Finetuning a LLM with Company Data

KNIME Analytics Platform
1

Hello,

I would love some help here as this is new to me! We have about 2 Million Rows of Data from ZenDesk - Customer Support Tickets and answers. Now we want to take this data and train / FineTune a LLM with this data. Can we do this in Knime? Are there any good starting point tutorials?

Thank you

2

Hey there,

fine tuning is possible within KNIME - at least for OpenAI models there is this node:

There’s also an example workflow that the KNIME Team has prepared:

As for Open Source models I’m not 100% sure.

2 Likes
3

This is working perfect for OpenAI. How can i do that locally with Llama?

Thank you

4

As far as I know and can tell based on some research this is not supported “out of the box” by any node yet.

That said you may be able to use the Python Integration combined with a custom Conda Env to set something up using a Python Script node.