API reference

The MATCHA API is organised into a handful of subpackages, each covering a different layer of the toolkit. The full reference is generated automatically from the source docstrings — use the links below as entry points, and follow the matcha.* tree from there.

Subpackages

• matcha.sklearn — scikit-learn-style estimators (single models, ensembles, finetuners) that are the main user-facing API.

• matcha.torch — PyTorch Module and LightningModule implementations backing the estimators.

• matcha.nn — reusable neural network building blocks (layers, heads, losses, activations).

• matcha.datamodules — Lightning DataModules and dataset wrappers for molecular data.

• matcha.calibration — post-hoc calibration and uncertainty estimation utilities.

• matcha.explainability — attribution and interpretability methods.

• matcha.cli — command-line entry points.

• matcha.utils — shared helpers used across the package.

Full module index

For the complete auto-generated reference, see the top-level matcha package.