Tomasz Grel (Nvidia): Faster Deep Learning with mixed precision and multiple GPUs
Industry Talk at the PL in ML: Polish View on Machine Learning 2018 Conference (plinml.mimuw.edu.pl).
Abstract:
The talk will focus on practical aspects of two common tricks for training deep learning models faster: mixed precision training and using multiple GPUs. Tomasz will demonstrate how to utilize half-precision operations to take advantage of the newest GPU architectures without compromising model accuracy. He will also briefly review common tools for using multiple GPUs to speed up the training. Examples of how to modify popular models to use both mixed precision and multigpu will be presented.
