I am trying to create an endpoint for an LLM in Vertex AI with a V100 gpu in the US-Central1 region. This will be my only endpoint, in any region. Here are my quotas and their limits:
- Committed NVIDIA V100 GPUs : 1
- NVIDIA V100 GPUs : 2
- Preemptible NVIDIA V100 GPUs : 1
- Managed Notebooks NVIDIA V100 GPUs per region : 1
- Custom model serving Nvidia V100 GPUs per region : 6
- Custom model training Nvidia V100 GPUs per region : 6
- Custom model training preemptible Nvidia V100 GPUs per region : 10
- Custom model serving Nvidia V100 GPUs per region : 6
- GPUs (all regions) : 6
When I try to create the endpoint, with a single V100 gpu, I get the error:
Error Messages: The following quotas are exceeded: CustomModelServingV100GPUsPerProjectPerRegion
I am using a Nvidia V100 GPU in a managed notebook, so I should have space for one more. There are other quotas I can't change, that are not connected to any region:
- Custom model serving Nvidia V100 GPUs per region (default) : 0
- Preemptible NVIDIA V100 GPUs (default) : 1
- Preemptible NVIDIA V100 GPUs (default) : Unlimited
- Committed NVIDIA V100 GPUs (default) : 0
- NVIDIA V100 GPUs (default) : 1
- NVIDIA V100 GPUs (default) : Unlimited
When I mouse over these quotas to change them, this message appears:
Edit is not allowed for this quota.
I have the Owner role for this project. Do I need to reach out to sales to try and get these changed? How can I do that when I only have the basic support plan? Is there another hidden quota somewhere? Do I just need to increase my quotas more?
LinkedIn
Twitter
