Hi,
I am trying to run MACE in LAMMPS with ML-IAP. Accordingly, I have followed the compilation instructions in the website. While the compilation was successful, I see the following error message when I try to start a simulation -
Traceback (most recent call last):
File "mliap_unified_couple_kokkos.pyx", line 501, in mliap_unified_couple_kokkos.mliap_unified_connect_kokkos
File "/gpfs/users/chattopadhyay/LAMMPS_MACE_NO_OMP/lib/python3.12/site-packages/torch/serialization.py", line 1097, in load
return _load(
^^^^^^
File "/gpfs/users/chattopadhyay/LAMMPS_MACE_NO_OMP/lib/python3.12/site-packages/torch/serialization.py", line 1526, in _load
del torch._utils._thread_local_state.map_location
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: '_thread._local' object has no attribute 'map_location'
ERROR: Running mliappy unified module failure. (src/KOKKOS/mliap_unified_kokkos.cpp:248)
Last input line: pair_style mliap unified mace-mpa-0-medium.model-mliap_lammps.pt 0
The cmake command that I used to configure the system is as follows -
cmake -C kokkos-cuda.cmake -D CMAKE_CXX_COMPILER=$(which g++) -D CMAKE_BUILD_TYPE=Release
-D CMAKE_INSTALL_PREFIX=$(pwd) -D Kokkos_ARCH_ZEN2=ON -D Kokkos_ARCH_AMPERE80=ON
-D BUILD_MPI=ON -D PKG_ML-IAP=ON -D PKG_ML-SNAP=ON -D MLIAP_ENABLE_PYTHON=ON
-D PKG_PYTHON=ON -D BUILD_SHARED_LIBS=ON -D PKG_RIGID=ON -D PKG_MANYBODY=ON ../cmake
The version of CUDA is 12.6 whereas that of PyTorch is 2.4.1+cu121. I am running the simulations in a single NVIDIA A100-PCIE-40GB GPU card with the following command -
../lammps/build-mliap_no_omp/lmp -k on g 1 -sf kk -pk kokkos newton on neigh half -in in.input
I request you to kindly help me solve this issue. Thank you in advance for your suggestions.
Hi,
I am trying to run MACE in LAMMPS with ML-IAP. Accordingly, I have followed the compilation instructions in the website. While the compilation was successful, I see the following error message when I try to start a simulation -
The cmake command that I used to configure the system is as follows -
The version of CUDA is 12.6 whereas that of PyTorch is 2.4.1+cu121. I am running the simulations in a single NVIDIA A100-PCIE-40GB GPU card with the following command -
I request you to kindly help me solve this issue. Thank you in advance for your suggestions.