Add Overthrust evaluation progress logs

Print method, device, dataset, batch progress, metric computation, and output paths during evaluation so long Colab runs show progress.

codes/eval_overthrust.py

@@ -104,8 +104,12 @@ def evaluate_overthrust(
     output_dir = Path(output_dir)
     output_dir.mkdir(parents=True, exist_ok=True)
     device = torch.device(device or ("cuda" if torch.cuda.is_available() else "cpu"))
+    print(f"[eval] method={method}, steps={num_inference_steps}, device={device}")
+    print(f"[eval] output_dir={output_dir}")
+    print("[eval] moving pipeline to device...")
     pipe = pipe.to(device)
 
+    print("[eval] building Overthrust dataset...")
     dataset = OverthrustTrueimpDataset(
         size=OVERTHRUST_CONFIG["size"],
         normalize=OVERTHRUST_CONFIG["normalize"],
@@ -125,6 +129,11 @@ def evaluate_overthrust(
         fixed_noise_snr=OVERTHRUST_CONFIG["noise_snr"],
         fixed_f0_phase=OVERTHRUST_CONFIG["f0_phase"],
     )
+    print(
+        "[eval] dataset ready: "
+        f"patches={len(dataset)}, batch_size={OVERTHRUST_CONFIG['batch_size']}, "
+        f"patch_indices={OVERTHRUST_CONFIG['patch_indices']}"
+    )
     loader = DataLoader(
         dataset,
         batch_size=OVERTHRUST_CONFIG["batch_size"],
@@ -135,8 +144,14 @@ def evaluate_overthrust(
     all_predictions: list[np.ndarray] = []
     all_targets: list[np.ndarray] = []
     all_reconstructions: list[np.ndarray] = []
-    for batch in loader:
+    total_batches = len(loader)
+    for batch_idx, batch in enumerate(loader, start=1):
         seeds = batch["seed"].tolist()
+        batch_size = len(seeds)
+        print(
+            f"[eval] batch {batch_idx}/{total_batches}: "
+            f"batch_size={batch_size}, seeds={seeds}"
+        )
         dipin = batch["dipin"].to(device)
         record = batch["record"].to(device)
         image = batch["image"].to(device)
@@ -151,6 +166,8 @@ def evaluate_overthrust(
                     device=device,
                 ),
             }
+            print(f"[eval] batch {batch_idx}/{total_batches}: CLDM operator ready")
+        print(f"[eval] batch {batch_idx}/{total_batches}: running pipeline...")
         output = pipe(
             dipin=dipin,
             record=record,
@@ -159,6 +176,7 @@ def evaluate_overthrust(
             seeds=seeds,
             **extra_kwargs,
         )
+        print(f"[eval] batch {batch_idx}/{total_batches}: collecting predictions...")
         prediction = output.impedance_samples
         reconstruction = output.impedance_reconstructed
         for local_idx in range(prediction.shape[0]):
@@ -166,6 +184,7 @@ def evaluate_overthrust(
             all_targets.append(image[local_idx, 0].detach().cpu().numpy())
             all_reconstructions.append(reconstruction[local_idx, 0].detach().cpu().numpy())
 
+    print("[eval] stitching patches...")
     full_target = stitch_patches(
         all_targets, dataset.splits, dataset.big_img.shape, OVERTHRUST_CONFIG["size"]
     )
@@ -176,10 +195,12 @@ def evaluate_overthrust(
         all_reconstructions, dataset.splits, dataset.big_img.shape, OVERTHRUST_CONFIG["size"]
     )
 
+    print("[eval] converting normalized predictions to impedance...")
     full_target_impedance = dataset.fan(full_target)
     full_prediction_impedance = dataset.fan(full_prediction)
     full_reconstruction_impedance = dataset.fan(full_reconstruction)
 
+    print("[eval] computing metrics...")
     metrics_summary = {
         "config": {
             **OVERTHRUST_CONFIG,
@@ -200,11 +221,13 @@ def evaluate_overthrust(
         "comparison": output_dir / "comparison_impedance.png",
         "metrics": output_dir / "metrics_summary.json",
     }
+    print("[eval] saving outputs...")
     np.save(paths["full_target"], full_target)
     np.save(paths["full_prediction"], full_prediction)
     np.save(paths["full_reconstruction"], full_reconstruction)
     save_comparison(full_target_impedance, full_prediction_impedance, paths["comparison"])
     paths["metrics"].write_text(json.dumps(metrics_summary, indent=2), encoding="utf-8")
+    print(f"[eval] done. metrics={paths['metrics']}")
     return {
         "metrics": metrics_summary,
         "paths": {key: str(value) for key, value in paths.items()},