Project done by Jan Latko, Artur Przybysz and Jonatan Cichawa for a Deep Learning DTU course under the supervision from Corti.
To run an RNN-VAE experiment use experiment.py file with appropriate configuration (experiments use Sacred).
To evaluate a saved model against different dataset use load_model.py.
To run a word frequency baseline download those english word frequencies and run baseline.py.
lang_model.py runs a RNN-LM experiment.
analyze_results.ipynb was used to load logs and metrics from experiments and analyze them.
Needed data and embeddings should download automatically.
Parts of the code were based on https://github.com/wiseodd/controlled-text-generation.