chipiron.players.boardevaluators.neural_networks package

Subpackages

• chipiron.players.boardevaluators.neural_networks.input_converters package
  • Submodules
  • chipiron.players.boardevaluators.neural_networks.input_converters.ModelInputRepresentationType module
    • InternalTensorRepresentationType
      • InternalTensorRepresentationType.BUG364
      • InternalTensorRepresentationType.NO
      • InternalTensorRepresentationType.NOBUG364
    • ModelInputRepresentationType
      • ModelInputRepresentationType.BOARD_PIECES_TWO_SIDES
      • ModelInputRepresentationType.BUG364
      • ModelInputRepresentationType.NOBUG364
      • ModelInputRepresentationType.PIECE_DIFFERENCE
      • ModelInputRepresentationType.PIECE_MAP
    • assert_compatibilities_representation_type()
    • get_default_internal_representation()
  • chipiron.players.boardevaluators.neural_networks.input_converters.board_representation module
    • BoardRepresentation
      • BoardRepresentation.get_evaluator_input()
    • Representation364
      • Representation364.get_evaluator_input()
      • Representation364.tensor_black
      • Representation364.tensor_castling_black
      • Representation364.tensor_castling_white
      • Representation364.tensor_white
    • Representation364_2
      • Representation364_2.get_evaluator_input()
      • Representation364_2.tensor
  • chipiron.players.boardevaluators.neural_networks.input_converters.board_to_input module
    • BoardToInput
      • BoardToInput.convert()
    • BoardToInputFunction
      • BoardToInputFunction.__call__()
    • create_board_to_input()
    • create_board_to_input_from_representation()
  • chipiron.players.boardevaluators.neural_networks.input_converters.board_to_transformer_input module
    • build_transformer_input()
  • chipiron.players.boardevaluators.neural_networks.input_converters.factory module
    • CreateFromBoard
    • CreateFromBoardAndFromParent
    • RepresentationFactory
      • RepresentationFactory.create_from_board
      • RepresentationFactory.create_from_board_and_from_parent
      • RepresentationFactory.create_from_transition()
  • chipiron.players.boardevaluators.neural_networks.input_converters.rep_364 module
    • a()
    • b()
    • c()
    • create_from_board()
    • create_from_board_and_from_parent()
    • d()
    • e()
  • chipiron.players.boardevaluators.neural_networks.input_converters.rep_364_bug module
    • a()
    • b()
    • c()
    • create_from_board()
    • create_from_board_and_from_parent()
    • d()
    • e()
  • chipiron.players.boardevaluators.neural_networks.input_converters.representation_364_bti module
    • RepresentationBTI
      • RepresentationBTI.convert()
  • chipiron.players.boardevaluators.neural_networks.input_converters.representation_factory_factory module
    • create_board_representation_factory()
  • chipiron.players.boardevaluators.neural_networks.input_converters.test_representation module
  • Module contents
• chipiron.players.boardevaluators.neural_networks.models package
  • Submodules
  • chipiron.players.boardevaluators.neural_networks.models.multi_layer_perceptron module
    • __init__()
    • print_param()
    • MultiLayerPerceptron
      • MultiLayerPerceptron.__getstate__()
      • MultiLayerPerceptron.forward()
      • MultiLayerPerceptron.init_weights()
      • MultiLayerPerceptron.log_readable_model_weights_to_file()
      • MultiLayerPerceptron.print_input()
      • MultiLayerPerceptron.print_param()
    • MultiLayerPerceptronArgs
      • MultiLayerPerceptronArgs.filename()
      • MultiLayerPerceptronArgs.list_of_activation_functions
      • MultiLayerPerceptronArgs.number_neurons_per_layer
      • MultiLayerPerceptronArgs.type
    • build_sequential()
    • extract_sequential_model_data()
    • print_piece_param()
  • chipiron.players.boardevaluators.neural_networks.models.transformer_one module
    • Block
      • Block.forward()
    • FeedFoward
      • FeedFoward.forward()
    • Head
      • Head.forward()
    • MultiHeadAttention
      • MultiHeadAttention.forward()
    • TransformerArgs
      • TransformerArgs.__str__()
      • TransformerArgs.dropout_ratio
      • TransformerArgs.filename()
      • TransformerArgs.len_square_tensor
      • TransformerArgs.n_embd
      • TransformerArgs.n_head
      • TransformerArgs.n_layer
      • TransformerArgs.number_occupancy_types
      • TransformerArgs.number_pieces_types
      • TransformerArgs.type
    • TransformerOne
      • TransformerOne.compute_representation()
      • TransformerOne.forward()
      • TransformerOne.get_nn_input()
      • TransformerOne.print_param()
  • Module contents
• chipiron.players.boardevaluators.neural_networks.output_converters package
  • Submodules
  • chipiron.players.boardevaluators.neural_networks.output_converters.factory module
    • create_output_converter()
  • chipiron.players.boardevaluators.neural_networks.output_converters.model_output_type module
    • ModelOutputType
      • ModelOutputType.point_of_view
  • chipiron.players.boardevaluators.neural_networks.output_converters.output_value_converter module
    • IdentityConverter
      • IdentityConverter.from_value_white_to_model_output()
      • IdentityConverter.to_board_evaluation()
    • OutputValueConverter
      • OutputValueConverter.from_value_white_to_model_output()
      • OutputValueConverter.to_board_evaluation()
    • PlayerToMoveValueToValueWhiteConverter
      • PlayerToMoveValueToValueWhiteConverter.convert_value_from_mover_viewpoint_to_value_white()
      • PlayerToMoveValueToValueWhiteConverter.from_value_white_to_model_output()
      • PlayerToMoveValueToValueWhiteConverter.to_board_evaluation()
  • Module contents

