Wrong GPU chosen -- running into "out of memory" issue

[StjernaBjoorn]

I'm trying to run the base version on a Windows 10 machine, but I am running into the common issue:

RuntimeError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 6.00 GiB total capacity; 5.20 GiB already allocated; 0 bytes free; 5.33 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

I have a NVIDIA RTX 3060, but also and integrated AMD Radeon GPU on the motherboard. The issue seems to be that SD is trying to run on the AMD card, and thus running out of memory.

Have anyone else run into this issue?

[kerlk]

У меня GTX 1660 ti, та же проблема. При этом в системных требованиях указано 4 гигабайта видеопамяти

[dagelf]

Duplicate tickets: https://github.com/CompVis/stable-diffusion/issues?q=is%3Aissue+is%3Aopen+4GB

@StjernaBjoorn: https://discuss.pytorch.org/t/how-to-change-the-default-device-of-gpu-device-ids-0/1041/9