How do I train and eval using my own audio&text data?

I have .wav files & script+label csv file.

Data on this github is a pickle type.

How can I handle my data to use your model?

please let me know.