MixtureDatapoint

class cosmolayer.MixtureDatapoint(temperature, mole_fractions, areas, volumes, probabilities, targets)

Base dataclass for a mixture datapoint.

Stores physical inputs (temperature, mole fractions, areas, volumes, and segment-type probabilities) and optional training targets. Shape metadata is computed and validated automatically on construction.

Parameters:

• temperature (float) – Temperature.

• mole_fractions (NumpyArray1D) – Mole fractions. Shape: (num_components,).

• areas (NumpyArray1D) – Segment surface areas per component. Shape: (num_components,).

• volumes (NumpyArray1D) – Molar volumes per component. Shape: (num_components,).

• probabilities (NumpyArray2D) – Sigma-profile probabilities. Shape: (num_components, num_segment_types).

• targets (NumpyArray1D) – Training targets. Shape: (num_targets,).

Variables:

• num_components (int) – Number of components.

• num_segment_types (int) – Number of segment-type probabilities.

• num_targets (int) – Number of training targets.

Raises:

ValueError – If array shapes are inconsistent.

Attributes

shape

Return the structural shape metadata for the datapoint.

Returns:

Tuple containing the number of components, segment types, and number of training targets.

Return type:

tuple[int, int, int]

Methods

get_inputs(dtype=torch.float64)

Convert physical inputs to torch tensors.

Parameters:

dtype (torch.dtype) – Data type used for all returned tensors. Default is torch.float64.

Returns:

Temperature, mole fractions, areas, volumes, and probabilities as torch tensors.

Return type:

InputsType

get_targets(dtype=torch.float64)

Convert target arrays to torch tensors.

Parameters:

dtype (torch.dtype) – Data type used for all returned tensors. Default is torch.float64.

Returns:

Training targets as torch tensors.

Return type:

Tensor1D