Databricks 7.3ML vs 7.0ML

Rohit Pillai 1 Microsoft Employee

Hi,

I'm trying to run a distributed training job using Horovod and Pytorch on Azure Databricks using 2 nodes with 2 gpus each. When I run my code on version 7.0ML, the code sees each node as having 2 GPUs which is expected behavior. However, on 7.3ML, the code sees a node as having only 1 GPU. Why is this happening? Is there someway to change this to show each node's original configuration with multiple GPUs?

PRADEEPCHEEKATLA-MSFT 85,586 Reputation points Microsoft Employee

2020-11-23T11:19:15.67+00:00

Hello @Rohit Pillai ,

Welcome to the Microsoft Q&A platform.

Thanks for bringing this to our attention. I’m working with the product team and get back to you when I have more information.
PRADEEPCHEEKATLA-MSFT 85,586 Reputation points Microsoft Employee

2020-11-24T11:48:38.477+00:00

Hello @Rohit Pillai ,

We are still waiting for a response from the product team. I will let you know once I heard back from the product team.

Stay Tuned!
PRADEEPCHEEKATLA-MSFT 85,586 Reputation points Microsoft Employee

2020-12-08T10:52:23.833+00:00

Hello @Rohit Pillai ,

Unfortunately I have not received much input on this from the internal team.

For a deeper investigation and immediate assistance on this issue, if you have a support plan you may file a support ticket, else could you please send an email to AzCommunity@Microsoft.com with the below details, so that we can create a one-time-free support ticket for you to work closely on this matter.

Subject of the email: <ATTN: PRADEEP - Databricks 7.3ML vs 7.0ML>

Thread URL: Databricks 7.3ML vs 7.0ML

Subscription ID:

Please let me know here once you have done the same.

Share via

Databricks 7.3ML vs 7.0ML