[danielmarkbruce]

It cuts to training data and effort. A lot of effort has been put in to optimize for python, even down to tokenization.