iconix / pytorch-text-vae

A conditional variational autoencoder (CVAE) for text

A partial reimplementation of "Generating Sentences From a Continuous Space", Bowman, Vilnis, Vinyals, Dai, Jozefowicz, Bengio (https://arxiv.org/abs/1511.06349).

Based on code from Kyle Kastner (@kastnerkyle) https://github.com/kastnerkyle/pytorch-text-vae, adapted to support the deephypebot project.

---

Changes in this detached fork:

• Update compatibility to Python 3 and PyTorch 0.4
• Add generate.py for sampling
• Add special support for JSON reading and thought vector conditioning
• Some code cleanup
• Add setup.py for package support as pytorchtextvae
• Train/test data split support