CUDA out of memory. Solução para quando a GPU fica sem memória e para de funcionar
Erro que ocorre no Stable Diffusion quando há falta de memória.
No meu caso, a seguinte mensagem apareceu:
CUDA out of memory. Tried to allocate 3.35 GiB (GPU 0; 8.00 GiB total capacity; 5.63 GiB already allocated; 843.24 MiB free; 5.65 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Edite o webui-user.bat e, se você não tiver o xformers, adicione-o.
Como isso também pode ocorrer ao usar VAE, adicione --no-half-vae.
set COMMANDLINE_ARGS=–xformers --no-half-vae
Fim.