Welcome to the Tadah! Website

Tadah! 1.6.0

This release is ready to use: the Tadah! authors fit and run potentials with this version, and feedback on it is welcome. The command-line interface and the pot.tadah format are stable across 1.x, except for the breaking changes listed in the Changelog. The C++ API is not stable and may change before 2.0.

The Changelog lists the changes of this release. An upgrade starts with its breaking changes, notably the error scales and priorities of a fit, which replace EWEIGHT, FWEIGHT and SWEIGHT (What a Fit Minimises). Issue reports are welcome.

This is the home of Tadah! software, used for developing and deploying machine learning interatomic potentials. This website provides documentation with examples to help users quickly get up to speed, along with an associated API for developers.

This site also publishes selected interatomic potentials developed by Prof. Ackland's group using Tadah! software, listed in Trained MLIPs.

General Information

• Introduction
  • Key features
  • Commands
  • Obtaining Tadah!
• Installation
  • Requirements
  • Installing Tadah!MLIP Software
  • Installing ML‑TADAH for LAMMPS (Tadah!LAMMPS)
  • Libraries and C++ Headers
  • Troubleshooting
• Trained MLIPs
  • N2
  • Kr
• Contributing to Tadah!
  • Getting Help
• Glossary and Abbreviations
• Abbreviations
• Changelog
  • [Unreleased]
  • [1.6.0] — unreleased
  • [1.5.0] — 2026-09-10
  • [1.4.0] — 2026-07-14
  • [1.3.0-beta.1] — 2026-05-20
  • Migrating from v1.2.0 to 1.3.0-beta.1
  • [1.2.0] and earlier (condensed)
• Licence

---

How to Use Tadah!

• Quick Start
  • Overview
  • Built-in Help
  • Training
  • Prediction
  • Units and Tadah!
  • Config File
  • Dataset Format
  • Tadah! and LAMMPS
  • Support for Multi-species Systems
• Command Line Interface
  • Basic invocation
  • One command per run
  • Short-hand rules
  • Top-level command map
  • Commands reference
• Tutorials
• Examples
  • Example 0 - Basics: Tadah!MLIP
  • Example 1 - Basics: Training and Prediction
  • Example 2 - Basics: Nested Fitting Procedure
  • Example 3 - DM_mBlipEnv: Training and Prediction
  • Example 4 - Reading the Fit Balance Table
• Configuration File
  • Key/Value Pairs
  • Common Usage
  • Key Specifics
  • Comments
  • Key Values and Formats
  • Value Limits
  • Supported KEYS
• Training
  • Command
  • Pipeline
  • Load-time data transforms
  • Output files
  • Related pages
• What a Fit Minimises
  • The sum a fit minimises
  • Error scales
  • Priorities
  • LAMBDA and the coefficients
  • Data that spans a wide range
  • Converting the retired keys
• The Balance Table
  • Where the table is
  • The lines of the table
  • The columns
  • Reading a quantity the others outweigh
• Regressors
  • Overview
  • Quick syntax guide
  • Non-linear models
  • Reading the autogenerated tables
  • M_KRR
  • M_BLR
• Descriptors
  • General workflow
  • The bias term
  • Descriptor key syntax
  • Composite Descriptors
  • Quick examples
  • Reading the tables below
  • Two-Body
  • Many-Body
• Modular scalar embedding descriptor
  • Configuration
  • Bases and coordinates
  • Cutoffs and envelopes
  • Named nonlinear parameters
  • Refit, editing and analysis
  • Further candidates
• Cutoff Functions
  • Spatial cutoffs — f(r_cut) = 0
  • Inverse envelope — f(0) = 0, f(r_cut) = 1
• Refitting tabulated EAM potentials
  • Typical workflow
  • Quick start
  • How the fit works
  • Reading the result
  • Configuration reference
  • Output files
  • Dense mode
  • Editing a refitted potential (--edit)
  • Rewriting search bounds (--retarget)
  • Cutoffs: knot channels are derived, blip channels are fixed

---

Nested Fitting

• Overview
• Quick Start
• The Target File
• Block and Key Directory
• The OPTIMIZER Block
• Settings Reference
  • Search Parameters and Bounds
  • Targets and Losses
  • Search Controls
  • Output Settings
• Worked examples
• Further Nested Fitting Topics
• Reference
  • Top-Level Keys and Blocks
  • OPTIMIZER Block Keys
  • Targets, Losses and Search Parameters
  • Optimisers
  • Annealing PARAM Keys
  • ASA PARAM Keys

---