Fine-Tuning example

#1
by stefan-it - opened

Hi @pdelobelle ,

I would like to ask if you accidentally plan to also release some demo example that shows how to fine-tune a model on that dataset. I would be super interested, how the training examples would like, especially how theconstraints column is used.

Many thanks in advance!

Hi @stefan-it ,

I was not planning on releasing any code, since it is normal SFT. I am not using the constraintscolumn, as most are empty anyways. I just SFT'd on the query, synthetic_reasoning and synthetic_answer from a filtered version of SYNTH for Dutch. This works reasonably well, but I might try to do something to switch the reasoning to Dutch instead of English.

Sign up or log in to comment