Empty cuda cache between inferences.

jasonppy / VoiceCraft

Zero-Shot Speech Editing and Text-to-Speech in the Wild

Other

7.66k stars 749 forks source link

Empty cuda cache between inferences. #75

Closed Ph0rk0z closed 7 months ago

Ph0rk0z commented 7 months ago

I noticed large amounts of memory allocated between runs and tried to find out why. After you generate some TTS it will now go back down. Tested it with the gradio ui.