Orochi VS Stable-Diffusion-ONNX-FP16

While you're waiting for the HIP SDK to release, check out Orochi as an alternative https://github.com/GPUOpen-LibrariesAndSDKs/Orochi

This isn't a full-fledged SDK but if you develop using the driver/runtime API and nvrtc on Linux, you could certainly use this library to make it run on Windows https://github.com/GPUOpen-LibrariesAndSDKs/Orochi Bonus - it also lets you compile a single binary that runs on both CUDA and HIP!

You can't run CUDA binaries directly. But you can use a wrapper library like Orochi to run both CUDA and HIP using a single binary that dynamically links with CUDA/HIP libraries at runtime https://github.com/GPUOpen-LibrariesAndSDKs/Orochi

I stumbled across orochi from AMD while looking for their FSR2.0 implementation, which basically boils down to being a wrapper over Cuda and HIP. I don't know if it is still maintained or functional, but heres the link if anyone is interested: https://github.com/GPUOpen-LibrariesAndSDKs/Orochi

There are a few versions of stable diffusion that someone ported and you can use them, I think I used https://github.com/Amblyopius/Stable-Diffusion-ONNX-FP16 with my amd card, then there's something like shark stable diffusion

Yes, the FP16 stuff in this latest release is a big part of that, and should tentatively support 8GB cards. I haven't pushed things quite as far as https://github.com/Amblyopius/Stable-Diffusion-ONNX-FP16 yet, but support for 4GB cards is possible and we've been discussing how it works: https://github.com/ssube/onnx-web/issues/241

Second option is: https://github.com/Amblyopius/Stable-Diffusion-ONNX-FP16