espnet/owls_4B_180K
Automatic Speech Recognition
•
Updated
OWLS: Scaling Laws for Speech Recognition and Translation
A suite of Whisper-style models from 250M to 18B parameters. Trained on up to 360K hours of data.