A Tensorflow based RNN with GRU cells trained on Shakespeare's complete works. This RNN is 3 layers deep with an internal size of 512. It is trained on sequences of 30 characters pulled from Shakespeare's complete works.
Take a look at the text generated from the trained RNN. Here's how the training and test error and cross entropy vary throughout the training iterations.
