PX915ReproducibleResult

Installation instructions (this is REALLY important)

The following instructions detail how to install everything you need for this package to work on one of the SCRTP nodes, e.g hetmathsys.

Loading modules

Once you have logged into an SCRTP node, load up the following modules:

module purge; module load GCC/11.2.0 OpenBLAS/0.3.18 CMake/3.22.1 OpenMPI/4.1.1 Python/3.9.6

Making a virtual python environment

The next step is to create a virtual environment, which is essentially makes it much easier to avoid any dependency clashes. Also, it means that once you're done you can just delete the virtual environment and it will get rid of everything.

python -m venv myvenv
source myvenv/bin/activate

Installing LAMMPS into venv

1. First, navigate to the directory that this readme is in once you've git cloned this repository, and download LAMMPS locally using git, e.g,

git clone -b release https://github.com/lammps/lammps.git mylammps

Where mylammps is the name of the directory that will be made.

2. There should also be an install_lammps.sh script in this directory. You need to make the script runable with

chmod u+x build_lammps.sh

3. Now just run

./build_lammps.sh

This might take a couple of minutes.

4. Deactivate and re-activate your virtual environment, with e.g,

deactivate
source myvenv/bin/activate

5. Check the install worked correctly. If you run the following, you should see the output below.

python
>>> from lammps import lammps
>>> lmp = lammps()
LAMMPS (15 Sep 2022)
using 1 OpenMP thread(s) per MPI task

If you don't see this, then something has gone wrong. Feel free to get in touch with me and I can help you sort it out (installing LAMMPS can be very fiddly). However, if you follow the above exactly step by step then it should work fine.

Installing matscipy and it's dependencies

We next need to install matscipy, which contains the bulk of the code that was written over the summer. Navigate to the directory that this readme is in and run the following:

pip install git+https://github.com/libAtoms/matscipy

This should install matscipy and it's dependencies.

Getting the Jupyter Notebook working

You now need to install ipython, and create a kernel for you virtual environment to use the jupyter notebook. Run:

pip install ipython==8.5.0
pip install ipykernel

Now deactivate and reactivate your virtual environment again, e.g,

deactivate
source myvenv/bin/activate

Now run

ipython kernel install --user --name=myvenv

and finally install all the remaining modules

pip install nglview h5py mpi4py

and then (making sure you're in the same directory as this readme), run

jupyter-notebook

Once the you've opened up the notebook (Walkthrough_Notebook.ipynb), select 'kernel > change kernel > myvenv'. Now you should be able to run the notebook. This notebook contains information on and examples of code written over the summer and how to interface with it, as well as instructions on how to run the reproducible result.

Restart guidance:

If you don't want to do this all in one sitting (or open another terminal), run the following commands to load up the virtual environment and modules:

module purge; module load GCC/11.2.0 OpenBLAS/0.3.18 CMake/3.22.1 OpenMPI/4.1.1 Python/3.9.6
source myvenv/bin/activate

POSSIBLE ERRORS

If you see the line 'command not recognised: mpirun', this means that you likely started a new terminal and did not reload the modules at the top of this page before entering the virtual environment. Run:

deactivate
module purge; module load GCC/11.2.0 OpenBLAS/0.3.18 CMake/3.22.1 OpenMPI/4.1.1 Python/3.9.6
source myvenv/bin/activate