[BUG] Building with GPU and Metatensor package fails

[PicoCentauri]

Summary

When trying to build lammps with the metatensor and the GPU package the final linking of the executable fails.

LAMMPS Version and Platform

Ubuntu 22.04

Expected Behavior

Lammps can be build successfully.

Actual Behavior

Lammps build fails.

/usr/bin/ld: liblammps.a(info.cpp.o): in function `LAMMPS_NS::Info::get_gpu_device_info()':
/home/loche/repos/lab-cosmo/lammps/src/info.cpp:1115: undefined reference to `lmp_gpu_device_info()'
/usr/bin/ld: liblammps.a(info.cpp.o): in function `LAMMPS_NS::Info::has_accelerator_feature(std::string const&, std::string const&, std::string const&)':
/home/loche/repos/lab-cosmo/lammps/src/info.cpp:1166: undefined reference to `lmp_gpu_config(std::string const&, std::string const&)'
collect2: error: ld returned 1 exit status
make[2]: *** [CMakeFiles/lmp.dir/build.make:114: lmp] Error 1
make[1]: *** [CMakeFiles/Makefile2:259: CMakeFiles/lmp.dir/all] Error 2
make: *** [Makefile:136: all] Error 2

Steps to Reproduce

Further Information, Files, and Links

TORCH_PREFIX=$(python -c "import torch; print(torch.utils.cmake_prefix_path)")

cmake \
    -DPKG_GPU=yes \
    -DGPU_API=cuda \
    -DGPU_PREC=single \
    -DPKG_ML-METATENSOR=ON \
    -DCMAKE_PREFIX_PATH="$TORCH_PREFIX" \
    ../cmake

make

[Luthaf]

I assume this works if you set -DPKG_ML-METATENSOR=OFF?

[PicoCentauri]

Yes.