Yoshiki

/

chime4_enh_asr1_wpd_wavlm_conformer

like 0

Hi, you worked with chime4 dataset, so please help me understand it. I am trying to run egs2/chime4/enh_asr1 recepie or this with your huggingface card. I downloaded the dataset (https://www.chimechallenge.org/challenges/chime4/download) and it doesn't look like the official site (https://www.chimechallenge.org/challenges/chime4/data#Audio), there is no
├── audio
├── transcriptions
└── WSJ0
only
├── annotations

same in with huggingface dataset card (https://huggingface.co/datasets/ksbai123/Chime4).
How can I access the missing data and make train, dev, test split? Can you provide me:
├── audio
├── transcriptions
└── WSJ0
if you have it? I need it form my phd thesis.