Batch prediction support for tuned Gemini models

Google Cloud Build with AI Custom ML & MLOps
, ,
1

I am trying to do batch prediction using a fine-tuned Gemini (gemini-1.5-flash-002) in Vertex AI. I am following this tutorial:

https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/batch-prediction-gemini

Is it supported? Because I always get various error messages (Model not found, or it tries to find it in Model Garden, or “Use publisher model reference instead of user uploaded model for prebuilt foundation models.”) depending on how I specify the model name (projects/012345678901/locations/europe-west3/models/0123456789012345678 or just 0123456789012345678)

2

Hi @marton-avrios ,

We hear you and I am happy to share that we now support fine-tuned Gemini models with Batch Prediction on Vertex AI.

Check out the documentation here.

Best