chipiron.players.boardevaluators.neural_networks.output_converters package

Submodules

chipiron.players.boardevaluators.neural_networks.output_converters.factory module

chipiron.players.boardevaluators.neural_networks.output_converters.factory.create_output_converter(model_output_type: ModelOutputType) → OutputValueConverter[source]

chipiron.players.boardevaluators.neural_networks.output_converters.model_output_type module

class chipiron.players.boardevaluators.neural_networks.output_converters.model_output_type.ModelOutputType(point_of_view: chipiron.players.boardevaluators.board_evaluation.board_evaluation.PointOfView)[source]

Bases: object

point_of_view: PointOfView

chipiron.players.boardevaluators.neural_networks.output_converters.output_value_converter module

Module for converting the output of the neural network to a board evaluation

class chipiron.players.boardevaluators.neural_networks.output_converters.output_value_converter.IdentityConverter[source]

Bases: OutputValueConverter

Converting from a NN that outputs a 1D value from the point of view of the player to move

from_value_white_to_model_output(board_value_white: float, board: IBoard) → Tensor[source]

This functions takes the value white and converts to the corresponding value from the NN model output. Remember some NN models output value_from_mover for instance This function is used in training where the value white is compared to taget value from datasets that are float.

to_board_evaluation(output_nn: Tensor, color_to_play: bool) → FloatyBoardEvaluation[source]

Convert the output of the neural network to a board evaluation.

Parameters:

  • output_nn (torch.Tensor) – The output of the neural network.

  • color_to_play (chess.Color) – The color of the player to move.

Returns:

The converted board evaluation.

Return type:

FloatyBoardEvaluation

class chipiron.players.boardevaluators.neural_networks.output_converters.output_value_converter.OutputValueConverter[source]

Bases: ABC

Converting an output of the neural network to a board evaluation and conversely converting a board evaluation to an output of the neural network

abstract from_value_white_to_model_output(board_value_white: float, board: IBoard) → Tensor[source]

This functions takes the value white and converts to the corresponding value from the NN model output. Remember some NN models output value_from_mover for instance This function is used in training where the value white is compared to taget value from datasets that are float.

abstract to_board_evaluation(output_nn: Tensor, color_to_play: bool) → FloatyBoardEvaluation[source]

Convert the output of the neural network to a board evaluation.

Parameters:

  • output_nn (torch.Tensor) – The output of the neural network.

  • color_to_play (chess.Color) – The color of the player to move.

Returns:

The converted board evaluation.

Return type:

FloatyBoardEvaluation

class chipiron.players.boardevaluators.neural_networks.output_converters.output_value_converter.PlayerToMoveValueToValueWhiteConverter[source]

Bases: OutputValueConverter

Converting from a NN that outputs a 1D value from the point of view of the player to move

convert_value_from_mover_viewpoint_to_value_white(turn: bool, value_from_mover_view_point: float) → float[source]

Convert the value from the mover’s viewpoint to the value from the white player’s viewpoint.

Parameters:

  • turn (chess.Color) – The color of the player to move.

  • value_from_mover_view_point (float) – The value from the mover’s viewpoint.

Returns:

The value from the white player’s viewpoint.

Return type:

float

from_value_white_to_model_output(board_value_white: float, board: IBoard) → Tensor[source]

This functions takes the value white and converts to the corresponding value from the NN model output. Remember some NN models output value_from_mover for instance This function is used in training where the value white is compared to taget value from datasets that are float.

to_board_evaluation(output_nn: Tensor, color_to_play: bool) → FloatyBoardEvaluation[source]

Convert the output of the neural network to a board evaluation.

Parameters:

  • output_nn (torch.Tensor) – The output of the neural network.

  • color_to_play (chess.Color) – The color of the player to move.

Returns:

The converted board evaluation.

Return type:

FloatyBoardEvaluation

Module contents