2. Compiling Amber#

Amber + QMhub + Torchmd-net

Compile Amber for QM/MM simulations with QMhub and Torchmd-net.

Using Amber22 and AmberTools23 on Pete.

Includes

  • sander.MPI

  • SINR thermostat

  • adaptive string method

2.1. Create Conda Environment#

  1. Create a torchmd-net conda environment, and load modules.

conda create -f torchmd-net.yml
conda activate torchmd-net

module load intel/2021.2.0
module load impi/2021.2.0
export MKLROOT=/opt/intel/oneapi/mkl/2021.2.0
export LD_PRELOAD=$MKLROOT/lib/intel64/libmkl_core.so:$MKLROOT/lib/intel64/libmkl_sequential.so

  1. Clone repository to Programs/ and change branch to torchnet-ext

git clone 
cd torchmd-net
pip install .

  1. Clone and to Programs/. We’ll compile qmhub first.

git clone
git clone

cd qmhub
pip install .

cd ../helpme
CXX=icpc PKG_CONFIG_PATH=~/miniforge3/lib/pkgconfig cmake .. -DFFTW_INCLUDES=$MKLROOT/include/fftw -DFFTW_LIBRARIES=$MKLROOT/lib/intel64
make helpmelib

  1. Copy the python/helpme*** file to minforge3/envs/torchmd-net/lib/python3.10/site-packages/qmhub/

2.2. Amber#

  1. In Programs/sources, decompress AmberTools then Amber

tar xvjf AmberTools
tar xvjf Amber

  1. Apply patches in amber22_src/, qmhub, sinr, sqm, and then asm.

patch -p 1 < patch_file

  1. Go to the build/ directory

cd build/

  1. Change the run_cmake and run it ./run_cmake:

Listing 2.4 run_cmake#
  cmake $AMBER_PREFIX/amber22_src \
    -DCMAKE_INSTALL_PREFIX=/scratch/van/.Programs/amber22 \
    -DCOMPILER=INTEL \
    -DMPI=FALSE -DCUDA=FALSE -DINSTALL_TESTS=FALSE \
    -DFORCE_EXTERNAL_LIB=mkl \
    -DDOWNLOAD_MINICONDA=FALSE -DBUILD_PYTHON=FALSE \
    2>&1 | tee  cmake.log

Check if everything is properly linked.

  1. Before we run make install -j 8, we’ll build sander first, and test it.

make sander -j 8

Test sander using the source in the AmberTools directory.

If the test went well, go back to the build/ and compile the MPI version.

make sander.MPI -j 8