daekeun-ml

/

Phi-4-multimodal-finetune-ko-speech

like 14

Hi,
I am following your fine-tuning guidelines to fine-tune the base model (microsoft/Phi-4-multimodal-instruct) for ASR task, for special words in Spanish language. The issue is that the model forgets prior knowledge and invents words and phrases in random cases.
Have you noticed this in your case? Does your fine-tuned model forget prior knowledge?
Thanks