Pytorch implementation of Pipeline Model Parallelism as described in Google AI article https://arxiv.org/pdf/1811.06965.pdf

this is still a work in progress and not yet functional