Tim Moon

/te-ci pytorch

Merging with approval from @ksivaman, @sudhakarsingh27, @ptrendx. This feature is still experimental and incomplete.

/te-ci pytorch

It looks like there's a compilation error when building the core C++ library. Can you provide more of the error message so we can figure out where the error is...

CMake is failing since it can't find your CUDA installation. You can reproduce this outside of TE by making a `CMakeLists.txt` file: ```cmake cmake_minimum_required(VERSION 3.18) project(myproject LANGUAGES CUDA CXX) ```...

For future reference, https://github.com/NVIDIA/TransformerEngine/issues/700#issuecomment-1979377899 provides instructions on installing CUDA so it is available to CMake. I'll close this issue so this guidance is the last in the thread and is...

The simplest solution would be to manually construct an MLP out of multiple `te.Linear`s, but this won't be able to do all of the kernel fusions in `te.LayerNormMLP`. Long-term, this...

The changes are reasonable, but we are in the process of moving Userbuffers to the `common` library: https://github.com/NVIDIA/TransformerEngine/pull/760 Pinging @denera

The CMake error message should already be printed to stderr, although it is somewhat buried within the Python stacktrace from `setup.py`. It may be helpful to search for "Building CMake...

If you are experiencing an error that looks like `RuntimeError: Error when running CMake`, then something has failed in the build process (probably a CMake configuration error or a compilation...