Jacob Kahn

This op remains unsupported — this is a work in progress with the oneDNN backend.

@deepwilson it is - we're in the process of prioritizing features before starting to work consistently again. We'll post a more comprehensive roadmap in the next several weeks.

@yurivict feel free to submit a fix, else I will soon. Working on getting a clang 14 baseline in CI.

@JinLi711 — tracking as part of `vcpkg` overhaul which is a work in progress.

@elfring feel free to send a PR here -- otherwise, I'll send a fix in the near future.

@vchagari — can you share more info about your setup? What is `inference.exe`?

What's your CUDA version? Looks like you're trying to use an arch version which is newer than your CUDA version's nvcc will support.

@jasonppy — this command works for me without issue: ``` cmake .. -DCMAKE_BUILD_TYPE=Release -DFL_USE_ONEDNN=OFF -DFL_USE_ARRAYFIRE=ON -DFL_ARRAYFIRE_USE_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=86 ``` a few odd things about yours -- `FL_BUILD_ARRAYFIRE` isn't a valid build...

@jasonppy ah, you're on 0.3.2 — looks like new versions of CMake that set the CUDA arch flag in the way they do don't work with the copied CUDA shims....

Renamed to `BUILD_WITH_BOOST` and re-enabled building query, fragment, and build_binary since they aren't boost dependent, as you say 👍 . I've tested in a few limited environments, but it would...