Datasets:

---

bofenghuang

/

stt-pseudo-labeled-whisper-large-v3-multilingual

like 2

Is it possible to pre-filter the dataset based on WER? Currently, it looks like if I am streaming, I would have to manually throw out instances above 5% WER after it has already been streamed in.