Accessing SageMaker model after registering it

Question

I can easily create and version a model using SM model registry: https://docs.aws.amazon.com/sagemaker/latest/dg/model-registry-version.html

After registering the model, it is also possible to access the metadata: https://docs.aws.amazon.com/sagemaker/latest/dg/model-registry-details.html but I do not understand how to programmatically access the internal of the model.

For example, I would like to manually inspect the weights of the last layer of my neural network. A model that I registered last week but that I do not have available on my machine anymore.

Accepted Answer

You can use [describe_model_package](https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/sagemaker.html#SageMaker.Client.describe_model_package) to get the description of the specified model package.

In the response, in `ModelDataUrl ` parameter you'll find the Amazon S3 path where the model artifacts, which result from model training, are stored. This path must point to a single gzip compressed tar archive (.tar.gz suffix).

You can download this file and unzip it. Inside will be the model files, as output of the training.

You can check [this example notebook](https://github.com/aws/amazon-sagemaker-examples/blob/main/sagemaker-python-sdk/scikit_learn_model_registry_batch_transform/scikit_learn_model_registry_batch_transform.ipynb) in which we show how to use Amazon SageMaker to train a Scikit-learn Random Forest model, register it in Model Registry, and run a Batch Transform Job.

Accessing SageMaker model after registering it

Contenus pertinents