"""
Threshold Network for 3-input XOR Gate

Cascade of two standard XORs (OR + NAND + AND structure).
"""

import torch
from safetensors.torch import load_file


class ThresholdXOR3:
    def __init__(self, weights_dict):
        self.w = weights_dict

    def __call__(self, a, b, c):
        inp1 = torch.tensor([float(a), float(b)])
        or1 = int((inp1 * self.w['xor1.layer1.or.weight']).sum() + self.w['xor1.layer1.or.bias'] >= 0)
        nand1 = int((inp1 * self.w['xor1.layer1.nand.weight']).sum() + self.w['xor1.layer1.nand.bias'] >= 0)
        h1 = torch.tensor([float(or1), float(nand1)])
        xor_ab = int((h1 * self.w['xor1.layer2.and.weight']).sum() + self.w['xor1.layer2.and.bias'] >= 0)

        inp2 = torch.tensor([float(xor_ab), float(c)])
        or2 = int((inp2 * self.w['xor2.layer1.or.weight']).sum() + self.w['xor2.layer1.or.bias'] >= 0)
        nand2 = int((inp2 * self.w['xor2.layer1.nand.weight']).sum() + self.w['xor2.layer1.nand.bias'] >= 0)
        h2 = torch.tensor([float(or2), float(nand2)])
        out = int((h2 * self.w['xor2.layer2.and.weight']).sum() + self.w['xor2.layer2.and.bias'] >= 0)

        return float(out)

    @classmethod
    def from_safetensors(cls, path="model.safetensors"):
        return cls(load_file(path))


if __name__ == "__main__":
    weights = load_file("model.safetensors")
    model = ThresholdXOR3(weights)

    print("3-input XOR Gate:")
    correct = 0
    for a in [0, 1]:
        for b in [0, 1]:
            for c in [0, 1]:
                out = int(model(a, b, c))
                expected = a ^ b ^ c
                if out == expected:
                    correct += 1
                status = "OK" if out == expected else "FAIL"
                print(f"  XOR3({a},{b},{c}) = {out} [{status}]")
    print(f"Total: {correct}/8")