Speech to text models training without other language performance degradation

[NIKHILDUGAR]

I have been reading this paper https://arxiv.org/pdf/2406.06619 on improving STT models like Whisper for specific language improvement via LoRA, and wanted to know how to do this using Unsloth? Need urgent help. Thanks in advance.

[NIKHILDUGAR]

diagram from the paper