CODEWITHSUNDEEP
amazon-web-servicesboto3amazon-sagemaker
Jacques Thibodeau
Jacques Thibodeau

Reputation: 959

How to choose from a list of Regions for GPU instances in Amazon SageMaker?

I'm currently trying to do distributed GPU training with 2 instances of ml.p3.8xlarge and after 4 attempts I have not been able to start a training job with the spot instances since AWS did not have any available instances in my region.

How do I increase the number of regions I'm willing to choose from in SageMaker? At the moment I'm only using:

sess.boto_region_name = us-east-1 (sagemaker session region)

But I'm assuming if I allow SageMaker to choose from other regions, I will be able to start a training job with spot instances.

Thanks!

Upvotes: 0

Views: 709

Answers (1)

John Rotenstein
John Rotenstein

Reputation: 270274

Most AWS services are regional-based, meaning they run in a given region and do not spread beyond one region.

If you wish to run SageMaker in multiple regions, you would need to launch it separately in each region. So, you would only be 'choosing' one region when requesting SageMaker to perform some work.

Upvotes: 1

Related Questions