Welcome to MLatom documentation!

MLatom is a program package for AI-enhanced computational chemistry. It is designed to leverage the power of ML to speed up and make more accurate common simulations and to create complex workflows.

The users can run simulations with input files, Python script, and command-line options.

MLatom is an open-source package that can be installed via pip and cloned from GitHub. Additional advanced features can be obtained via the Aitomic add-ons. The calculations with MLatom can be performed on the XACS cloud computing service and Aitomistic Hub.

A video overview of the MLatom capabilities:

To get quickly started, please check a simple example illustrating the use of MLatom.

See a detailed overview of capabilities of MLatom for more information.

Overview

• Overview
• License and citations
• Releases
• About
• Contact and support

Setup

• Installation
• Source code
• Use
• Run on XACS cloud

Tutorials: Introduction

• Get started
• Data
• Workshop tutorials

Tutorials: Methods

• UAIQM
• AIQM1
• AIQM2
• Universal ML models
• DFT
• DFT ensembles

Tutorials: (ML) models

• Building models
• Machine learning potentials
• Active learning
• Transfer learning
• Delta-learning
• Learning molecular dynamics

Tutorials: Simulations

• Single-point calculations
• Geometry optimization
• Transition states
• Frequencies and thermochemistry
• Infrared spectra
• Raman spectra
• Molecular dynamics
• Quasi-classical molecular dynamics
• Vibrational spectra from MD
• Periodic boundary conditions

Tutorials: Excited states

• Methods for excited states
• Single-point calculations
• Data with excited states
• ML of excited states
• Absorption and emission spectra
• Surface-hopping dynamics
• Active learning for TSH

Manual for input file/CLI

• Overview
• Simulations
  • Single-point calculations
  • Geometry optimization
  • Frequencies and thermochemistry
  • IRC
  • Molecular dynamics
  • IR and power spectra from MD
  • Simulations with universal ML-based models
  • Simulations with QM methods
  • Simulations with user-trained models
  • Quantum dynamics with machine learning
  • UV/vis spectra
  • Two-photon absorption cross sections
• Learning
  • Training popular ML models
  • Training generic ML models
  • Optimizing hyperparameters
  • Evaluating ML models
  • Δ-learning
  • Self-correction
• Data
  • Converting XYZ coordinates to molecular descriptor
  • Analyzing data sets
  • Sampling and splitting

Manual for API

• Overview
• Data
  • atom
  • molecule
  • molecular_database
  • molecular_trajectory
  • h5md
• Models
  • methods
  • ml_model
  • model_tree_node
  • Interfaces
• Simulations
  • optimize_geometry
  • freq
  • thermochemistry
  • dmc
  • Initial conditions
  • Molecular dynamics
  • Surface-hopping dynamics
• Spectra
• Active learning
  • Initial data sampling
  • User-defined ML models
  • Sampler
• Analysis

Indices and tables

• Index

• Module Index

• Search Page