Tried the HIP build install instructions and now my Stable Diffusion model doesn't compute

[CurseWave]

So while trying to install llama.cpp (and failing), I tried the instructions for the HIP section (https://github.com/ggerganov/llama.cpp/blob/master/docs/build.md#hip) of the build.md. Doing so seemingly did nothing to enable llama.cpp, but that's a matter I've given up on.

What I'm trying to troubleshoot now is that my memory is apparently tanked by the addition of HIP to my system. I get the following message when trying to gen reasonable images:

20:33:00-759347 ERROR VAE decode: HIP out of memory. Tried to allocate 8.86 GiB. GPU 0 has a total capacity of 15.98 GiB of which 5.58 GiB is free. Of the allocated memory 9.44 GiB is allocated by PyTorch, and 597.74 MiB is
reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_HIP_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory
Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

NGL, I don't even know why HIP is detected by my app (SD Next) and I'm not sure where to begin with fixing this. I'm actually contemplating doing a fresh install of my arch linux set up.