Chimera: Efficiently Training Large-Scale Neural Networks with Bidirectional Pipelines

Open in new window