Model-Aware Parallelization Strategy for Deep Neural Networks' Distributed Training | IEEE Conference Publication | IEEE Xplore