Merge branch 'kevin-cgc-main' into dev (#3)

src/aac_datasets/datasets/functional/audiocaps.py

@@ -642,6 +642,15 @@ def _download_from_youtube(
         return False
     _video_link, audio_link = lines[:2]
 
+    # if yt-dlp only returns one link, it is a combined video audio
+    if len(audio_link) == 0:
+        if verbose >= 2:
+            pylog.debug(
+                f"youtube_id={youtube_id} is combined video audio only (cant download)"
+            )
+        # audio_link = video_link # this does not work, not sure why. probably requires changes to ffmpeg command
+        return False
+
     # Download and extract audio from audio_link to fpath_out with ffmpeg
     extract_command = [
         ffmpeg_path,

src/aac_datasets/datasets/functional/wavcaps.py

@@ -362,7 +362,7 @@ def download_wavcaps_dataset(
     ign_patterns = [
         pattern
         for source in ign_sources
-        for pattern in (f"json_files/{source}/*.json", "Zip_files/*")  # {source}/
+        for pattern in (f"json_files/{source}/*.json", f"Zip_files/{source}/*")
     ]
     if verbose >= 2:
         pylog.debug(f"ign_sources={ign_sources}")
@@ -395,7 +395,7 @@ def download_wavcaps_dataset(
     del snapshot_dpath
 
     # Build symlink to hf cache
-    if osp.exists(wavcaps_root):
+    if osp.lexists(wavcaps_root):
         if not osp.islink(wavcaps_root):
             raise RuntimeError("WavCaps root exists but it is not a symlink.")
         link_target_abspath = osp.abspath(osp.realpath(wavcaps_root))

src/aac_datasets/utils/download.py

@@ -54,7 +54,10 @@ def safe_rmdir(
     ):
         if not rm_root and dpath == root:
             continue
-        elif len(fnames) == 0 and len(dnames) == 0:
+        elif len(fnames) == 0 and (
+            len(dnames) == 0
+            or all(osp.join(dpath, dname) in to_delete for dname in dnames)
+        ):
             to_delete.append(dpath)
         elif error_on_non_empty_dir:
             raise RuntimeError(f"Cannot remove non-empty directory '{dpath}'.")