Model deployment with internal data

Question

Is it possible to deploy a model to an endpoint along with a large amount of data for it to use? (~50GB), this can be though of as an internal database for my model to use.
I only require such functionality temporary, since I am trying to avoid using database services (e.g DynamoDB etc) for the moment.

Thanks

Answer

Your best bet is to put the data in S3 if possible.  S3 integrates directly with SageMaker so it should be seamless.

Answer

Yes the data is stored in S3. Do you mean I can simply download the data from S3 bucket using boto3 (say in the function the loads the model)?

Answer

Yes.  Here is a sample snippet of code

import boto3

s3 = boto3.client('s3')

s3.download_file('BUCKET_NAME', 'OBJECT_NAME', 'FILE_NAME')

s3 = boto3.client('s3')

with open('FILE_NAME', 'wb') as f:
    s3.download_fileobj('BUCKET_NAME', 'OBJECT_NAME', f)

Model deployment with internal data

Conteúdo relevante