File size: 3,178 Bytes
0ef4ba2 | 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 | """Physical and runtime audit for the refined conditional DENIM v2."""
from __future__ import annotations
import json
import time
from collections import defaultdict
import torch
from src.t2_denim_conditional import ConditionalDENIM, rollout
from src.train_t2_denim_conditional_v2 import ARTIFACT_DIR, MODEL_DIR, load_all_trajectories
def run() -> dict[str, object]:
checkpoint = torch.load(
MODEL_DIR / "denim_conditional_v2_refined.pt",
map_location="cpu",
weights_only=False,
)
model = ConditionalDENIM()
model.load_state_dict(checkpoint["state_dict"])
model.eval()
trajectories = load_all_trajectories()
groups: dict[str, list] = defaultdict(list)
for trajectory in trajectories:
groups[trajectory.evaluation].append(trajectory)
selected = (
groups["hidden_frozen_test"][:8]
+ groups["test_path_ood"][:8]
+ groups["test_long_horizon"][:8]
)
strain = torch.stack([item.strain for item in selected])
started = time.perf_counter()
with torch.no_grad():
result = rollout(
strain,
torch.tensor([item.young for item in selected]),
torch.tensor([item.poisson for item in selected]),
torch.tensor([item.yield_stress for item in selected]),
torch.stack([item.descriptor for item in selected]),
model,
bisection_iterations=24,
)
seconds = time.perf_counter() - started
reference = torch.stack([item.stress for item in selected])
error = result["stress"] - reference
active = result["plastic_increment"] > 1.0e-11
report = {
"model": "material-conditioned DENIM v2 refined",
"parameter_count": sum(parameter.numel() for parameter in model.parameters()),
"audit_trajectory_count": len(selected),
"audit_state_point_count": int(strain.shape[0] * strain.shape[1]),
"stress_rmse_mpa": float(torch.sqrt(error.square().mean()) / 1.0e6),
"maximum_yield_residual_pa": float(result["yield_residual"][active].abs().max()),
"minimum_peeq_increment": float(torch.diff(result["peeq"], dim=1).min()),
"maximum_plastic_strain_trace": float(
result["plastic_strain"][..., :3].sum(dim=-1).abs().max()
),
"maximum_memory_trace_pa": float(
result["memories"][..., :3].sum(dim=-1).abs().max()
),
"all_finite": all(torch.isfinite(value).all() for value in result.values()),
"inference_seconds": seconds,
"state_points_per_second": float(strain.shape[0] * strain.shape[1] / seconds),
"descriptor_hidden_hardening_entries_zero": bool(
torch.count_nonzero(torch.stack([x.descriptor[3:10] for x in selected])) == 0
),
"scope": "material-point audit; conditional v2 is not yet an AgentFEM global provider bundle",
}
ARTIFACT_DIR.mkdir(parents=True, exist_ok=True)
(ARTIFACT_DIR / "physical_runtime_validation.json").write_text(
json.dumps(report, indent=2) + "\n", encoding="utf-8"
)
print(json.dumps(report, indent=2))
return report
if __name__ == "__main__":
run()
|