Getting original file mapping of GTZAN dataset

[jichan3751]

Hi, thanks for the great package!

For the GTZAN dataset, it seems like the dataset is provided as pre-processed array pkl file, and labels are all shuffled.
Is there way to know which file each wav array is generated from the original GTZAN dataset?
For example, for the i th data in train_wav_list, what is the filename in GTZAN data as "(label)/xxxx.wav"?

If you can also provide preprocessing script, I can try to infer from it.

Thanks!