I have retrained the whisper-large-v3-turbo model on my local data (there are few of them, 55 audio for several minutes and text to them), but the model that I retrained works worse than the original one. Why?
1 Like