Submodules

chipiron.players.boardevaluators.neural_networks.NNModelType module

class chipiron.players.boardevaluators.neural_networks.NNModelType.ActivationFunctionType(value, names=<not given>, *values, module=None, qualname=None, type=None, start=1, boundary=None)

Bases: str, Enum

ParametricRelu = 'parametric_relu'

Relu = 'relu'

TangentHyperbolic = 'hyperbolic_tangent'

class chipiron.players.boardevaluators.neural_networks.NNModelType.NNModelType(value, names=<not given>, *values, module=None, qualname=None, type=None, start=1, boundary=None)

Bases: str, Enum

MultiLayerPerceptron = 'multi_layer_perceptron'

Transformer = 'transformer'

chipiron.players.boardevaluators.neural_networks.NNModelTypeArgs module

chipiron.players.boardevaluators.neural_networks.board_to_tensor module

Module to transform the board into a tensor representation.

chipiron.players.boardevaluators.neural_networks.board_to_tensor.get_tensor_from_tensors(tensor_white: Tensor, tensor_black: Tensor, tensor_castling_white: Tensor, tensor_castling_black: Tensor, color_to_play: bool) → Tensor

Get the final tensor representation from individual tensors.

Parameters:

• tensor_white (torch.Tensor) – The tensor representation for white pieces.

• tensor_black (torch.Tensor) – The tensor representation for black pieces.

• tensor_castling_white (torch.Tensor) – The tensor representation for white castling.

• tensor_castling_black (torch.Tensor) – The tensor representation for black castling.

• color_to_play (chess.Color) – The color to play.

Returns:

The final tensor representation.

Return type:

torch.Tensor

chipiron.players.boardevaluators.neural_networks.board_to_tensor.transform_board_pieces_one_side(board: IBoard, requires_grad_bool: bool) → Tensor

Transform the board pieces for one side into a tensor representation.

Parameters:

• board (BoardChi) – The chess board.

• requires_grad_bool (bool) – Whether the tensor requires gradient.

Returns:

The transformed board pieces tensor.

Return type:

torch.Tensor

chipiron.players.boardevaluators.neural_networks.board_to_tensor.transform_board_pieces_two_sides(board: IBoard, requires_grad_bool: bool) → Tensor

Transform the board pieces for both sides into a tensor representation.

Parameters:

• board (BoardChi) – The chess board.

• requires_grad_bool (bool) – Whether the tensor requires gradient.

Returns:

The transformed board pieces tensor.

