3 ms·GPipe: Efficient Training of Giant Neural Networks Using Pipeline Parallelism2 points by pplonski86 8y agoibusworker 8y agoGoogle is splitting of the AI research community and creating a monopoly in deep learning.