How much VRAM is required to run it without CPU offloading? Specifically, will it work on an RTX 3060 with 12GB of VRAM without relying on the CPU? #152
We haven't tested it on the 3060, but we have received reports that it won't run on a 24GB GPU with fp32 (instead of bf16) without CPU offloading. This suggests that it may not run on a 12GB GPU with bf16.
We haven't tested it on the 3060, but we have received reports that it won't run on a 24GB GPU with fp32 (instead of bf16) without CPU offloading. This suggests that it may not run on a 12GB GPU with bf16.