Return type:

torch.Tensor

chipiron.players.boardevaluators.neural_networks.factory module

Module for creating neural networks and neural network board evaluators.

class chipiron.players.boardevaluators.neural_networks.factory.NeuralNetModelsAndArchitecture(model_weights_file_name: str | PathLike[str], nn_architecture_args: NeuralNetArchitectureArgs)

Bases: object

Class to hold the neural network models and architecture. :ivar model_weights_file_name: The file name of the model weights. :vartype model_weights_file_name: path :ivar nn_architecture_args: The neural network architecture arguments.

classmethod build_from_folder_path(folder_path: str | PathLike[str]) → NeuralNetModelsAndArchitecture

Build an instance of NeuralNetModelsAndArchitecture from the given folder path.

Parameters:

folder_path (Path) – Path to the folder containing ‘architecture.yaml’ and model weights.

Returns:

An initialized instance.

Return type:

NeuralNetModelsAndArchitecture

model_weights_file_name: str | PathLike[str]

nn_architecture_args: NeuralNetArchitectureArgs

chipiron.players.boardevaluators.neural_networks.factory.create_nn(nn_type_args: MultiLayerPerceptronArgs | TransformerArgs) → ChiNN

Create a neural network.

chipiron.players.boardevaluators.neural_networks.factory.create_nn_board_eval_from_architecture_args(nn_architecture_args: NeuralNetArchitectureArgs) → NNBoardEvaluator

chipiron.players.boardevaluators.neural_networks.factory.create_nn_board_eval_from_folder_path_and_existing_model(path_to_nn_folder: str | PathLike[str]) → tuple[chipiron.players.boardevaluators.neural_networks.nn_board_evaluator.NNBoardEvaluator, chipiron.players.boardevaluators.neural_networks.neural_net_board_eval_args.NeuralNetArchitectureArgs]

Create a neural network board evaluator.

Parameters:

path_to_nn_folder (path) – the path to the folder where the model is defined.

Returns:

The created neural network board evaluator.

Return type:

NNBoardEvaluator

chipiron.players.boardevaluators.neural_networks.factory.create_nn_board_eval_from_nn_and_architecture_args(nn_architecture_args: NeuralNetArchitectureArgs, nn: ChiNN) → NNBoardEvaluator

chipiron.players.boardevaluators.neural_networks.factory.create_nn_board_eval_from_nn_parameters_file_and_existing_model(model_weights_file_name: str | PathLike[str], nn_architecture_args: NeuralNetArchitectureArgs) → NNBoardEvaluator

chipiron.players.boardevaluators.neural_networks.factory.create_nn_from_folder_path_and_existing_model(folder_path: str | PathLike[str]) → tuple[chipiron.utils.chi_nn.ChiNN, chipiron.players.boardevaluators.neural_networks.neural_net_board_eval_args.NeuralNetArchitectureArgs]

chipiron.players.boardevaluators.neural_networks.factory.create_nn_from_param_path_and_architecture_args(model_weights_file_name: str | PathLike[str], nn_architecture_args: NeuralNetArchitectureArgs) → tuple[chipiron.utils.chi_nn.ChiNN, chipiron.players.boardevaluators.neural_networks.neural_net_board_eval_args.NeuralNetArchitectureArgs]

chipiron.players.boardevaluators.neural_networks.factory.get_architecture_args_from_file(architecture_file_name: str | PathLike[str]) → NeuralNetArchitectureArgs

chipiron.players.boardevaluators.neural_networks.factory.get_architecture_args_from_folder(folder_path: str | PathLike[str]) → NeuralNetArchitectureArgs

chipiron.players.boardevaluators.neural_networks.factory.get_nn_architecture_file_path_from(folder_path: str | PathLike[str]) → str

Get the file path for the architecture parameters.

Parameters:

folder_path (str) – The folder path for the architecture parameters.

Returns:

The file path for the architecture parameters.

Return type:

str

chipiron.players.boardevaluators.neural_networks.factory.get_nn_param_file_path_from(folder_path: str | PathLike[str], file_name: str | None = None) → tuple[str, str]

Get the file path for the neural network parameters.

Parameters:

folder_path (str) – The folder path for the neural network parameters.

