Skip to content

Failure to use correct version of Sklearn #170

@mcoplan-hash

Description

@mcoplan-hash

When deploying a mme endpoint, the sklearn version does not seem to use the correct version. I specify the

IMAGE_NAME = amazonaws.com/sagemaker-scikit-learn:1.0-1-cpu-py3
sklearn_estimator = SKLearnModel(
    source_dir= source_dir
    model_data= model_data,
    entry_point="inference.py",  # all models in MultiDataModel need to have inference.py
    role=ROLE,
    image_uri=IMAGE_NAME,
    env={
        "SAGEMAKER_SUBMIT_DIRECTORY": "/opt/ml/model/"
        # 'SAGEMAKER_PROGRAM':
    },
    sagemaker_session=sagemaker_session,
)

I would expect this to use sklearn==1.0.1, but when i check logs it seems to be using sklearn==0.23.2

Any ideas that would cause this?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions