Review for NeurIPS paper: HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis