sagemaker.image_uri.retrieve "container_version" parameter does not work as expected

[arjkesh]

Describe the bug This API doc https://sagemaker.readthedocs.io/en/stable/api/utility/image_uris.html#sagemaker.image_uris.retrieve states that there is a parameter called "container_version" that can be used to retrieve a particular DLC. However, with certain versions, it creates an invalid image URI.

To reproduce A clear, step-by-step set of instructions to reproduce the bug.

>>> import sagemaker
>>> sagemaker.image_uris.retrieve('pytorch', 'us-west-2', version='1.12', image_scope="inference", instance_type='ml.p3.8xlarge', container_version='1.1')

'763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-inference:1.12-gpu-py38-1.1'

The above image URI does not exist, even though there is a 1.1 version of the container. Image URIs for v1.1 are included here:

https://github.com/aws/deep-learning-containers/releases/tag/v1.1-pt-sagemaker-1.12.1-inf-py38

Expected behavior Expected behavior would be for the function to return a valid image URI with container version for the given framework from here - https://github.com/aws/deep-learning-containers/releases/tag/v1.1-pt-sagemaker-1.12.1-inf-py38

Additional context It make not make sense to have this parameter, as there are likely only a few use cases where a user would specify the container version and not just reference the image URI directly. This could be useful for major version pinnings, but unfortunately as constructed, it would not retrieve the correct image URI.