fix: source RUASD from official dataset

Author: dylanuys

src/gasbench/dataset/config.py

@@ -59,6 +59,10 @@ class BenchmarkDatasetConfig:
 
     include_paths: Optional[List[str]] = None
     exclude_paths: Optional[List[str]] = None
+    # Filter media entries inside ZIP/TAR archives. Unlike include_paths and
+    # exclude_paths, these do not affect which remote archive files are listed.
+    archive_include_paths: Optional[List[str]] = None
+    archive_exclude_paths: Optional[List[str]] = None
 
     # For parquet datasets: specify column name(s) containing media bytes
     # Works for any modality (image, audio, video)
@@ -380,6 +384,8 @@ def _dataset_dict_to_config(d: dict, **overrides) -> BenchmarkDatasetConfig:
         "archives_per_dataset": d.get("archives_per_dataset", 5),
         "include_paths": d.get("include_paths"),
         "exclude_paths": d.get("exclude_paths"),
+        "archive_include_paths": d.get("archive_include_paths"),
+        "archive_exclude_paths": d.get("archive_exclude_paths"),
         "data_columns": d.get("data_columns"),
         "notes": d.get("notes"),
         "filter_column": d.get("filter_column"),

src/gasbench/dataset/configs/synthetic_audio.yaml

@@ -785,12 +785,14 @@ datasets:
     notes: Malayalam train split, released from v22 bmcore holdouts
 
   - name: ruasd-audio
-    path: 34data/ruasd-audio
+    path: lab260/RuASD
     modality: audio
     media_type: synthetic
-    source_format: zip
-    media_per_archive: -1
-    archives_per_dataset: -1
+    source_format: tar
+    media_per_archive: 500
+    archives_per_dataset: 2
+    hf_revision: fcbc87c57b54ef4f58e1135e2813f6f000c2b739
+    archive_include_paths: [fake]
     generator_family: mixed-generation
     content_category: speech
-    notes: Released from v22 bmcore holdouts
+    notes: Synthetic subset of the official RuASD anti-spoofing dataset

src/gasbench/dataset/download/extract.py

@@ -317,6 +317,20 @@ def open_entry(m):
                 if any(get_name(e).lower().endswith(ext) for ext in valid_exts)
                 and "MACOSX" not in get_name(e)
             ]
+            archive_include_paths = getattr(dataset, "archive_include_paths", None)
+            archive_exclude_paths = getattr(dataset, "archive_exclude_paths", None)
+            if archive_include_paths:
+                candidates = [
+                    entry
+                    for entry in candidates
+                    if any(segment in get_name(entry) for segment in archive_include_paths)
+                ]
+            if archive_exclude_paths:
+                candidates = [
+                    entry
+                    for entry in candidates
+                    if not any(segment in get_name(entry) for segment in archive_exclude_paths)
+                ]
             if not candidates:
                 logger.warning(f"No matching files found in {source_path}")
                 return
@@ -754,4 +768,3 @@ def extract_number(filepath):
         logger.warning(f"Error processing frame directory {frame_dir}: {e}")
         return
 
-

tests/unit/test_archive_member_filtering.py

@@ -0,0 +1,40 @@
+import io
+import tarfile
+
+from gasbench.dataset.config import BenchmarkDatasetConfig
+from gasbench.dataset.download.extract import yield_media_from_source
+
+
+def _write_tar(path, members):
+    with tarfile.open(path, "w") as archive:
+        for name, payload in members.items():
+            info = tarfile.TarInfo(name)
+            info.size = len(payload)
+            archive.addfile(info, io.BytesIO(payload))
+
+
+def test_archive_member_filtering_keeps_only_matching_media(tmp_path):
+    archive_path = tmp_path / "mixed.tar"
+    _write_tar(
+        archive_path,
+        {
+            "raw/fake/tts/audio/fake.wav": b"synthetic audio",
+            "raw/real/speech/audio/real.wav": b"captured audio",
+            "raw/fake/tts/metadata/fake.json": b"{}",
+        },
+    )
+    dataset = BenchmarkDatasetConfig(
+        name="mixed-audio",
+        path="example/mixed-audio",
+        modality="audio",
+        media_type="synthetic",
+        source_format="tar",
+        archive_include_paths=["/fake/"],
+    )
+
+    samples = list(yield_media_from_source(archive_path, dataset, num_items=-1))
+
+    assert [sample["member_path"] for sample in samples] == [
+        "raw/fake/tts/audio/fake.wav"
+    ]
+    assert samples[0]["audio_bytes"] == b"synthetic audio"