One-year overview: from MLatom 3.0 to 3.10

A year ago, we released MLatom 3, making MLatom the fully-fledged Python package. This shift endowed MLatom with much of its current flexibility that also allowed us to rapidly do method development and new implementations: in one year we had 10 big upgrades of the package (12 releases on PyPI). So do not forget to upgrade your MLatom! Just run the command:

python3 -m pip install -U MLatom

Here we make a summary of the upgrades in case you missed them (if you do not want to miss these upgrades in the future, please subscribe to our channels as described at https://xacs.xmu.edu.cn/docs/mlatom/contact.html#social-media):

We will start with extra updates of the MLatom@XACS (i.e., MLatom which is only available on the http://XACScloud.com and provides you with the latest experiences not available yet on the GitHub’s version):

• Universal and updatable AI-enhanced quantum mechanical methods (UAIQM) – our latest offering in the universal interatomic potentials. They are considered as the successors to AIQM1: they offer coupled-cluster accuracy for many molecules as well as ~DFT-level accuracy for the entire periodic table. With them, you can, e.g., run overnight hundreds and thousands of MD trajectories with better-than-DFT accuracy. And they are getting better with more usage!
• IR and Raman spectra
• Calculations with DFT ensembles – getting results better than with any individual DFT functional.

Updates of the MLatom on GitHub/PyPI:

• MLatom 3.10.0–3.10.1 (21-22.08.2024) – active learning for surface hopping MD, multi-state ANI for excited states, gapMD for efficient exploration of the conical intersection regions, quality of life improvements such as viewing molecules, databases, and trajectories in Jupyter, easier load of molecules
• MLatom 3.9.0 (23.07.2024) – periodic boundary conditions
• MLatom 3.8.0 (17.07.2024) – directly learning dynamics
• MLatom 3.7.0-3.7.1 (03-04.07.2024) – active learning & batch parallelization of MD
• MLatom 3.6.0 (15.05.2024) – new universal ML models (ANI-1xnr, AIMnet2, DM21)
• MLatom 3.5.0 (08.05.2024) – quasi-classical trajectory/molecular dynamics
• MLatom 3.4.0 (29.04.2024) – usability improvements with a focus on geometry optimizations
• MLatom 3.3.0 (03.04.2024) – surface-hopping dynamics
• MLatom 3.2.0 (19.03.2024) – diffusion Monte Carlo and energy-weighted training
• MLatom 3.1.0 (12.29.2023) – MACE interface

I cordially thank the amazing MLatom developers team which is growing and making possible all this progress.I also thank the users who motivate our work and give crucial feedback helping us to make MLatom better.

MLatom has become so mature for computational chemistry that it has allowed me to make a hands-on mini-course ‘Modern computational chemistry and AI with MLatom@XACS’, which I use to teach the courses. The feedback of its readers shows that you or your students can use the course for self-education and as a quick reference. If there is enough interest (please let us know!), I am going to make a ‘living’ updatable course too.

If you like MLatom, please share this post with your friends and colleagues.