TensorOpt: Exploring the Tradeoffs in Distributed DNN Training with Auto-Parallelism

Open in new window