Building a Recurrent Neural network implementing LSTM cells in order to generate a new Simpsons TV script. The Simpsons dataset of scripts from 27 seasons was used. The Neural Network generates a new TV script for a scene at Moe's Tavern. The RNN was trained on GPU using FloydHub.
In order to run it, setup Conda, then activate an environment:
$ conda env create -f environment.yaml
Setup floyd hub and launch it in GPU mode:
$ floyd login
$ floyd init
$ floyd run --mode jupyter --gpu
You can open the jupyter notebook directly in github TV script generation with RNN