parler-tts icon indicating copy to clipboard operation
parler-tts copied to clipboard

Published 20 hours ago • verified publisher icon huggingface

Slow inference on apple silicone

Open SaKiEQ opened this issue 9 months ago • 0 comments

Howdy,

this is more of question to validate what I have been doing as a error. The setup I am using runs just fine.

Apple Silicone M3 - macBook - OS X Sequoia latest 15.3 (24D60) version - 18GB ram

Using "parler-tts/parler-tiny-v1-jenny" voice for testing. torch_device = "mps"

Yet my inference times on a 100 char sentence is around 90seconds. It averages to about 10x per second of audio generated.

Is this the expected speed, or is something totally wrong at my end ?

SaKiEQ avatar Jan 23 '25 16:01 SaKiEQ