This repository revisits text generation as detailed in The Unreasonable Effectiveness of Recurrent Neural Networks and tries to bring more understanding about RNN's and how to optimize them.

This repository contains support material to solve the text generation problem as detailed in The Unreasonable Effectiveness of Recurrent Neural Networks. The objective is to read a large text file, one character at a time, and then be able to generate text (one character at a time) with the same style.

Every single experiment is automatically logged onto Weighs & Biases for easier analysis/interpretation of results. We also want to bring more understanding about RNN's and how to optimize them.

Dependencies can be installed through requirements.txt.

The following files are present:

• notebook.ipynb is a Jupyter Notebook used to prototype our solution ;
• train.py is a script to run several experiments and log them on Weighs & Biases.

See my results and conclusions:

• Results & Conclusions
• W&B report
• W&B runs

And here is a little sample of what we can generate.