search

aws / aws-step-functions-data-science-sdk-python

Step Functions Data Science SDK for building machine learning (ML) workflows and pipelines on AWS
Apache License 2.0
285 stars 87 forks source link

Unable set ModelClientConfig in TransformerStep #171

Open ysgit opened 2 years ago

ysgit commented 2 years ago

Not sure if this qualifies as a bug or a feature request...

When creating a transform job one can pass a ModelClientConfig containing the invocation timeout and number of retries. See https://docs.aws.amazon.com/sagemaker/latest/APIReference/API_CreateTransformJob.html#sagemaker-CreateTransformJob-request-ModelClientConfig

This is implemented in the sagemaker python sdk, you can set model_client_config when calling Transformer.transform.

There is currently no option to set this in the TransformerStep

It should be relatively easy to implement by adding model_client_config as a param to the TransformerStep and setting parameters['ModelClientConfig'] to the value passed there.

This is :bug: Bug Report

ca-nguyen commented 2 years ago

Hi @ysgit

Thank you reporting this!

We are currently working on exposing the TransformStep parameters in the step constructor. With this change, any parameters documented in CreateTransformJob will be dynamically configurable.

Current PR for this change: #157

shivlaks commented 2 years ago

@ca-nguyen - PR #157 is currently classified as a feature. If it is a bug, let's make sure we add the template information captured for posterity (repro steps, expected behaviour, observed behaviour, etc).

The template information really helps in clarifying the problem and resolves ambiguity :)

RGirish commented 2 years ago

Any updates on this?