Upload scripts/example_dataloader.py

scripts/example_dataloader.py

@@ -0,0 +1,109 @@
+#!/usr/bin/env python3
+"""
+Minimal HDF5WaveformDataset usage example.
+
+Run:
+    python example_dataloader.py --h5_input path/to/data.h5
+"""
+
+import argparse
+import numpy as np
+import sys
+from pathlib import Path
+from torch.utils.data import DataLoader
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+from utils.hdf5_waveform_dataset import HDF5WaveformDataset, waveform_collate_fn
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--h5_input", default="data/hdf5/continuous_waveform_usa_20190701.h5",
+                        help="HDF5 file, directory, or glob pattern")
+    parser.add_argument("--n_samples", type=int, default=3,
+                        help="Number of samples to print")
+    parser.add_argument("--response_json", default="data/response/instrument_responses.json",
+                        help="Instrument-response JSON used with --remove_response")
+    parser.add_argument("--remove_response", action="store_true",
+                        help="Remove the native instrument response before resampling")
+    parser.add_argument("--response_output", default="VEL",
+                        help="Physical output unit for response removal: DISP, VEL, or ACC")
+    parser.add_argument("--response_pre_filt", nargs=4, type=float, default=None,
+                        metavar=("F1", "F2", "F3", "F4"),
+                        help="Four-corner pre-filter passed to ObsPy remove_response")
+    parser.add_argument("--response_water_level", type=float, default=60.0,
+                        help="ObsPy water level; use a negative value to pass None")
+    args = parser.parse_args()
+    water_level = None if args.response_water_level < 0 else args.response_water_level
+
+    # ── 1. Build dataset ───────────────────────────────────────────────────
+    dataset = HDF5WaveformDataset(
+        h5_file=args.h5_input,
+        mode="three",                        # returns [T, 3] waveform per station-day
+        allowed_families=("HH", "BH", "EH", "HN"),
+        allowed_z_only_channels=("EHZ",),
+        allow_z_only=True,
+        replicate_z_only=True,               # Z-only → [Z, Z, Z]
+        target_sampling_rate=100.0,          # resample everything to 100 Hz
+        instrument_response_json=args.response_json if args.remove_response else None,
+        remove_instrument_response=args.remove_response,
+        response_output=args.response_output,
+        response_pre_filt=tuple(args.response_pre_filt) if args.response_pre_filt else None,
+        response_water_level=water_level,
+    )
+
+    print(f"HDF5 files  : {len(dataset.h5_files)}")
+    print(f"Total samples: {len(dataset)}")
+    print()
+
+    # ── 2. Build DataLoader ────────────────────────────────────────────────
+    loader = DataLoader(
+        dataset,
+        batch_size=1,
+        shuffle=False,
+        num_workers=0,           # 0 = single-process, safest for h5py
+        collate_fn=waveform_collate_fn,
+    )
+
+    # ── 3. Iterate and print ───────────────────────────────────────────────
+    for i, batch in enumerate(loader):
+        if i >= args.n_samples:
+            break
+
+        item = batch[0]          # batch_size=1, so one item per batch
+
+        w = item["waveform"]     # torch.Tensor [T, 3]
+        sr = item["sampling_rate"]
+        duration_sec = w.shape[0] / sr if sr and sr > 0 else float("nan")
+
+        print(f"── Sample {i + 1} ──────────────────────────────────────────")
+        print(f"  station_id    : {item['station_id']}")
+        print(f"  network       : {item['station_info'].get('network', '')}."
+              f"{item['station_info'].get('station', '')}")
+        print(f"  channels      : {item['channels']}")
+        print(f"  starttime     : {item['starttime']}")
+        print(f"  sampling_rate : {sr} Hz")
+        print(f"  waveform shape: {tuple(w.shape)}  "
+              f"({duration_sec:.1f} s  ×  3 components)")
+        print(f"  waveform dtype: {w.dtype}")
+        print(f"  Z-only        : {item.get('is_z_only', False)}")
+        if args.remove_response:
+            print(f"  response      : {item.get('instrument_processing', {})}")
+        print(f"  location      : "
+              f"lon={item['station_info'].get('longitude', float('nan')):.4f}  "
+              f"lat={item['station_info'].get('latitude',  float('nan')):.4f}")
+        # Quick per-channel stats
+        for c, name in enumerate(["E/1", "N/2", "Z/3"]):
+            ch = w[:, c].numpy()
+            print(f"  ch[{name}]  "
+                  f"min={float(np.min(ch)):+.3e}  "
+                  f"max={float(np.max(ch)):+.3e}  "
+                  f"std={float(np.std(ch)):.3e}")
+        print()
+
+    dataset.close()
+
+
+if __name__ == "__main__":
+    main()