selectorrrr/wav2vec2mos

import onnxruntime as ort
import librosa
import numpy as np

sess = ort.InferenceSession("wav2vec2mos.onnx")
signal, _ = librosa.load("example.wav", sr=16000)
x = signal[np.newaxis, :].astype(np.float32)
outs = sess.run(None, {"input_values": x})
mos = float(outs[0].reshape(-1).mean())

print("MOS:", mos)