what is the best procedure to finetune toto2 ?

[Major1994]

Hello there,

I find your work fascinating.

I’m trying to improve Toto‑2’s performance for my own scenario and have come across a few issues.

Could you share the best practice for fine‑tuning Toto‑2 on custom datasets?

What dataset size do you suggest?

Would LoRA or adding new layers be good option?

Thanks.
Bisou bisou!!