How to run a batch transformation with custom inference code in sagemaker pipeline?

Question

based on the docs here , https://sagemaker-examples.readthedocs.io/en/latest/sagemaker-pipelines/tabular/abalone_build_train_deploy/sagemaker-pipelines-preprocess-train-evaluate-batch-transform.html#Define-a-Transform-Step-to-Perform-Batch-Transformation, after creating a model,  the example shows a batch transform job is run. how to include any custom inference script that is required to get an inference from the model.

AWS-User-5210721 · Answer

The inference code plus model artifacts information is defined with SageMaker Model. In SageMaker Model, you have the option of which model artifact to use plus which inference script to use, usually defined as `entry_point` in the model object.

To be more specific, after training, you need a step to create/register the model as described [here](https://sagemaker-examples.readthedocs.io/en/latest/sagemaker-pipelines/tabular/abalone_build_train_deploy/sagemaker-pipelines-preprocess-train-evaluate-batch-transform.html#Define-a-Create-Model-Step-to-Create-a-Model). During this process, you could choose appropriate model artifacts with inference scripts.

Then this model could be defined to be used in a transform job.

How to run a batch transformation with custom inference code in sagemaker pipeline?

相關內容