torchaudio.save("t.wav", tensor.unsqueeze(0), sampling_rate, bits_per_sample=16)

m

add train.py