GitLab

Co-authored-by: zenghongtai <1518445275@qq.com>

* static constexpr should be used in cuda/reducer.cuh, otherwise compilation fails in VS

* modify torch/include/ATen/Parallel.h in testing.yml

* update

* update

* update

* typo

* typo

* typo

* remove prints
Co-authored-by: zenghongtai <1518445275@qq.com>
Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de>

* update

* update

* update

* update

* update

* update

* update

* update

* update

* typo

* typo

* first step, everything compiles

* fix rebuilds; skip cuda version check for rocm

* use macro for __shfl_up_sync __shfl_down_sync

* add BFloat16 support for ROCm and CUDA

* add USE_ROCM definition to setup.py

* flake8 fixes

* Mark exported symbols with SCATTER_API so they are available in the DLL on Windows

* macro definition for python package installation

* move macros definitions to separate file

* unlink?

* update

* linting
Co-authored-by: Matthias Fey <matthias.fey@tu-dortmund.de>

* compile with half

* Fix

* fix

* rename

* update

* update

* update

* update

* typo
Co-authored-by: rusty1s <matthias.fey@tu-dortmund.de>

This reverts commit 0be33ffa.