sumit-t
commented
4 years ago Thank you for your feedback.
You can find the default service quotas (aka limits) for any AWS Service including Amazon SageMaker on AWS Service Quotas page. Follow the instructions on the page to navigate to SageMaker default quotas page. You can also discover this documentation from Amazon SageMaker Developer Guide. Visit the Supported Regions and Quotas page for details.
Regarding instance types supported for each algorithm, you can find this information for each built-in algorithm in SageMaker Developer Guide.
Hope this is helpful.
Lately running into too many Sagemaker issues. Is there any unambiguous documentation on Sagemakers Instances? I could glean the following from different sources:
Sagemaker Instances, Sagemaker being a managed service, have nothing to do with EC2 instances.
Unlike EC2 console, Sagemaker console has no option to view limits or increase limits. One has to go directly to the support page and select instance limit increase and pick Sagemaker as the service. But must one run into Resource Limit Exceeded errors before to find out instance limits?
Sagemaler notebook instances are different from the training, deployment instances.
Not every algorithm in Sagemaker supports every(training/deployment) instance type. It doesn't make sense for XGboost to be GPU accelerated.
I wonder if all of this is documented anywhere at all, in simple language. It'd be immensely helpful to have it included in a one-page document that shown up prominently to users of Sagemaker, at least new ones.