Its been 2 days i am trying to create a Vertex AI workspace instance with Nvidia T4 GPU attached. But no luck. I have checked the quotas, but still no success. Without the nvidia GPU T4 selected i was able to create a instance, but that does not help for model training. Not sure if i am doing something wrong, or this is simply google not having enough nvidia t4 resources to attach.
Hi @sandeepnjk,
Welcome to the Google Cloud Community!
It looks like you’re running into an issue where your Vertex AI instance with a T4 GPU won’t launch even though your quota is fine. You can create one without the GPU, but that doesn’t suit your training needs likely because your selected zone is out of T4 resources.
Here are the potential ways that might help with your use case:
- Try Different GPU Regions and Zones: T4 GPU availability can vary widely, not just between regions but also between individual zones within the same region. If you’re having trouble in ‘us-central1’, it’s worth trying other regions like ‘us-east4’, ‘us-west1’, or ‘europe-west4’. And within any chosen region, you might get better results by manually selecting specific zones (e.g. ‘us-central1-a’ versus ‘us-central1-f’) rather than letting the system auto assign one.
- Attempt During Off-Peak Hours: You can try creating your instance during non-business hours like late at night or early in the morning in the time zone of your selected region. This might increase your chances of accessing GPU resources when demand is lower.
- Re-verify Quotas: To verify your quotas, go to your IAM & Admin > Quotas page in the Google Cloud Console. Filter by the Compute Engine API, then review your quotas for NVIDIA T4 GPUs, GPUs (all regions), and CPUs. Make sure none of your quotas are set to zero or fully consumed. If any are too low, you can request an increase directly through your console.
Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.