-
I ran the finetune.py on my dataset, but looking into the metadata_train.csv and metadata_eval.csv, capitalisation is ignored. When prompting the model, does capitalization matter at all? Also, can I just change the .csv files after running whisper to modify things in case whisper got it wrong? Is that how it works? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
Hi @Weroxig I ran the finetune.py on my dataset, but looking into the metadata_train.csv and metadata_eval.csv, capitalisation is ignored. When prompting the model, does capitalization matter at all? Also, can I just change the .csv files after running whisper to modify things in case whisper got it wrong? Is that how it works? Thanks |
Beta Was this translation helpful? Give feedback.
Hi @Weroxig
I ran the finetune.py on my dataset, but looking into the metadata_train.csv and metadata_eval.csv, capitalisation is ignored. When prompting the model, does capitalization matter at all?
It doesn't matter in any way that I am aware of.
Also, can I just change the .csv files after running whisper to modify things in case whisper got it wrong? Is that how it works?
Yes you can change/edit these before moving to the training step.
Thanks