Installation
====================
The easiest way to use MLatom is not to install it locally but run on the :doc:`XACS cloud <cloud>`.
If you want to install it anyway, please follow instruction below.

MLatom is a Python package and can be easily installed on Linux using this shell command:

.. code-block:: bash

   pip install mlatom 

You also need to install required dependences in your Python environment as described below.

Dependencies
------------

**Minimal setup**

.. code-block::
    
   pip install numpy scipy torch torchani tqdm matplotlib statsmodels h5py pyh5md

**Useful optional modules**

.. code-block::
   
   pip install sgdml rmsd openbabel xgboost scikit-learn pyscf rmsd rdkit pandas \
      ase fortranformat tensorflow

**Full Anaconda environment setup:**

Download :download:`mlatom.yml <files/mlatom.yml>`.

Then,

.. code-block::

   conda create -N mlatom --file mlatom.yml
   conda activate mlatom

**Additional software packages**

Many MLatom features are relying on other third-party software packages which are not Python modules and should be installed and setup separately as `described here <http://mlatom.com/download/#Installation_of_third-party_packages>`_. 

The third-party packages below are optional and can be installed separately to enable specific features.

- Newton-X
`Newton-X <https://newtonx.org/>`_ is required for `ML-NEA <https://doi.org/10.1021/acs.jpca.0c05310>`_ calculations.

1. Install `Newton-X <https://amubox.univ-amu.fr/s/aBMnEq7dXeZPm2H>`_ (NX, preferably version==2.2 for which our implementations were tested)
2. use ``export NX=/path/to/Newton-X`` to define the ``$NX``

- TorchANI
TorchANI is required for calculations with AIQM1 and ANI family of potentials.

1. install Numpy and nightly version of PyTorch (if you do not have them already):

.. code-block::

   pip install numpy tensorboard
   pip install --pre torch torchvision -f https://download.pytorch.org/whl/nightly/cu100/torch_nightly.html

2. install TorchANI:

.. code-block::

   pip install torchani

Visit https://aiqm.github.io/torchani/ for more info. The latest version of TorchANI used for testing was v2.2, you can 
install this version by ``pip install torchani==2.2`` if there are any problems when running with the newest version of 
TorchANI. The CUDA extension for AEV calculation is not supported for the NN part of AIQM1 and ANI-1ccx now.

- DeePMD-kit
Required for DPMD and DeepPot-SE potentials.

1. download installer for DeePMD-kit from `GitHub <https://github.com/deepmodeling/deepmd-kit/releases>`_ (tested v1.2.2)
2. run installer
3. add environmetal variable ``$DeePMDkit`` that point to the where dp binary is located (``bin/`` in your installation directory), 
   e.g. ``export DeePMDkit=/export/home/fcge/deepmd-kit-1.2/bin``.

- GAP and QUIP
Required for GAP-SOAP potentials.

1. compile QUIP and GAP from source

1.1 install prerequisites

.. code-block::

   sudo apt-get install gcc gfortran python python-pip libblas-dev liblapack-dev # for system uses apt, do equivalent for your OS
   pip install numpy ase f90wrap

1.2 get source code of QUIP and GAP

.. code-block::

   git clone --recursive https://github.com/libAtoms/QUIP.git

Get source code of GAP from http://www.libatoms.org/gap/gap_download.html (form-filling required).

Then put source code in ``QUIP/src/``.

1.3 compile

.. code-block::

   cd QUIP
   export QUIP_ARCH=linux_x86_64_gfortran_openmp # enable multi-threading, use 'export QUIP_ARCH=linux_x86_64_gfortran' if no OpenMP thus no MT capability
   export QUIPPY_INSTALL_OPTS=--user # omit for a system-wide installation
   make config 

Enter ``Y`` for gap or edit ``build/linux_x86_64_gfortran/Makefile.inc`` with ``HAVE_GAP=1``, then: ``make``.
Built binaries are in ``QUIP/build/linux_x86_64_gfortran/quip`` and ``QUIP/build/linux_x86_64_gfortran/gap_fit``.

2. add environmetal variable ``$quip`` and ``$gap_fit`` for quip and gap_fit, e.g.

