Training not starting

Hello!

First of all, thank you for your work on that fantastic tool.

I've got a problem on Windows though : sometimes, the training just doesn't start. I've tried twice with a dataset of 11 hours (which represents approximately 8600 clips), after preprocessing and feature extracting, nothing happens. In the console, there's no error message, there's just :

`load model(s) from assets/hubert/hubert_base.pt

move model to cuda

all-feature-16811

all-feature-done
`

And then nothing happens. I've been leaving it for an hour but nothing moves. So I've started with a smaller dataset of about 1 hour audio, and there the training started.

I've seen in the readme that a dataset between 10mn and 50mn was recommended, so maybe I went a bit over the top there, but I saw another thread here with someone who said they had trained over 4 hours audio. Have you guys tried training on really big datasets as well? Do you have an idea why the training doesn't start ? Should I try without the GUI?

Thank you very much!

No file or test is named. First reproduce the Windows training run with the reported 11-hour dataset and compare it with the smaller dataset that starts successfully, while recording where execution stops after feature extraction. Done means identifying why large datasets do not start training and documenting or correcting the reproducible behavior.