How do I generate inferences locally from a model fit in Sagemaker?

Question

I built a custom container on Sagemaker to allow me to tune a Catboost model.

I then fit the model with the best hyperparameters (as if I were to deploy on Sagemaker).

I downloaded the tar.gz file onto my machine.

I can read the file:

import tarfile

file = tarfile.open('model.tar.gz')

I can extract from the file:

file.extractall('./output')

I now have a file named catboost_model.dump, but I am unsure where to go from here.

Is it possible load in this .dump file to generate inferences?

Answer 1

You'd want to load your model artifact into Catboost, using load_model() (probably). or you can use the SageMaker Python SDK to run an inference endpoint locally. You need to have Docker installed to do this