Tune Gemini 2.5 models by using supervised fine-tuning not working with function calling modality

Google Cloud Build with AI Generative AI & Foundational Models
1

Hello.

Based on documentation trying to fine-tune model to improve extraction json from text. I created needed jsonl files and try to train model, but for both gemini-2.5-flash-lite and gemini-2.5-flash system return error

Dataset example 1 of 20 contains an unsupported modality [function call] in the "parts" field, which is not supported by this model. The supported modalities are: [image, PDF document, audio, video]

But based on documentation - Function calling reference  |  Generative AI on Vertex AI  |  Google Cloud Documentation - both model should support this. For gemini-2.0-flash-001 no such error.

Thanks

2

Hey, same issue here. Did you find a solution?