Speech to text: Create en empty json file
rohanjuneja opened this issue · comments
Rohan Juneja commented
The command to download and preprocess the LibriSpeech dataset creates an empty json file.
brevitas_quartznet_preprocess --data_root=/path/to/validation/folder --data_set=DEV_OTHER
This causes the evaluation command to fail.
Giuseppe Franco commented
Thanks for pointing this out.
We will investigate the issue.
Giuseppe Franco commented
This should be fixed now. Thanks for opening the issue!