transformer_lens.conversion_utils.conversion_steps.arithmetic_tensor_conversion module¶

Weight conversion that performs arithmetic operations on weights.

class transformer_lens.conversion_utils.conversion_steps.arithmetic_tensor_conversion.ArithmeticTensorConversion(operation: OperationTypes, value: float | int | Tensor, input_filter: Callable | None = None, output_filter: Callable | None = None)¶

Bases: BaseTensorConversion

handle_conversion(input_value, *full_context)¶

revert(input_value, *full_context)¶: Revert the arithmetic operation (apply inverse operation).

class transformer_lens.conversion_utils.conversion_steps.arithmetic_tensor_conversion.OperationTypes(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)¶

Bases: Enum

ADDITION = 0¶

DIVISION = 3¶

MULTIPLICATION = 2¶

SUBTRACTION = 1¶