Purfview
Purfview
FYI, you shouldn't run it on CUDA as the model is not meant to run on it. Benchmark on ~2h audio with RTX4090: CUDA: 72.22 seconds CPU: 15.15 seconds
5 hours shouldn't be a problem for the modern hardware. Try smaller model if you don't have GPU with CUDA.
Share your command.
> the first line of the result is literally the same as the issue mention in... One line of hallucination is nothing to worry about. > i guess my program...
`beam_size` is not related to segments duration, it's size of [beam search](https://en.wikipedia.org/wiki/Beam_search).
On some other audio you can observe opposite effect.
`0.10.0` is broken on pypi. `0.10.1` is unbroken 0.10.0.
Sometimes it stops without error when it's actually "out of memory". How much VRAM there is in your GPU?
Here it doesn't have CLI nor an output writer. Use "forks" with CLI: https://github.com/Softcatala/whisper-ctranslate2 https://github.com/Purfview/whisper-standalone-win
Are you comparing with same quantization type?