Building LAMMPS with CUDA 12.6

[MicPellegrino]

Hi, I am trying to adapt the instructions in https://mace-docs.readthedocs.io/en/latest/guide/lammps.html to my workstation (Xeon w9-3495X + RTX A4000) but I am failing to build LAMMPS with the provided (and adapted) CMake flags.

After running:

git clone --branch=mace --depth=1 https://github.com/ACEsuit/lammps
wget https://download.pytorch.org/libtorch/nightly/cu126/libtorch-shared-with-deps-latest.zip
unzip libtorch-shared-with-deps-latest.zip
mv libtorch libtorch-gpu
cd lammps
mkdir build-ampere
cd build-ampere

cmake \
    -D CMAKE_BUILD_TYPE=Release \
    -D CMAKE_INSTALL_PREFIX=$(pwd) \
    -D CMAKE_CXX_STANDARD=17 \
    -D CMAKE_CXX_STANDARD_REQUIRED=ON \
    -D BUILD_MPI=ON \
    -D BUILD_SHARED_LIBS=ON \
    -D PKG_KOKKOS=ON \
    -D Kokkos_ENABLE_CUDA=ON \
    -D CMAKE_CXX_COMPILER=$(pwd)/../lib/kokkos/bin/nvcc_wrapper \
    -D Kokkos_ARCH_NATIVE=ON \
    -D Kokkos_ARCH_AMPERE86=ON \
    -D CMAKE_PREFIX_PATH=$(pwd)/../../libtorch-gpu \
    -D PKG_ML-MACE=ON \
    ../cmake

I get the following warning:

CMake Warning at /home/michele/lammps-mace/libtorch-gpu/share/cmake/Caffe2/public/cuda.cmake:184 (message):
  Cannot find NVTX3, find old NVTX instead
Call Stack (most recent call first):
  /home/michele/lammps-mace/libtorch-gpu/share/cmake/Caffe2/Caffe2Config.cmake:86 (include)
  /home/michele/lammps-mace/libtorch-gpu/share/cmake/Torch/TorchConfig.cmake:68 (find_package)
  Modules/Packages/ML-MACE.cmake:3 (find_package)
  CMakeLists.txt:586 (include)

and the following error:

CMake Error in CMakeLists.txt:
  Imported target "torch" includes non-existent path

    "/home/michele/lammps-mace/lammps/cmake/MKL_INCLUDE_DIR-NOTFOUND"

  in its INTERFACE_INCLUDE_DIRECTORIES.  Possible reasons include:

  * The path was deleted, renamed, or moved to another location.

  * An install or uninstall procedure did not complete successfully.

  * The installation package was faulty and references files it does not
  provide.

I am wondering if (and how) I can solve this issue or if MACE simply doesn't work with CUDA 12.6.

Cheers,
Michele

[wcwitt]

Here's something you could try: #660 (comment) .

[MicPellegrino]

Hi, I don't have a intel-mkl module installed, but I can try the 'CMake trick'.

[MicPellegrino]

It actually worked! Thanks!