Azure ML Studio Quotas for Assisted Labeling and Training Object Detection Model

Question

In an Azure ML Studio labeling project, I have tried to enable Assisted Labeling. I get the error message "Error: There is insufficient quota to create a gpu compute target. You can request more quota and create a custom compute to enable ML assisted labeling."

I also get a similar error message when I try to train an object detection model, "“STANDARD_D2AS_V4 is not supported for image tasks. Please choose a VM type that is in the NC-family or the ND-family."

I requested and was granted "Standard NC Family Cluster Dedicated vCPUs/GPUs. These show up in my quota, but if I go to create a compute target and select GPU, I get a message saying: "You do not have enough quota for the following VM sizes." followed by a list of all the VMs that it says aren't in my quota including the NC family VMs.

And I still get the same two error messages saying I don't have enough quota for either Assisted Labeling or training an object detection model.

Does anyone know what I need to do to get these two services to work?

Thanks.

Answer 1

Hi there

I have the same quota issue when I just created my project. I contacted the support team and they solved issue for me. You should try it too.

Answer 2

Thanks Alexandre,

Do you have any info on how they solved this issue?

I opened a case a week or so ago. The initial support team thought they were the wrong team, and referred me to the Global Capacity team. The global capacity team has been slow to respond, most likely due to the Memorial day weekend.

Hopefully they will get back to me soon.

But I think this may be more of an issue with Azure ML Studio rather than the quota team.

Any thoughts?

Answer 3

@YutongTie-MSFT this continues unresolved. TrackingID#2205240040008434