why the inference need to take so much time, one image need more 1 minute?

[wmmin]

[USTC-JialunPeng]

Indeed, the inference time of our autoregressive model is long compared to GAN-based or VAE-based models. The majority of the time is spend iteratively running the structure generator for 32x32 times. Please refer to the Supplementary Material for details on Inference Time.