Mixed-Precision Training of Deep Neural Networks

Originally published at: https://developer.nvidia.com/blog/mixed-precision-training-deep-neural-networks/

Deep Neural Networks (DNNs) have lead to breakthroughs in a number of areas, including image processing and understanding, language modeling, language translation, speech processing, game playing, and many others. DNN complexity has been increasing to achieve these results, which in turn has increased the computational resources required to train these networks. Mixed-precision training lowers the required…

May be I missed it - but what is the real speedup achieved for these models?

What is the accuracy drop if FP8 or FP10 are used ?

Have not tried the mixed precision in training but we did so with simulations so can share few remarks about the overall speedup here https://medium.com/@marcroj...

Hi, where I can find the pre-trained models for Faster R-CNN and Multibox SSD (FP16 and FP32)?

What's the `VGG-D` model in Table.1?