too big to run
#320
by
karan963
- opened
is there any quantized version of this available so we can replace this model with smaller version?
Get a GPU with more than VRAM and it will work, no need replacing as lower quantized versions will not output same quality/cohesive images
@LancerMaster Recommend how much GB?
I have a 4090 with 24GB of VRAM and that does not seem to be enough. When I try to load the model to my cuda device, I always get:
torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 90.00 MiB. GPU 0 has a total capacity of 23.62 GiB of which 94.94 MiB is free. Process 2033 has 488.00 MiB memory in use.
but the values of RAM it says it is trying to allocate and that are free varies. What is the minimum VRAM needed for this model?