks ago
I also ran into this when a deploy failed unexpectedly. Looks like there was a request to make it optional that wasn't addressed before merge - https://github.com/aws/sagemaker-python-sdk/pull/4880#discussion_r1784907626
Edit: mlflow actually puts us over the lambda max deployment size so I'll need to avoid this version entirely until this is made optional
https://github.com/aws/sagemaker-python-sdk/pull/4903 open for review
Describe the bug The latest update to the SageMaker SDK introduces sagemaker-mlflow which then brings in a constellation of dependencies that users might not need. This is a minor change (https://github.com/aws/sagemaker-python-sdk/compare/v2.232.1...v2.232.2) that leads to installing 30 new dependencies.
To reproduce Install the latest update,
pip install sagemaker==2.232.2
.Expected behavior I would have liked if this release had been included via Python "extras" where you can specify an additional set of installation requirements. For example, rather than installing
sagemaker-mlflow
automatically, only install it if someone specifies `pip install "sagemaker[mlflow]>=2.232.2".Screenshots or logs If applicable, add screenshots or logs to help explain your problem.
System information A description of your system. Please provide:
Additional context The package could be more usable and leaner if it implemented this as a Python "extra".