Error in creating GCP deep learning VM image

1
votes

I am getting the following error while trying to create GCP deep learning VM instance using my free $300 credits. I have tried with multiple server locations (US-west, US-central, etc.)

tensorflow-1 has resource warnings tensorflow-1-vm: {"ResourceType":"compute.v1.instance","ResourceErrorCode":"403","ResourceErrorMessage":{"code":403,"errors":[{"domain":"usageLimits","message":"Quota 'GPUS_ALL_REGIONS' exceeded. Limit: 0.0 globally.","reason":"quotaExceeded"}],"message":"Quota 'GPUS_ALL_REGIONS' exceeded. Limit: 0.0 globally.","statusMessage":"Forbidden","requestPath":"https://www.googleapis.com/compute/v1/projects/dltest-232312/zones/us-central1-a/instances","httpMethod":"POST"}}

2
google-cloud-platformgoogle-compute-engine
You will need to upgrade your account first. The free trial does NOT cover all services and configurations. The free trial has a GPU quota of 0. – John Hanley
Thanks a lot. After upgrading, we also have to request for a quota upgrade if GPU quota is set to zero. – ravvv

2 Answers

2
votes
  1. Upgrade your account.
  2. Increase the global gpu quota.
  3. Increase the quota in the specific region you want.
1
votes

As mentioned in the Always Free usage limits docs and by John Hanley, GPUs are not included when using Free Trial:

GPUs and TPUs are not included in the Always Free offer. You will always be charged for GPUs and TPUS that you add to instances.

In case you want to use this kind of resource, you have to upgrade your account and check your existing quotas to validate that you have enough GPUs available in your project.

Additionally, it is important to consider that when you request a GPU quota, it is must required to request the quota for the specific GPU model you want to create per region, as well as an additional global quota for the total GPUs in all the zones.

Here you can find the pricing details that you need to consider when using this components.