Open matbee-eth opened 4 months ago
I can't load (app.py) on my 24gb vram gpu, is there a way to split it across multiple cuda devices?
Just skimming the code, seems like it may be possible to move just some of the models in CPU, and keep the slowest ones on GPU that fit within 24GB.
I can't load (app.py) on my 24gb vram gpu, is there a way to split it across multiple cuda devices?