create_safetensors.py

import torch
from safetensors.torch import save_file

# Cascade of two XOR2 gates: parity(a,b,c) = XOR(XOR(a,b), c)
# Each XOR uses OR-NAND-AND structure

def xor_block(prefix):
    return {
        f'{prefix}.or.weight': torch.tensor([1.0, 1.0], dtype=torch.float32),
        f'{prefix}.or.bias': torch.tensor([-1.0], dtype=torch.float32),
        f'{prefix}.nand.weight': torch.tensor([-1.0, -1.0], dtype=torch.float32),
        f'{prefix}.nand.bias': torch.tensor([1.0], dtype=torch.float32),
        f'{prefix}.and.weight': torch.tensor([1.0, 1.0], dtype=torch.float32),
        f'{prefix}.and.bias': torch.tensor([-2.0], dtype=torch.float32),
    }

weights = {}
weights.update(xor_block('xor1'))
weights.update(xor_block('xor2'))

save_file(weights, 'model.safetensors')

def xor2(a, b, prefix):
    or_out = int(a * weights[f'{prefix}.or.weight'][0] + b * weights[f'{prefix}.or.weight'][1] + weights[f'{prefix}.or.bias'] >= 0)
    nand_out = int(a * weights[f'{prefix}.nand.weight'][0] + b * weights[f'{prefix}.nand.weight'][1] + weights[f'{prefix}.nand.bias'] >= 0)
    and_out = int(or_out * weights[f'{prefix}.and.weight'][0] + nand_out * weights[f'{prefix}.and.weight'][1] + weights[f'{prefix}.and.bias'] >= 0)
    return and_out

def parity3(a, b, c):
    xor_ab = xor2(a, b, 'xor1')
    return xor2(xor_ab, c, 'xor2')

print("Verifying parity3...")
errors = 0
for i in range(8):
    a, b, c = (i >> 2) & 1, (i >> 1) & 1, i & 1
    result = parity3(a, b, c)
    expected = a ^ b ^ c
    if result != expected:
        errors += 1
        print(f"ERROR: parity({a},{b},{c}) = {result}, expected {expected}")
if errors == 0:
    print("All 8 test cases passed!")
print(f"Magnitude: {sum(t.abs().sum().item() for t in weights.values()):.0f}")