Here I try to get audio and label from raw_dataset(tfrec)
feature_description = { \'recording_id\': tf.io.FixedLenFeature([], tf.string, default_value=\'\