Perform in-place resampling during read_audio.

utils_vad.py

@@ -122,16 +122,12 @@ class Validator():
 def read_audio(path: str,
                sampling_rate: int = 16000):
 
-    wav, sr = torchaudio.load(path)
+    effects = [
+        ['channels', '1'],
+        ['rate', str(sampling_rate)]
+    ]
 
-    if wav.size(0) > 1:
-        wav = wav.mean(dim=0, keepdim=True)
-
-    if sr != sampling_rate:
-        transform = torchaudio.transforms.Resample(orig_freq=sr,
-                                                   new_freq=sampling_rate)
-        wav = transform(wav)
-        sr = sampling_rate
+    wav, sr = torchaudio.sox_effects.apply_effects_file(path, effects=effects)
 
     assert sr == sampling_rate
     return wav.squeeze(0)