def load_audio()

in utils/helpers.py [0:0]

• 8 lines of code
• 3 McCabe index (conditional complexity)

def load_audio(wave_file: str):
    """
    :param wave_file: .wav file containing the audio input
    :return: 1 x T tensor containing input audio resampled to 16kHz
    """
    audio, sr = ta.load(wave_file)
    if not sr == 16000:
        audio = ta.transforms.Resample(sr, 16000)(audio)
    if audio.shape[0] > 1:
        audio = th.mean(audio, dim=0, keepdim=True)
    # normalize such that energy matches average energy of audio used in training
    audio = 0.01 * audio / th.mean(th.abs(audio))
    return audio