Full threadtempaccount420·Doesn't this use as much VRAM as llama.cpp (with int4 models) uses RAM? RAM is a lot cheaper than VRAM.View on HN