kosmos.ml.config.factories.model

Classes

class ModelConfig

Bases: abc.ABC

Model configuration.


Methods

get_instance(input_dim: int, output_dim: int) → kosmos.ml.models.model.Model

Get the model instance.

Parameters:

  • input_dim (int) – Model input dimension.

  • output_dim (int) – Model output dimension.

Returns:

Model instance.

Return type:

Model

class NeuralNetworkConfig(hidden_layers: list[int])

Bases: ModelConfig

Neural network configuration.

Initialize the neural network configuration.

Parameters:

hidden_layers (list[int]) – Sizes of the hidden layers. The length of the list defines the number of hidden layers, and each element specifies the size of a layer.


Methods

get_instance(input_dim: int, output_dim: int) → kosmos.ml.models.neural_network.NeuralNetwork

Get the neural network instance.

Parameters:

  • input_dim (int) – Model input dimension.

  • output_dim (int) – Model output dimension.

Returns:

Neural network instance.

Return type:

NeuralNetwork

class VQCConfig(num_layers: int, encoding_config: kosmos.ml.config.factories.encoding.EncodingConfig, weight_mapping_func: kosmos.ml.typing.TensorMapping | None = None, input_mapping_func: kosmos.ml.typing.TensorMapping | None = None, weight_init_range: tuple[float, float] = (-1.0, 1.0), bias_init_range: tuple[float, float] | None = (-0.001, 0.001), *, data_reuploading: bool = False, output_scaling: bool = False)

Bases: ModelConfig

Variational quantum circuit configuration.

Initialize the VQC configuration.

Parameters:

  • num_layers (int) – The number of variational layers.

  • encoding_config (EncodingConfig) – The encoding configuration.

  • weight_mapping_func (TensorMapping | None) – The mapping function for the weights. Defaults to None.

  • input_mapping_func (TensorMapping | None) – The mapping function for the input. Defaults to None.

  • weight_init_range (tuple[float, float]) – Lower and upper bounds for initializing the trainable weight parameters. Defaults to (-1.0, 1.0).

  • bias_init_range (tuple[float, float] | None) – Lower and upper bounds for initializing the trainable bias parameters applied to each output unit. If None, no bias parameters are used. Defaults to (-0.001, 0.001).

  • data_reuploading (bool) – Whether to use data re-uploading. Defaults to False.

  • output_scaling (bool) – Whether to use output scaling. Defaults to False.


Methods

get_instance(input_dim: int, output_dim: int) → kosmos.ml.models.vqc.vqc.VQC

Get the VQC instance.

Parameters:

  • input_dim (int) – Model input dimension.

  • output_dim (int) – Model output dimension.

Returns:

VQC instance.

Return type:

VQC