Transformer-Based Conditioned Variational Autoencoder (T-CVAE) for Paraphrase Generation. T-CVAE was originally proposed for story completion task in IJCAI'19 paper: T-CVAE: Transformer-Based Conditioned Variational Autoencoder for Story Completion by Wang, Tianming and Wan, Xiaojun. The implementation is based on @sodawater's T-CVAE.
zhiming-xu/transformer-vae-paraphrase
Transformer-based conditional variational autoencoder for paraphrase generation.
Python