This is a well-structured VRAE implementation for future research uses. It has been tested on Pytorch 4.0 and Python 3.6. An example dataset located in examples/. Use this to try training and generating samples.

To train, run the following command from the root repository:

python -m train --data-path examples/sents.txt --save-dir examples/

To generate, run the following command:

python -m generate --vocab examples/vocab.pkl --ckpt-path examples/checkpoint-e01 --save-dir examples/

For more available options, check out python -m train --help and python -m generate --help.