Returns:

The file path for the neural network parameters.

Return type:

str

chipiron.players.boardevaluators.neural_networks.neural_net_board_eval_args module

Module that contains the NeuralNetBoardEvalArgs class.

class chipiron.players.boardevaluators.neural_networks.neural_net_board_eval_args.NeuralNetArchitectureArgs(model_type_args: chipiron.players.boardevaluators.neural_networks.models.multi_layer_perceptron.MultiLayerPerceptronArgs | chipiron.players.boardevaluators.neural_networks.models.transformer_one.TransformerArgs, model_input_representation_type: chipiron.players.boardevaluators.neural_networks.input_converters.ModelInputRepresentationType.ModelInputRepresentationType, model_output_type: chipiron.players.boardevaluators.neural_networks.output_converters.model_output_type.ModelOutputType)

Bases: object

filename() → str

model_input_representation_type: ModelInputRepresentationType

model_output_type: ModelOutputType

model_type_args: MultiLayerPerceptronArgs | TransformerArgs

class chipiron.players.boardevaluators.neural_networks.neural_net_board_eval_args.NeuralNetBoardEvalArgs(nn_type: NNModelType, nn_param_folder_name: str)

Bases: object

Represents the arguments for a neural network board evaluator.

Variables:

• nn_type (str) – The type of the neural network.

• nn_param_folder_name (str) – The name of the folder containing the neural network parameters.

nn_param_folder_name: str

nn_type: NNModelType

chipiron.players.boardevaluators.neural_networks.nn_board_evaluator module

Module for the Neural Network Board Evaluator

class chipiron.players.boardevaluators.neural_networks.nn_board_evaluator.NNBoardEvaluator(net: ChiNN, output_and_value_converter: OutputValueConverter, board_to_input_convert: BoardToInputFunction)

Bases: object

The Generic Neural network class for board evaluation

Variables:

• net (ChiNN) – The neural network model

• output_and_value_converter (OutputValueConverter) – The converter for output values

• board_to_input_converter (BoardToInputFunction) – The converter for board to input tensor

board_to_input_convert: BoardToInputFunction

evaluate(input_layer: Tensor, color_to_play: bool) → FloatyBoardEvaluation

Evaluate the board position

Parameters:

• input_layer (torch.Tensor) – The input tensor representing the board position

• color_to_play (chess.Color) – The color to play

Returns:

The evaluation of the board position

Return type:

FloatyBoardEvaluation

net: ChiNN

output_and_value_converter: OutputValueConverter

value_white(board: IBoard) → float

Evaluate the value for the white player

Parameters:

board (BoardChi) – The chess board

Returns:

The value for the white player

Return type:

float

Module contents

Module for neural network board evaluators.

class chipiron.players.boardevaluators.neural_networks.NNBoardEvaluator(net: ChiNN, output_and_value_converter: OutputValueConverter, board_to_input_convert: BoardToInputFunction)

Bases: object

The Generic Neural network class for board evaluation

Variables:

• net (ChiNN) – The neural network model

• output_and_value_converter (OutputValueConverter) – The converter for output values

• board_to_input_converter (BoardToInputFunction) – The converter for board to input tensor

board_to_input_convert: BoardToInputFunction

evaluate(input_layer: Tensor, color_to_play: bool) → FloatyBoardEvaluation

Evaluate the board position

Parameters:

• input_layer (torch.Tensor) – The input tensor representing the board position

• color_to_play (chess.Color) – The color to play

Returns:

The evaluation of the board position

Return type:

FloatyBoardEvaluation

net: ChiNN

output_and_value_converter: OutputValueConverter

value_white(board: IBoard) → float

Evaluate the value for the white player

Parameters:

board (BoardChi) – The chess board

Returns:

The value for the white player

Return type:

float

class chipiron.players.boardevaluators.neural_networks.NeuralNetBoardEvalArgs(nn_type: NNModelType, nn_param_folder_name: str)

Bases: object

Represents the arguments for a neural network board evaluator.

Variables:

• nn_type (str) – The type of the neural network.

• nn_param_folder_name (str) – The name of the folder containing the neural network parameters.

nn_param_folder_name: str

nn_type: NNModelType