GCP 错误: 超出配额‘ GPUS_ALL_REGION’。全局限制: 0.0

我已经增加了我的 NVIDIA K80配额，并收到了请求批准。为什么我仍然不能使用与 GPU 虚拟机？

标题中显示的错误。

72317 次浏览
小开

Follow the instructions at GCE Quota Increase.

Remember to select correct filters. If you're not sure or you don't see GPUs(all regions) (under the name "Compute Engine API", see the image) in the list, try to select all for those filters, and scroll through the list to look for it. It took me some time to realize this. In my case, it is listed in 601 - 716th items, almost at the last of the list.

Just select it and edit the quota according to you needs. GPUs(all regions)

小开

You musty file a request that will include GPUS_ALL_REGIONS. New projects have a global GPU quota, which limits the total number of GPUs you can create in any supported zone. When you request a GPU quota, you must request a quota for the GPU models that you want to create in each region, and an additional global quota for the total number of GPUs of all types in all zones. Please see this link[1] for more information.

[1] https://cloud.google.com/compute/docs/gpus/#restrictions

小开
最佳答案

Increasing the region-specific GPU quota won't work. GPUs(all regions) have to be increased from 0. Steps as of December 2018 are below:

  1. From top left burger menu, select, IAM & admin -> Quotas. enter image description here
  2. In filter type, select metric to be GPUs(all regions) and Location as Global enter image description here
  3. Click edit quotas and new quota limit to 1 or more enter image description here
小开

I just came across this error and none of these directions were working for me (I think the GCP website has been updated since these answers).

In the filter at the top of the quotas page, I typed in Service: Compute Engine API and Limit Name: GPUs (all regions) and changed the limit from 0 to 1 or more. Hope this helps someone!

enter image description here

小开

The provided answers did not work for me. After searching for "GPUs (all regions)", there was no checkbox next to the filtered results. Clicking on the results and then clicking "edit quotas" as suggested did nothing.

Instead, here is the ridiculous process I had to go through:

  1. As with other answers, go to the Quotas page and filter for GPUs (all regions).
  2. Click All Quotas in the Details field of the result.
  3. A new table appears with only a single entry. The checkbox is grayed out.
  4. Instead of selecting and editing Quotas, click View hierarchy in the last field of the entry.
  5. A new table should appear with all of your projects.
  6. Finally, the checkboxes are not grayed out. Select the appropriate project.
  7. NOW you can click Edit Quotas.

Google. WTF mate? Now, according to one answer you should do the same for the regions you want to use.

小开

Also, GCP error message does not make sense. For example, it kept on showing me the above error message even after I got the global quota increased. Later I found out from this documentation:- https://cloud.google.com/compute/docs/gpus/#restrictions

.. that the above GPU configuration is not supported with the particular Intel Skylake or later CPU platforms. Error message never talks about this incompatibility issue ( under the notification). It still says:

    Quota 'GPUS_ALL_REGIONS' exceeded. Limit: 0.0 globally

folks@GCP should improve the error messaging.

小开

The top rated answer did not work out for me.

Here is what I did to get increase the global GPU quotas as of January 2021:

  1. Go to https://console.cloud.google.com/iam-admin/quotas
  2. Filter by GPUs (All Regions)

enter image description here

  1. Click on "Edit Quotas"

enter image description here

  1. Requested limit 1 in the upcoming dialogue

enter image description here

  1. Hit "Next" button

  2. Filled the next form and hit "Submit Request" button

enter image description here

After these steps you have requested access to GPUs and should receive on your email account an email from Google Compute Engine Quota Support with either approval or denial of your quota increase request.


提交答案