[Request] support for lite whisper series
#3
by
pzoltowski
- opened
Would be great to add support for some lite whisper series:
https://huggingface.co/models?search=litewhisper
They are supported already in transformers and at least 'lite-whisper-large-v3-turbo-acc' looks interesting: Almost same WER, 50% smaller encoder size, lower latency, faster inference - especially useful on iOS.