How to add custom bitwise XOR operator?

[binseq]

I want to create a bitwise operators like in the ^ operator in python. I've tried many things, and asking AI nothing works. I keep getting MethodError. Copilot gave me the following that doesn't work...

# Custom bitwise XOR operator function
def bitwise_xor_operator(x1, x2):
    # Convert inputs to integers and use absolute values
    x1_int = int(abs(x1))
    x2_int = int(abs(x2))
    return x1_int ^ x2_int

from pysr import PySRRegressor

# Define the PySRRegressor with the custom bitwise XOR operator
model = PySRRegressor(
    niterations=1000,
    binary_operators=["+", "-", "*", "/", "xor"],  # Add "xor" to the list of binary operators
    unary_operators=[],
    extra_sympy_mappings={"xor": lambda x, y: x ^ y},  # Define how the operator should be represented in SymPy
    extra_torch_mappings={"xor": bitwise_xor_operator},  # Use the custom bitwise XOR operator function defined earlier
    extra_jax_mappings={"xor": bitwise_xor_operator},  # Use the custom bitwise XOR operator function defined earlier
)

# Example usage with non-negative inputs
X = [[1, 2], [3, 4], [5, 6]]
y = [1, 0, 1]

# Run the symbolic regression
model.fit(X, y)

[MilesCranmer]

Maybe try this?

"my_xor(x, y) = convert(typeof(x), xor(x > 0, y > 0))"

i.e.,

model = PySRRegressor(
    niterations=1000,
    binary_operators=["+", "-", "*", "/", "my_xor(x, y) = convert(typeof(x), xor(x > 0, y > 0))"],
    unary_operators=[],
    extra_sympy_mappings={"my_xor": lambda x, y: x ^ y},  # Define how the operator should be represented in SymPy
)

logical_or and logical_and are built-in: https://ai.damtp.cam.ac.uk/pysr/operators

[binseq]

Thank you!