CogView: Mastering Text-to-Image Generation via Transformers - 42Papers