File size: 5,211 Bytes
47709ed | 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 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 | #!/usr/bin/env python3
"""CUDA static, gradient, and dequantization checks for QH032."""
from __future__ import annotations
import json
import sys
from datetime import datetime, timezone
from pathlib import Path
import torch
PROJECT = Path(__file__).resolve().parents[2]
sys.path.insert(0, str(PROJECT / "code/src"))
from quantum_qwen38.qh032_pauli_observable_replacement import ( # noqa: E402
DequantizedPauliObservableCore,
EqualParameterClassicalObservableCore,
ExactPauliObservableCore,
QH032Config,
qh032_parameter_audit,
)
def relative(a: torch.Tensor, b: torch.Tensor) -> float:
return float((a - b).float().norm() / b.float().norm().clamp_min(1.0e-12))
def main() -> None:
if not torch.cuda.is_available():
raise RuntimeError("QH032 static validation requires CUDA")
torch.manual_seed(20260849)
device = torch.device("cuda:0")
config = QH032Config(hidden_size=32, output_size=16, rank=16, layer_indices=(0, 1), num_qubits=4, observable_offsets=(1, 2))
qh = ExactPauliObservableCore(config).to(device)
dq = DequantizedPauliObservableCore(config).to(device)
cc = EqualParameterClassicalObservableCore(config).to(device)
no_ent = ExactPauliObservableCore(config, use_entanglers=False).to(device)
if {module.parameter_count for module in (qh, dq, cc, no_ent)} != {32}:
raise AssertionError("small-core parameter counts are not equal")
identity_input = torch.randn(3, 5, config.rank, device=device, requires_grad=True)
identity_error = float((qh(identity_input) - identity_input).float().abs().max())
if identity_error > 1.0e-7:
raise AssertionError(f"zero-alpha QH032 is not identity: {identity_error}")
with torch.no_grad():
qh.theta.uniform_(-0.25, 0.25)
qh.alpha.uniform_(-0.08, 0.08)
dq.theta.copy_(qh.theta)
dq.alpha.copy_(qh.alpha)
xq = torch.randn(2, 3, config.rank, device=device, requires_grad=True)
xd = xq.detach().clone().requires_grad_(True)
yq, yd = qh(xq), dq(xd)
probe = torch.randn_like(yq)
(yq * probe).sum().backward()
(yd * probe).sum().backward()
checks = {
"forward_max_abs": float((yq - yd).float().abs().max()),
"forward_relative_l2": relative(yq, yd),
"input_gradient_max_abs": float((xq.grad - xd.grad).float().abs().max()),
"input_gradient_relative_l2": relative(xq.grad, xd.grad),
"theta_gradient_max_abs": float((qh.theta.grad - dq.theta.grad).abs().max()),
"theta_gradient_relative_l2": relative(qh.theta.grad, dq.theta.grad),
"alpha_gradient_max_abs": float((qh.alpha.grad - dq.alpha.grad).abs().max()),
"alpha_gradient_relative_l2": relative(qh.alpha.grad, dq.alpha.grad),
}
if checks["forward_max_abs"] > 2.0e-5 or checks["input_gradient_max_abs"] > 2.0e-5:
raise AssertionError(f"QH/DQ mismatch: {checks}")
if checks["theta_gradient_max_abs"] > 3.0e-4 or checks["alpha_gradient_max_abs"] > 3.0e-4:
raise AssertionError(f"QH/DQ gradient mismatch: {checks}")
gradient_audit = {}
for name, module in (("qh", qh), ("dq", dq), ("cc", cc), ("no_ent", no_ent)):
module.zero_grad(set_to_none=True)
with torch.no_grad():
module.theta.uniform_(-0.2, 0.2)
module.alpha.uniform_(-0.1, 0.1)
value = torch.randn(4, config.rank, device=device)
module(value).float().square().mean().backward()
gradient_audit[name] = {
"theta_norm": float(module.theta.grad.norm()),
"alpha_norm": float(module.alpha.grad.norm()),
"all_finite": bool(torch.isfinite(module.theta.grad).all() and torch.isfinite(module.alpha.grad).all()),
}
if gradient_audit[name]["theta_norm"] == 0 or gradient_audit[name]["alpha_norm"] == 0 or not gradient_audit[name]["all_finite"]:
raise AssertionError(f"invalid {name} gradients")
cpu_rejected = False
try:
ExactPauliObservableCore(config)(torch.randn(1, config.rank))
except RuntimeError:
cpu_rejected = True
if not cpu_rejected:
raise AssertionError("CPU exact-quantum path was not rejected")
full_audit = qh032_parameter_audit(QH032Config())
if full_audit["trainable_parameters"] != 82 or full_audit["net_parameter_reduction"] != 3145646:
raise AssertionError(f"full parameter audit changed: {full_audit}")
payload = {
"status": "pass",
"candidate": "QH-032 Pauli-observable nonlinear shared-basis replacement",
"small_config": config.__dict__,
"small_parameter_count_per_core": qh.parameter_count,
"zero_alpha_identity_max_abs": identity_error,
"dequantization_checks": checks,
"gradient_audit": gradient_audit,
"cpu_quantum_path_rejected": cpu_rejected,
"full_parameter_audit": full_audit,
"simulator_calls": int(qh.circuit_calls.item()),
"created_at": datetime.now(timezone.utc).isoformat(),
}
output = PROJECT / "artifacts/qh032-static-validation.json"
output.write_text(json.dumps(payload, indent=2), encoding="utf-8")
print(json.dumps(payload, indent=2))
if __name__ == "__main__":
main()
|