Cannot use custom model with Google Vertex Chat Model

Questions
1

I have fine-tuned a model in Google Vertex AI. The model is called quotes-test2. It was fine-tuned using gemini-1.5-flash as the base.

The message I’m getting is ‘Unsupported Model’.

If I change the model to gemini-1.5-flash in my node, it works fine.

How can I use a fine-tuned model in n8n? I’m open to using other platforms other than Vertex if any work.

Information on your n8n setup

  • n8n version: 1.80.3
  • Running n8n via n8n cloud
2

I am not confident as I have not personally encountered/tested, but this could be because fine tuned models from google do not support json mode.

Tuned models
Tuned models have the following limitations:

  • The input limit of a tuned Gemini 1.5 Flash model is 40,000 characters.
  • JSON mode is not supported with tuned models.
  • Only text input is supported.
3

Thanks for the reply @ThinkBot. Do you have any ideas on how to use any custom fine-tuned models with n8n?

This is currently a major limitation for me using n8n because I need better training options other than RAG and prompt engineering.

4

I know you can for sure with just using HTTP requests or some base AI nodes, I am unsure if its possible with the AI Agent node.

I was reading update logs. This could be a bug, as they did fix openai fine tuned models to now appear in the list for AI Agent nodes.

5

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.