Opening model directory from s3 for sagemaker scheduled jobs

[ColinFerguson]

Hi all, thanks in advance for any help. I've used simpletransformers for several different projects but now I've come up against a new (to me) issue.

I am using AWS Sagemaker Studio, and have no problems training models and then using them for inference, e.g.

'''model = ClassificationModel('bert', 'path_to_model_directory', num_labels=2, use_cuda=True, args=model_args)'''

However, now I need to schedule a recurring job via sagemaker to run inference once per day. The sagemaker scheduler does not mount the local file system, so 'path_to_model_directory' effectively does not exist to a scheduled job.

The solution that makes the most sense to me would be to create a tar.gz file from the model directory, save that file to s3, and then somehow open it in my script (putting a tar.gz file in the call to ClassificationModel doesn't work). Does any one have any suggestions for how to do that, or another idea for how to access the model?

Thanks.