search

sidharthanup / Audio-Analyzer-and-Transcription-Tool

End to end information extraction from audio input.
MIT License
0 stars 2 forks source link

Pyannote Diarization Bottleneck #20

Closed sidharthanup closed 5 months ago

sidharthanup commented 5 months ago

Pyannote took a while to complete diarization - A 40 sec audio file with two speakers took 6mins.

sidharthanup commented 5 months ago

Solved. Pyannote was not using GPU as specified in the repo. Had to explicitly move the pipeline object to GPU by

pipeline = pipeline.to(torch.device(0))