A pytorch implementation of the paper "Latent Variable Dialogue Models and their Diversity" https://arxiv.org/abs/1702.05962 It introduces a Variational Sequence to Sequence Model to generate diverse dialogues.
Extract train.txt
and valid.txt
from opensubtitle(https://s3.amazonaws.com/opennmt-trainingdata/opensub_qa_en.tgz) to data/
$ python prepare_data_opensub.py
$ python main.py