in basic_pitch/data/datasets/maestro.py [0:0]
def create_input_data(source: str) -> List[Tuple[str, str]]:
import apache_beam as beam
filesystem = beam.io.filesystems.FileSystems()
with tempfile.TemporaryDirectory() as tmpdir:
maestro = mirdata.initialize("maestro", data_home=tmpdir)
metadata_path = maestro._index["metadata"]["maestro-v2.0.0"][0]
with filesystem.open(
os.path.join(source, metadata_path),
) as s, open(os.path.join(tmpdir, metadata_path), "wb") as d:
d.write(s.read())
return [(track_id, track.split) for track_id, track in maestro.load_tracks().items()]