Pre-load all models in RAM

#12
by multimodalart HF staff - opened

Pre-load all models in RAM and swap between RAM and VRAM for faster inference

anzorq changed pull request status to merged

Awesome, thanks!

Sign up or log in to comment