NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:2711
Title:Hybrid 8-bit Floating Point (HFP8) Training and Inference for Deep Neural Networks

This paper demonstrates a method for low precision 8-bit deep neural network training without loss of performance across a number of popular tasks, including classification, speech recognition, and translation. The reviewers agree that this is a strong paper with solid analysis and a number of novel contributions to low precision training.