Which documentation/build environment are you using?

I’m using Ubuntu(to follow what AMD has for ROCm) and building the entirety of (gfx803 patched) ROCm from source.

It works with some forks but not others.

IIRC it’s just the standard AMD build from https://repo.radeon.com/rocm/apt/5.2.3/ I think.

It’s possible I had to do something weird, but I think it supports Polaris OOB. rocminfo certainly thinks so.

I’m running it on Debian testing, with an equivs package to get it to install cleanly:

    $ cat rocm-equivs 
    Package: amdgpu-driver-fixes
    Provides: python,libstdc++-7-dev,libgcc-7-dev
    Architecture: all
    Description: Fixes the AMD GPU driver installation on Debian testing

I had to compile my own version of pytorch to get gfx803 support there.

I‘ll see if I can recreate the steps & create a runbook.

I am in the same boat with a gfx03 card. What patch did you use? The ones here? https://github.com/xuhuisheng/rocm-build

I also tried to compile pytorch with its Vulkan backend, but ended throwing the towel as LDFLAGS are a mess to get right (I successfully compiled it, but that was only part of the build chain, and decided I had better things to spend time on). I wonder how that would perform; ncnn works pretty decently.

Yes. I went through their build instruction shell scripts with some issues but it did work as below:

It works with an older fork that has img2img / txt2img separated from each other, but nothing else did.

When you built it, you did at least source the env.sh to pick up the build environment that it needed?

I did try that, but I don’t think the pytorch Vulkan backend is complete enough. I never did get a working install going down that route though, so I could be wrong.