Can't configure profile with AWS CLI for using AWS Built-in sagemaker algorithms about studio-lab-examples HOT 4 CLOSED

Hi! The use case you are describing is exactly why we have this example notebook:

• https://github.com/aws/studio-lab-examples/blob/main/connect-to-aws/Access_AWS_from_Studio_Lab.ipynb

Your net net is:
1/ Ensure you have proper training and authorization to use your AWS access and secret keys. If you are an AWS account admin, you can find this in your IAM console. If you are not, work with your admin to determine if this pattern is appropriate for you.

2/ If you are qualified to manage your AWS keys, create a new file called ~/.aws/credentials. There, copy and paste in your access and secret keys.

3/ Remove any cells you ran in a notebook to create or verify those files.

4/ Create a SageMaker execution role. The easiest way to do this is via the console - you can create a new SageMaker execution role when create a notebook instance or a Studio user profile. Once this is done, paste in the arn (Amazon Resource Name), for this execution role.

5/ Go forth and scale up on SageMaker! After that, once you are using the SageMaker Python SDK, you should be able to use all code-based features within SageMaker, such as training, hosting, tuning, autopilot, pipelines, etc.

from studio-lab-examples.

Hi @EmilyWebber. Might you also have an example that doesn't require AWS account information yet avoids "ValueError: Must setup local AWS configuration with a region supported by SageMaker" in the code below for "role"?

# create Hugging Face Model Class
huggingface_model = HuggingFaceModel(
	transformers_version='4.6.1',
	pytorch_version='1.7.1',
	py_version='py36',
	env=hub,
	role=role, 
)

Code source: https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english, where Deploy is Amazon SageMaker rather than Accelerated Inference — the latter currently returns an error with or without SageMaker

Since the SageMaker Studio Lab website emphasizes no costs and no need to sign up for an AWS account, a team and I are exploring whether to have students try. Thank you in advance.

from studio-lab-examples.

Hi @derekschan - thanks for the comment. This is actually expected behavior right now - Studio Lab defaults to not having access to any AWS API's unless explicitly granted. To date, that is solved only by the pattern mentioned above in this issue, explicitly installing AWS key permissions to utilize them.

Should that ever change in the future we will be sure to let you know! I'll mark your comment as a feature enhancement.

from studio-lab-examples.

Thank you, @EmilyWebber.

from studio-lab-examples.