1

I'm running JupyterLab via. AWS SageMaker.

I've been taking AWS certifications, but this is my first time actively using AWS.

Update:

I have changed the Notebook instance type to ml.g4dn.xlarge, a GPU.

Will run and see what happens.

How do I change the instance types of EC2 to GPU?

In Google Colab, e.g., you can select which hardware accelerator to use, one of which being GPU.

Error:

MisconfigurationException: You requested GPUs: [0]
 But your machine only has: []

SageMaker environment: enter image description here

List of Kernels: I'm on conda_python3. enter image description here

Pranav Hosangadi
  • 23,755
  • 7
  • 44
  • 70

1 Answers1

1

First thing is to determine if you are you using a SageMaker Studio or SageMaker notebook instance.

Since you are using SageMaker notebooks, you first need to go back to SageMaker console, select the correct notebook, and stop it.

enter image description here

Once the notebooks is stopped, you can edit the configuration and select an instance that has GPU. You can find the list of all instances here.

enter image description here

In Sagemaker studio, you can select a PyTorch GPU optimized kernel

enter image description here

and then select an instance that has GPU.

enter image description here

Amir Imani
  • 3,118
  • 2
  • 22
  • 24
  • Will try this all now and get back to you. Ty @AmirImani –  Aug 24 '21 at 07:54
  • On AWS Console, top navigation states: `Amazon SageMaker` > `Notebook instances`. Also, `URL` contains "notebook": `https://.notebook.eu-central-1.sagemaker.aws/lab`. –  Aug 24 '21 at 07:58
  • I've attached a screenshot, to my post, of my `SageMaker` enviroment and list of `kernels`. –  Aug 24 '21 at 08:03
  • 1
    so you are using SageMaker notebook instances. unlike studio, you can't change the underlying instance in runtime. you need to stop the instance and change the instance type to one with GPU and restart it. I will update my answer as well – Amir Imani Aug 24 '21 at 16:10
  • 1
    Yes, I learnt this. I selected GPU instance types with `ResourceLimitExceeded` errors. So, I have made a service requested to AWS. I'll keep posted. Thanks @AmirImani –  Aug 25 '21 at 08:11