Mismatch of audio file name between VoiceMOS Challenge 2022 audio dataset and mos-list

[bear1349]

Hello,

I created a runtime environment according to "VoiceMOS_baseline_README" and then downloaded the VoiceMOS Challange 2022 dataset from the specified website (https://zenodo.org/record/6572573#.Yphw5y8RprQ) and ran "python run_inference_for_challenge.py --datadir path/to/DATA". However, after displaying "Starting prediction", I got the following error message.
"RuntimeError: Error loading audio file: failed to open file . /main/DATA/wav/~.wav"
This may be due to a mismatch between the audio file name in the downloaded VoiceMOS Challenge 2022 dataset and in /main/DATA/sets/val_mos_list.txt.
Could you please confirm the facts and provide a solution?
Thank you!

[ecooper7]

Hello,

Have you run the scripts in the VoiceMOS dataset to gather all the data? The materials on Zenodo are not all of the data, but scripts to obtain the data. If you have already done these steps and still get the error, please let me know.