Convert PEFT Model to GGML Format

[Mariomavridis]

Hello this project is awesome!
I want to fine tune whisper with PEFT-Lora in a Colab on multiple languages.
I've managed to make a few adjustments and run the trainer with large-v3 on turkish audio from common voice 17.
After the training finished i merged my model with model.merge_and_unload() and pushed it to the hub.
However the file size is much smaller then the original model and it has a ".safetensor" format and like i expected the convert-scripts didn't work.
Has anyone some ideas how I could convert this format into ggml?
Thank you for help! :)