create_safetensors.py

import torch
from safetensors.torch import save_file

# 7-bit parity using tree of XOR gates
# Structure: parity7 = XOR(XOR(XOR(x0,x1), XOR(x2,x3)), XOR(XOR(x4,x5), x6))
# 6 XOR2 gates total

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 = {}
# Level 1: XOR pairs
weights.update(xor_block('xor_01'))    # XOR(x0, x1)
weights.update(xor_block('xor_23'))    # XOR(x2, x3)
weights.update(xor_block('xor_45'))    # XOR(x4, x5)
# Level 2: Combine
weights.update(xor_block('xor_0123'))  # XOR(xor01, xor23)
weights.update(xor_block('xor_456'))   # XOR(xor45, x6)
# Level 3: Final
weights.update(xor_block('xor_final')) # XOR(xor0123, xor456)

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)
    return int(or_out * weights[f'{prefix}.and.weight'][0] + nand_out * weights[f'{prefix}.and.weight'][1] + weights[f'{prefix}.and.bias'] >= 0)

def parity7(x0, x1, x2, x3, x4, x5, x6):
    xor01 = xor2(x0, x1, 'xor_01')
    xor23 = xor2(x2, x3, 'xor_23')
    xor45 = xor2(x4, x5, 'xor_45')
    xor0123 = xor2(xor01, xor23, 'xor_0123')
    xor456 = xor2(xor45, x6, 'xor_456')
    return xor2(xor0123, xor456, 'xor_final')

print("Verifying parity7...")
errors = 0
for i in range(128):
    bits = [(i >> j) & 1 for j in range(7)]
    result = parity7(*bits)
    expected = sum(bits) % 2
    if result != expected:
        errors += 1
        print(f"ERROR: parity({bits}) = {result}, expected {expected}")

if errors == 0:
    print("All 128 test cases passed!")
else:
    print(f"FAILED: {errors} errors")

print(f"Magnitude: {sum(t.abs().sum().item() for t in weights.values()):.0f}")
print(f"Parameters: {sum(t.numel() for t in weights.values())}")