I have trained Whisper large-v3 model on my dataset of around 5-6 hours. The results are really good. But i noticed that the text is all lowercased and without punctuation. Is there any parameter that i missed during finetuning or during inference ?
I used this https://huggingface.co/blog/fine-tune-whisper blog to finetune on my dataset.
1 Like
I found a similar case.
1 Like
Thank you for replying. But in my case normal whisper-large-v3 works perfectly and the transcript is always with punctuation. But when i fine tune it only then i get the results in lower case. I think may be because i fine tune it with the dataset that is lowercased and normalized thats why the resulting transcript is also lowercased.
1 Like