Finetuning example
#3
by
RASMUS
- opened
Do you have some finetuning example you would like to share?
I have been thinking of finetuning this or the nllb one for Finnish pdf(image) retrieval
Any SigLIP/SigLIP2 finetuning script should work, just use model/processor initialization code from readme. Regarding PDFs, I'd recommend using this model because it has higher resolution (512px vs 384px) and potentially crop PDF images if they have high resolution and/or small text.
RASMUS
changed discussion status to
closed