Aitomic add-ons¶
Aitomic add-ons provide UAIQM and other unpublished methods for use with open-source MLatom. Version 2.0.0 is available on PyPI and is free for academic, non-commercial research and teaching under the Aitomic Academic License. Commercial use requires a separate license.
Published methods including AIQM3 and OMNI-P2x are part of open-source MLatom; AIQM3 does not require the add-ons in MLatom 3.27.0 or later. See the AIQM2/AIQM3 tutorial.
You can also use UAIQM on the Aitomistic Lab@XMU and Aitomistic Hub.
Installation¶
Use Python 3.9, 3.10 or 3.11; Python 3.11 is recommended. Binary wheels are available for Linux x86-64 and ARM64, and macOS Intel and Apple Silicon. Windows wheels and source distributions are not provided. UAIQM models based on GFN2-xTB* require Linux for local calculations. They can also be used on the online platforms linked above.
Install in a fresh environment:
python3.11 -m venv uaiqm-env
source uaiqm-env/bin/activate
python -m pip install --upgrade aitomic-addons
This installs MLatom and the required Python packages automatically. There is no need to pin PyTorch to the older version 2.1.2. The first calculation downloads the selected model’s parameters and prints a one-time license notice.
Dispersion requirements depend on the selected UAIQM model and version:
The example below,
uaiqm_gfn2xtbstar@ccversion20250115, uses D3(BJ). Install D3.Other models and older versions use D4, including the ODM2*-based models and GFN2-xTB*-based version
20240106. Install D4 to use them.
Install both D3 and D4 if you plan to use automatic model selection or switch between these models. They are separate external programs.
Check the installation¶
The following calculation uses one explicitly selected UAIQM model and version:
import mlatom as ml
water = ml.data.molecule.from_xyz_string('''3
Water
O 0.00000 0.00000 0.11779
H 0.00000 0.75545 -0.47116
H 0.00000 -0.75545 -0.47116
''')
model = ml.methods(method='uaiqm_gfn2xtbstar@cc', version='20250115')
model.predict(molecule=water, calculate_energy=True,
calculate_energy_gradients=True)
print(f'Energy (Hartree): {water.energy:.8f}')
print('Gradient shape:', water.energy_gradients.shape)
Energy (Hartree): -76.37869477
Gradient shape: (3, 3)
Final digits can vary between platforms. Model version 20250115 is written
as a string in Python; in a command-line input use uversion=20250115.
Omitting the version selects the newest available version of the chosen model.
For optimization, frequencies and other tasks, see the UAIQM tutorial. It includes a sodium-containing example.
If installation fails¶
No matching distribution found for aitomic-addons means that PyPI has no
wheel matching your Python version, operating system or processor. Check
python --version against the supported combinations above. Python 3.12 or
newer is not supported by the current add-on wheels.
If D3 or D4 is missing, follow the corresponding installation instructions above. For assistance or commercial licensing, contact contact@aitomistic.com.
Citation¶
Please cite the UAIQM preprint, MLatom, the add-ons and the selected methods listed in the calculation output:
Yuxinxin Chen, Yi-Fan Hou, Olexandr Isayev, Pavlo O. Dral. Universal and Updatable Artificial Intelligence-Enhanced Quantum Chemical Foundational Models. ChemRxiv 2024 (preprint).
Pavlo O. Dral, Yuxinxin Chen, Mikołaj Martyka, Jingbai Li. Aitomic add-ons to MLatom; version 2.0.0. Aitomistic, Shenzhen, China, 2025–2026. https://aitomistic.com/mlatom/addons.html.
The MLatom citation and references for the underlying methods are also printed in the calculation output.