Hello everyone,
I’m trying to use Imagen on Vertex AI for image generation, but I’ve run into a confusing issue and am hoping someone in the community can offer some insight.
Goal:
I want to generate an image using the vertexai Python library, calling the imagen-3.0-fast-generate model in the us-central1 region.
The Problem:
On my Google Cloud Console “Quotas” page, I can clearly see that I have a quota of “20 requests per minute” for the Vertex AI API with the following dimensions:
- Service:
Vertex AI API - Region:
us-central1 - Model (
base_model):imagen-3.0-fast-generate - Quota: 20 / minute
However, when my Python script calls the API with this exact configuration, it doesn’t return a quota error (429), but instead returns a 404 Not Found error:
404 Publisher Model `projects/(PII Removed by User)/locations/us-central1/publishers/google/models/imagen-3.0-fast-generate` not found.
This seems to be a contradiction: the quota exists, but the API claims the model can’t be found.
What I’ve Tried So Far:
- Double-checked that the Project ID, model name, and region string are all correct.
- Confirmed the Vertex AI API is enabled and a valid billing account is linked to the project.
- To isolate the problem, I created a brand new Google Cloud project and repeated the entire setup process from scratch, but the exact same
404error occurs. - Created a new service account in the new project and granted it the “Vertex AI User” role.
- Confirmed the
GOOGLE_APPLICATION_CREDENTIALSenvironment variable is set correctly. - Tried other model/region combinations that also showed a valid quota, but they all resulted in similar errors.
My Questions:
- Has anyone else experienced this kind of discrepancy between the Quota page and the actual API response for Vertex AI models?
- Is it possible this is a known issue or a temporary problem on the Google Cloud side?
- Is there a step I might be missing to get the quota properly recognized by the Vertex AI API?
Any help or suggestions would be greatly appreciated. Thank you!