Credit

This project was forked from https://github.com/martin-gorner/tensorflow-rnn-shakespeare

assume all code is attributed to and owned by Martin Gorner.

Original Code And Presentation

Code for the Recurrent Neural Network in the presentation "Tensorflow and deep learning - without a PhD, Part 2"

The presentation itself is available here:

• Video
• Slides

This sample has now been updated for Tensorflow 1.1. Please make sure you redownload the checkpoint files if you use rnn_play.py.

Usage:

To Train

Step 1 - Train on Twitter Data

$ python3 rnn_train.py This will train the model and save it with checkpoint files.

After training, you can see what the model outputs by using

$ python3 rnn_play.py

You will need to replace the files with your checkpoint files.

Step 2 - Train on pickup line data

$ python3 rnn_retrain.py

You will need to replace the checkpoint file names for this to work.

To Play

Point to the correct checkpoint files and do $ python3 rnn_play.py

Each training session will generate saved checkpoints as well as save a final model when it has finished. you can monitor the progress using Tensorboard

$ tensorboard --logdir="your/log/path"