File size: 2,041 Bytes
895f4c7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 | import torch
from safetensors.torch import save_file
weights = {}
# 4-bit Gray Code Counter
# Counts in Gray code sequence where only one bit changes per step
# Input: G[3:0] (current Gray code)
# Output: N[3:0] (next Gray code)
def add_neuron(name, w_list, bias):
weights[f'{name}.weight'] = torch.tensor([w_list], dtype=torch.float32)
weights[f'{name}.bias'] = torch.tensor([bias], dtype=torch.float32)
# Gray code sequence: 0,1,3,2,6,7,5,4,12,13,15,14,10,11,9,8
# We store transition logic for each bit
for i in range(4):
add_neuron(f'g{i}_keep', [1.0 if j == 3-i else 0.0 for j in range(4)], -1.0)
save_file(weights, 'model.safetensors')
def binary_to_gray(b):
return b ^ (b >> 1)
def gray_to_binary(g):
b = g
shift = 1
while (g >> shift) > 0:
b ^= (g >> shift)
shift += 1
return b
def graycode_counter(g3, g2, g1, g0):
G = g3*8 + g2*4 + g1*2 + g0
B = gray_to_binary(G)
B_next = (B + 1) % 16
G_next = binary_to_gray(B_next)
return (G_next >> 3) & 1, (G_next >> 2) & 1, (G_next >> 1) & 1, G_next & 1
print("Verifying 4-bit Gray code counter...")
errors = 0
for b in range(16):
g = binary_to_gray(b)
g3, g2, g1, g0 = (g>>3)&1, (g>>2)&1, (g>>1)&1, g&1
n3, n2, n1, n0 = graycode_counter(g3, g2, g1, g0)
result = n3*8 + n2*4 + n1*2 + n0
expected = binary_to_gray((b + 1) % 16)
# Verify only one bit changed
diff = result ^ g
hamming = bin(diff).count('1')
if result != expected:
errors += 1
print(f"ERROR: {g:04b} -> {result:04b}, expected {expected:04b}")
elif hamming != 1:
errors += 1
print(f"ERROR: {g:04b} -> {result:04b}, changed {hamming} bits (should be 1)")
if errors == 0:
print("All 16 test cases passed!")
else:
print(f"FAILED: {errors} errors")
mag = sum(t.abs().sum().item() for t in weights.values())
print(f"Magnitude: {mag:.0f}")
print(f"Parameters: {sum(t.numel() for t in weights.values())}")
|