from __future__ import annotations import argparse import json from pathlib import Path from typing import Any import numpy as np import yaml try: import matplotlib matplotlib.use("Agg") import matplotlib.pyplot as plt except ImportError as exc: # pragma: no cover raise SystemExit(f"matplotlib is required for visualization: {exc}") from exc PROJECT_ROOT = Path(__file__).resolve().parents[1] def load_config(path: str | Path) -> dict[str, Any]: with open(path, "r", encoding="utf-8") as handle: return yaml.safe_load(handle) def resolve_path(path: str | Path) -> Path: path = Path(path) return path if path.is_absolute() else PROJECT_ROOT / path def compute_numpy_metrics(pred: np.ndarray, target: np.ndarray) -> dict[str, float]: diff = pred - target mse = float(np.mean(diff**2)) mae = float(np.mean(np.abs(diff))) rel = np.linalg.norm(diff.reshape(diff.shape[0], -1), axis=1) / np.maximum( np.linalg.norm(target.reshape(target.shape[0], -1), axis=1), 1.0e-12 ) return {"mse": mse, "relative_l2": float(np.mean(rel)), "mae": mae} def plot_sample(pred: np.ndarray, target: np.ndarray, output_path: Path, sample_id: int) -> None: channels = pred.shape[0] fig, axes = plt.subplots(channels, 3, figsize=(10, 3 * channels), squeeze=False) for ch in range(channels): fields = [target[ch], pred[ch], np.abs(pred[ch] - target[ch])] titles = [f"target ch{ch}", f"prediction ch{ch}", f"abs error ch{ch}"] for col, (field, title) in enumerate(zip(fields, titles)): ax = axes[ch, col] image = ax.imshow(field, origin="lower", cmap="viridis") ax.set_title(title) ax.set_xticks([]) ax.set_yticks([]) fig.colorbar(image, ax=ax, fraction=0.046, pad=0.04) fig.suptitle(f"FlowBench FNO sample {sample_id}") fig.tight_layout() output_path.parent.mkdir(parents=True, exist_ok=True) fig.savefig(output_path, dpi=160) plt.close(fig) def main() -> None: parser = argparse.ArgumentParser(description="Summarize and visualize FlowBench FNO predictions.") parser.add_argument("--config", default=str(PROJECT_ROOT / "config/config.yaml")) parser.add_argument("--predictions", default=None) parser.add_argument("--output-dir", default=None) parser.add_argument("--max-samples", type=int, default=None) args = parser.parse_args() config = load_config(args.config) vis_cfg = config["visualization"] predictions_path = resolve_path(args.predictions or vis_cfg["predictions_path"]) output_dir = resolve_path(args.output_dir or vis_cfg["output_dir"]) max_samples = args.max_samples or int(vis_cfg.get("max_samples", 4)) archive = np.load(predictions_path) pred = archive["predictions"] target = archive["targets"] metrics = compute_numpy_metrics(pred, target) output_dir.mkdir(parents=True, exist_ok=True) for idx in range(min(max_samples, pred.shape[0])): plot_sample(pred[idx], target[idx], output_dir / f"sample_{idx:03d}.png", idx) summary_path = output_dir.parent / "result_summary.json" with open(summary_path, "w", encoding="utf-8") as handle: json.dump({"metrics": metrics, "figures_dir": str(output_dir)}, handle, indent=2) print(json.dumps({"metrics": metrics, "figures_dir": str(output_dir)}, indent=2), flush=True) if __name__ == "__main__": main()