Help required to prepare the dataset to train the model.

Hello,

I am having a hard time to train the model due to the dataset. I have downloaded the [mgb3](https://arabicspeech.org/mgb3-asr-2/) dataset and loaded according to the error logs I came across. Could you please write the steps to prepare the dataset for train the model.

Thanks!

Start with the MGB3 dataset documentation linked in the issue and the error logs referenced by the reporter; no log details, files, or tests are provided. Done means documenting reproducible dataset-preparation steps that allow the model to train.

Written by the indexing model from the issue text.