Introduction

Ta-dah! is a modular and fast machine learning software and C++ library specifically designed for the interatomic potential development in mind. It is written in modern C++ with the intention to provide an easy to use, modular and extensible state-of-the-art toolkit.

Ta-dah provides LAMMPS interface which works with all provided descriptors and models.

Ta-dah! can be used as a software from the command line to train models or use already developed machine learning potentials (with Ta-dah!) for prediction. Alternatively, it can be included as a C++ library for the more advanced use.

Why use Ta-dah!?

  • Ta-dah! is community driven and new ideas are always welcome. If it is possible to implement it into Ta-dah! it will be done.

  • Ta-dah! is fast. Typical model development cycle is much faster - no more waiting for hours to refit a model.

  • Ta-dah! is improving almost every day. New descriptors and models are being added. Bugs are squashed and issues are resolved.

  • Ta-dah! is open-sourced.

  • Ta-dah! allows to mix and match various descriptors with cutoffs and models. Once model is trained it can be immediately tested with LAMMPS. It just works!

  • Ta-dah! is extensible. Adding new descriptors is fairly easy and they will work seemingly with the rest of the code (including LAMMPS) interface.

Requirements

Ta-dah! does not require any external libraries to be build or to be downloaded.

Obtaining Ta-dah!

The code lives here:

To download use git and clone from the stable branch:

git clone -b stable https://git.ecdf.ed.ac.uk/s1351949/ta-dah.git

Installation

The library uses CMake to control configuration and compilation process. The CMake will try to guess number of system depended variables based on the config file CMakeList.txt located in the project root directory.

# cd into project dir
cd ta-dah

# create build directory and cd to it
mkdir build && cd build

# Use cmake to configure
cmake ..

# compile and install
make && make install

To change default library installation location run below command instead of cmake ... This is useful when you don’t have root privileges.

cmake .. -DCMAKE_INSTALL_PREFIX=/your/path

Binary file

The executable file ta-dah will be installed to bin directory inside your installation location. On most UNIX systems the default is /usr/local/bin/ta-dah.

In case of using -DCMAKE_INSTALL_PREFIX=/your/path flag, the binary file location is /your/path/bin/ta-dah.

This is usually the end of the installation process for most users. If you are intending to use Ta-dah! as a C++ library continue to the next section.

Shared Library and C++ Headers

The header files and shared library will be installed to the system default location (on UNIX headers are usually installed to: /usr/local/include and libtadah.so shared libary to /usr/local/lib) or the path specified by the user with -DCMAKE_INSTALL_PREFIX flag.

When compiling code with Ta-dah! make sure that linker can find shared library object. On UNIX machines this is usually done by exporting LD_LIBRARY_PATH variable. Similarly compiler must know where header files are stored. For example, for GCC compiler this is done by exporting CPATH. This step is usually not required when library is installed to the system default location.

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/lib
export CPATH=$CPATH:/path/to/include

Compiling and Linking

The library must be compiled with C++17 or higher. To include library header remember to use prefix tadah/. For example to use methods defined in atom.h

#include <tadah/atom.h>

To compile code written with Ta-Dah! link ltadah. If OpenMP is available link fopenmp. See Go Faster Stripe for details.

For example, to compile test.cpp with GNU g++

g++ -O3 -std=c++17 -o output_file test.cpp -ltadah -fopenmp

If the library is not installed in the standard search path you will have to link it manually, for example:

g++ -O3 -std=c++17 -o output_file test.cpp -I/home/$USER/.local/include -L/home/$USER/.local/lib -ltadah -fopenmp

Go Faster Stripe

The library will be compiled with OpenMP if it is available on the system. OpenMP uses the system variable to control the number of threads.

To set number of threads to be used by the library on a UNIX machine:

export OMP_NUM_THREADS=num_of_cores

Always link fopenmp when compiling code which uses Ta-dah! library.

Installing LAMMPS interface

First, install Ta-dah! library as described in Installation.

To build Ta-dah with lammps you need to download LAMMPS source code if you don’t have it.

Copy tadah/build/USER-TADAH directory to LAMMPS src directory.

Warning

Make sure you are copying USER-TADAH from the build directory and NOT from the src/lammps/USER-TADAH.

Hint

The file USER-TADAH/Makefile.lammps specifies paths to libtadah. If it is not installed in the default location you might have to edit this file. Cmake will try to guess correct locations.

Go to lammps source directory (e.g. mylammps/src) and run make yes-user-tadah and then recompile LAMMPS (e.g. make serial).

Important

Ta-dah should work with LAMMPS versions from June 2022 onwards. It has been tested with the following LAMMPS version: 23 Jun 2022 - Update 2.

See also

Detailed instructions on how to build LAMMPS with extra features:

https://lammps.sandia.gov/doc/Build_package.html

Troubleshooting

If you have any issues with a compilation process do not hesitate to drop me an email.

  1. Issue: Permission denied

    This error might happen when running make install.

    Install the project...
-- Install configuration: "Release"
-- Installing: /usr/local/bin/ta-dah
CMake Error at src/cmake_install.cmake:55 (file):
  file INSTALL cannot copy file
  "/home/$USER/ta-dah/build/src/ta-dah" to
  "/usr/local/bin/ta-dah": Permission denied.
Call Stack (most recent call first):
  cmake_install.cmake:42 (include)


make: *** [Makefile:129: install] Error 1

    You are installing to root directory without root privileges hence files cannot be copied to the required destination. Either run the command with root privileges (sudo or equivalent) or change install location to user directory (e.g. $HOME/.local for most UNIX systems):

    cmake .. -DCMAKE_INSTALL_PREFIX=~/.local
make install

  2. Issue: Shared object ImportError

    ImportError: libtadah.so: cannot open shared object file: No such file or directory

    libtadah is a dynamic library and your system does not know where to find it. Adding the directory containing libtadah.so to LD_LIBRARY_PATH should solve the problem.

    export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/your/path/

    It usually make sense to add above command to ‘~/.bashrc’ file so it is executed each time a new terminal window is opened.

  3. Issue: error: ‘…class LAMMPS_NS::Comm’ has no member named… or similar

    ../pair_tadah_tadah.cpp: In member function ‘void LAMMPS_NS::PairTadah::compute_2b_mb_half(int, int)’:
../pair_tadah_tadah.cpp:242:15: error: ‘class LAMMPS_NS::Comm’ has no member named ‘reverse_comm_pair’; did you mean   ‘reverse_comm’?
242          comm->reverse_comm_pair(this);
../pair_tadah_tadah.cpp:263:15: error: ‘class LAMMPS_NS::Comm’ has no member named ‘forward_comm_pair’; did you mean     ‘forward_comm_array’?
263          comm->forward_comm_pair(this);

    This is likely to be an incopatibility between your version of LAMMPS and Ta-dah! See Installing LAMMPS interface for more information.

  4. Issue: error: ‘…error: ‘int LAMMPS_NS::NeighRequest::half’ is protected within this context… or similar

    ../pair_tadah_tadah.cpp: In member function ‘virtual void LAMMPS_NS::PairTadah::init_style()’:
../pair_tadah_tadah.cpp:888:43: error: ‘int LAMMPS_NS::NeighRequest::half’ is protected within this context
../pair_tadah_tadah.cpp:263:15: error: ‘class LAMMPS_NS::Comm’ has no member named ‘forward_comm_pair’; did you mean ‘forward_comm_array’?

    This is likely to be an incopatibility between your version of LAMMPS and Ta-dah! See Installing LAMMPS interface for more information.

  5. Issue: New issue

    If you have an issue which is not listed here please get in touch.