.. code-block::

   export quip='/export/home/fcge/GAP-SOAP/QUIP/build/linux_x86_64_gfortran_openmp/quip'
   export gap_fit='/export/home/fcge/GAP-SOAP/QUIP/build/linux_x86_64_gfortran_openmp/gap_fit'

visit https://libatoms.github.io/GAP/index.html for more information.

- PhysNet
Required for PhysNet models.

1. clone form PhysNet's GitHub page

.. code-block::

   git clone https://github.com/MMunibas/PhysNet.git

2. install TensorFlow:

.. code-block::

   pip install tensorflow

3. if you use TensorFlow v2, you need to execute the command below in PhysNet's directory to make the scripts compatible with TFv2.

.. code-block::

   for i in `find . -name '*.py'`; do sed -i -e 's/import tensorflow as tf/import tensorflow.compat.v1 as tf\ntf.disable_v2_behavior()/g'
   -e 's/import tensorflow as tf/import tensorflow.compat.v1 as tf\ntf.disable_v2_behavior()/g' $i; done

4. add environmetal variable ``$PhysNet`` to the directory, e.g.

.. code-block::

   export PhysNet=/export/home/fcge/PhysNet/

- sGDML
Required for GDML and sGDML potentials.

1. install sGDML

.. code-block::

   pip install sgdml==0.4.4

2. add the path of sGDML binary to environmetal variable ``$sGDML``, e.g.

.. code-block::

   export sGDML=/export/home/fcge/.linuxbrew/bin/sgdml

Visit http://quantum-machine.org/gdml/doc/ for more information.

.. note::
   In our tests we found that installation is more stable with: ``pip install scipy==1.7.1``.

- MNDO
MNDO program is required to provide the ODM2* part of AIQM1. Alternatively, a `(development) version <https://zenodo.org/record/7362585>`_ 
of `SCINE Sparrow <https://scine.ethz.ch/download/sparrow>`_ can be used in the future (see a `paper <https://zenodo.org/record/7362585>`_ on Sparrow; 
note that the development version of Sparrow also implements single-point AIQM1 calculations).

The free binary and open-source code of the MNDO program is available from the official distributors of the MNDO code as described at https://mndo.kofo.mpg.de.

After the MNDO program is installed, you need to set the environmental variable pointing to the MNDO executable (typically ``mndo99``), e.g., in bash:

.. code-block::

   export mndobin=[path to the executable]/mndo99

- dftd4
dftd4 program is required to provide the D4 part of AIQM1.

The dftd4 program can be `obtained <https://github.com/dftd4/dftd4>`_ as both executable and open-source code. We recommend to use `dftd4 v2.5.0 <https://github.com/dftd4/dftd4/releases/tag/v2.5.0>`_, 
which can calculate Hessian needed for thermochemical calculations. To install the dftd4 program from source code, please see the README.md file 
on dftd4 GitHub page for more details.

After the dftd4 program is installed, you need to set the environmental variable pointing to the ``dftd4`` executable, e.g., in bash:
   
.. code-block::

   export dftd4bin=[path to the executable]/dftd4

- Gaussian
Required for geometry optimizations, freq, TS search, IRC, thermochemistry, and ML-NEA. For some of these tasks, alternatively, ASE can be used, see below.

Our implementation work with both Gaussian 09 and Gaussian 16. It is a commercial program, which can be `obtained and installed separately <https://gaussian.com/>`_.

To use Gaussian interface, make sure that your environmental variable ``$GAUSS_EXEDIR`` points to the right place.

- ASE
Required for geometry optimizations, freq, and thermochemistry. Alternatively, Gaussian can be used, see above.

The ASE (Atomic Simulation Environment) are Python modules, which can be installed as described on `ASE website <https://wiki.fysik.dtu.dk/ase/>`_, i.e.:

.. code-block::

   pip install ase

- hyperopt
To enable hyperopt, please run ``pip install hyperopt`` to install the `hyperopt <http://hyperopt.github.io/hyperopt/>`_ package.


Source code
-----------

Source code is on `GitHub <https://github.com/dralgroup/mlatom>`_:

.. code-block:: bash

   git clone https://github.com/dralgroup/mlatom.git