matcha.sklearn.clm.base_sklearn_clm

Base class for sklearn-compatible chemical language model (CLM) wrappers.

Classes

BaseScikitLearnCLM

Base class for all sklearn-compatible CLM models.

Module Contents

class matcha.sklearn.clm.base_sklearn_clm.BaseScikitLearnCLM(params)[source]

Bases: matcha.sklearn.base_sklearn_model.BaseScikitLearnModel

Base class for all sklearn-compatible CLM models.

Not meant to be instantiated directly; serves as a parent class for each CLM model variant. Adapts BaseScikitLearnModel for chemical language inputs by configuring CLM-specific datamodules and training managers.

fit(x: list[rdkit.Chem.rdchem.Mol] | list[str] | torch.utils.data.StackDataset, y: numpy.ndarray | None = None, bound_mask: list[str] | None = None, validation_set: torch.utils.data.StackDataset | None = None)[source]

Runs training with the desired model architecture.

Parameters:

  • x (list[Mol] | StackDataset) – either a list of molecules or a StackDataset computed by the appropriate datamodule

  • y (np.ndarray | None) – property labels in a numpy array or None

  • bound_mask (list[str] | None) – list of str (N) defining whether the value is exact or is ‘less than’ / ‘more than’

  • validation_set (StackDataset | None) – pre-transformed data for early-stopping