Overview
QML is a Python toolkit for quantum machine learning, with quantum chemistry also identified in its package metadata. It is a software toolkit rather than a supplied predictive model or hosted service. The repository is now archived: its README directs further development, issues and pull requests to qmllib, while retaining the master and develop branches as an archive. This makes QML principally relevant to inspecting older implementations or evaluating existing workflows that depend on it.
The build configuration exposes a Python package with compiled Fortran components for representations, kernels, distances and mathematical solvers. It also lists FCHL and ARAD kernel components, SLATM and ACSF representation components, and specialized gradient, force and electric-field kernel source files. These components indicate a role within representation-based scientific machine-learning workflows. However, the source excerpts do not document callable interfaces, accepted molecular input formats, array layouts or returned output structures; those details need inspection before a concrete data-processing pipeline can be specified.
The setup source uses numpy.distutils and includes compiler and numerical-library linking configuration, including BLAS/LAPACK and conditional MKL handling. These are build details, not evidence of successful installation or current platform compatibility. The README describes qmllib as a streamlined continuation of QML's Fortran core with fewer dependencies and NumPy 2.0 compatibility; that compatibility statement applies to qmllib, not archived QML. No benchmark results, prediction accuracy, training datasets or complete end-to-end examples are supplied here.
Key Features
- Python package structure covering representations, kernels, mathematical routines, models and workflow-related modules.
- Compiled Fortran extensions for kernel calculations, distances and mathematical solvers.
- Representation extension sources include SLATM and ACSF components.
- FCHL extension sources include scalar, force and electric-field kernel components; separate gradient-kernel and ARAD-kernel extensions are also listed.
- Build configuration includes BLAS/LAPACK linking and conditional MKL handling.
Use Cases
- Intended evaluation: inspect the archived implementation when maintaining or reproducing an existing workflow that depends on QML.
- Intended evaluation: examine representation and kernel components as possible building blocks for a quantum-chemistry machine-learning pipeline, after verifying their interfaces and data requirements.
- Intended evaluation: assess migration from archived QML to qmllib by comparing the functions required by an existing project with the successor's documented scope.
How to Use
- Read the archive notice first. Decide whether your task requires historical QML code or should instead target its continuation.
- Inspect the retained master and develop branches in the QML repository. Record which branch and revision your existing workflow depends on rather than assuming the archive is actively maintained.
- Review setup.py to identify relevant representation, kernel or solver extensions and their compiler and numerical-library dependencies. The excerpt does not establish a working installation procedure.
- Before evaluating your own data, locate the selected component's interface in the repository and verify input formats, output structures and any examples. These details are not specified in the source excerpts.
- Check the repository code licence, and consult qmllib for future development or issue reporting. Treat its NumPy 2.0 statement as successor-specific, not a compatibility guarantee for QML.