The command
sage deploy pytorch text-classification arn:aws:iam::xxxxxxxxxxxx:role/service-role/AmazonSageMaker-ExecutionRole-xxxxxxxxxxxx--model-path s3://datalabs-data/sagemaker_test/dummy.pt.tar.gz --endpoint-name test --instance-type ml.t2.medium --entry-point sage/example_entrypoints/pytorch_dummy_entrypoint.py
(with arn:aws:iam::xxxxxxxxxxxx my role number)
works neatly for me now!
It does make me think is quite a faff to write all arguments. I really like the way Matt's remotepy (https://github.com/ivyleavedtoadflax/remote.py/blob/main/remotepy/config.py) uses a config file to store variables which repeat regulary, in this case, it could be the arn role but even the s3 bucket or instance type
Would it be much effort to add this in? I think it would make the tool more user friendly
The command
(with
arn:aws:iam::xxxxxxxxxxxxmy role number)works neatly for me now!
It does make me think is quite a faff to write all arguments. I really like the way Matt's remotepy (https://github.com/ivyleavedtoadflax/remote.py/blob/main/remotepy/config.py) uses a config file to store variables which repeat regulary, in this case, it could be the arn role but even the s3 bucket or instance type
Would it be much effort to add this in? I think it would make the tool more user friendly