Debug for mono input & support all dtypes

Signed-off-by: Jael Gu <mengjia.gu@zilliz.com>

torch_vggish.py

@@ -64,12 +64,14 @@ class Vggish(NNOperator):
         return [AudioOutput(outs.detach().numpy())]
 
     def preprocess(self, frames: List[numpy.ndarray], sr, layout):
-        audio = numpy.hstack(frames)
         if layout == 'stereo':
-            audio = audio.reshape(-1, 2)
+            frames = [frame.reshape(-1, 2) for frame in frames]
+            audio = numpy.vstack(frames)
+        else:
+            audio = numpy.hstack(frames)
+            audio = audio.transpose()
         audio = self.int2float(audio)
         try:
-            audio = audio.transpose()
             audio_tensors = vggish_input.waveform_to_examples(audio, sr, return_tensor=True)
             return audio_tensors
         except Exception as e:

vggish_input.py

@@ -28,8 +28,8 @@ def waveform_to_examples(data, sample_rate, return_tensor=True):
     """Converts audio waveform into an array of examples for VGGish.
 
   Args:
-    data:
-        np.array of 2 dimension, second of which is number of channels.
+    data: np.array of either one dimension (mono) or two dimensions
+      (multi-channel, with the outer dimension representing channels).
       Each sample is generally expected to lie in the range [-1.0, +1.0],
       although this is not required.
     sample_rate: Sample rate of data.
@@ -43,10 +43,8 @@ def waveform_to_examples(data, sample_rate, return_tensor=True):
 
   """
     # Convert to mono.
-    if data.shape[1] > 1:
+    if len(data.shape) > 1:
         data = np.mean(data, axis=1)
-    else:
-        data = data.squeeze(1)
     # Resample to the rate assumed by VGGish.
     if sample_rate != vggish_params.SAMPLE_RATE:
         data = resampy.resample(data, sample_rate, vggish_params.SAMPLE_RATE)