File size: 1,642 Bytes
1c1d9aa | 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 | import torch, sys, json, math
sys.path.insert(0, "..")
from src.moo import cagrad_update_k2, raco_update_k2, _grads_dot, _solve_lambda_k2
seeds = [0, 1, 7, 13, 42, 99, 123, 256]
w1, w2 = 0.8, 0.2
results = []
for seed in seeds:
for c in [0.3, 0.5, 0.7]:
torch.manual_seed(seed)
g1 = torch.randn(100) * 3.0
g2 = torch.randn(100) * 2.0
H11 = _grads_dot([g1], [g1])
H12 = _grads_dot([g1], [g2])
H22 = _grads_dot([g2], [g2])
g0_norm_sq = (w1 * w1) * H11 + (2.0 * w1 * w2) * H12 + (w2 * w2) * H22
g0_norm = torch.sqrt(torch.clamp(g0_norm_sq, min=0.0))
b1 = (w1 * H11) + (w2 * H12)
b2 = (w1 * H12) + (w2 * H22)
s_val = float(g0_norm) * c
lam_raw, _ = _solve_lambda_k2(b1, b2, H11, H12, H22, s_val)
p2_raw = 1.0 - float(lam_raw)
if p2_raw > w2:
cagrad_g = cagrad_update_k2([g1], [g2], w1, w2, c=c)
raco_g = raco_update_k2([g1], [g2], w1, w2, c=c)
ca_alignment_g2 = float((cagrad_g[0] @ g2).item())
ra_alignment_g2 = float((raco_g[0] @ g2).item())
results.append({
"seed": seed, "c": c,
"p2_raw": p2_raw, "w2": w2,
"overcorrection": p2_raw > w2,
"cagrad_g2_improvement": ca_alignment_g2,
"raco_g2_improvement": ra_alignment_g2,
"raco_reduces_overcorrection": abs(ra_alignment_g2) < abs(ca_alignment_g2),
})
for r in results[:5]:
print(json.dumps(r, indent=2))
print(f"\nFound {len(results)} cases with overcorrection out of {len(seeds) * 3} attempts")
|