A Comparison of Deep Learning MOS Predictors for Speech Synthesis Quality