Unload model

[nixsrv]

Hi! How to unload faster-whisper model from VRAM? I need to use another model(NER) after work of F-W.

[guillaumekln]

You could delete the model object and then force the Python garbage collector to run. Try this:

import gc
from faster_whisper import WhisperModel

model = WhisperModel("large-v2")
del model
gc.collect()

[nixsrv]

Thanks. That's work.

[Denis-Kazakov]

Don't I also have to run torch.cuda.empty_cache()?