Wrong Enum Values for Transformer `strategy`

[jaredleekatzman]

What did you find confusing? Please describe.
When trying to build a transformer from a TensorFlowModel, e.g.

transformer = model.transformer(
    instance_count=1, instance_type='ml.m5.large', assemble_with='Line',
    accept='application/json', output_path=output_uri, strategy='SingleRecord'
)

The doc string gave wrong information for the parameter strategy. The Doc says the values must be of either [MULTI_RECORD, SINGLE_RECORD]. I tried them but then the API throws an error.

When you look at the Transformer API doc string the value of the strings mus be of either [MultiRecord, SingleRecord], and there doesn't seem to be a value translation.

Describe how documentation can be improved
The right values of the strategy parameter should be mentioned correctly across the sagemaker-python-sdk documentation. There seems to be 7-8 occurrses across the GitHub Repo link:

• entities.py
• pipeline.py
• algorithm.py
• model.py
• tensorflow/estimator.py
• airflow.py
• estimator.py
• session.py

[knakad]

Oh no! Thanks for bringing this to our attention!

It looks like there were already unit tests and integ tests validation the functionality using the properly cased version of the strings (SingleRecord/MultiRecord). However, as you pointed out, the docstrings were wrong (likely because the backend uses the screaming snake case version, which is also reflected in our local mode code).

I submitted and merged PR to fix this: #1340
Note that the readthedocs pages will only be updated on the next release, which is the morning after the commit is merged (on a non_Friday weekday). In this case, tomorrow morning, barring unforeseen circumstances.

Thanks again for letting us know!