Webinar: Automatic Mixed Precision in Tensorflow

Automatic Mixed Precision is an approach to simplify ease of use and provide up to 3x performance improvements by utilizing NVIDIA Tensor Cores. We’ll review the theory behind mixed-precision, the functionality of the Automatic Mixed Precision feature, how to use it in model training scripts, and some customer examples of realized performance. Compared to single precision, mixed precision offers many benefits: 2x better use of the available DRAM bandwidth, smaller memory footprints which allow larger batch sizes or network architectures with more parameters to fit in GPU memory, and allow usage of Volta’s Tensor Cores to boost raw math throughput by up to 8x. We describe ways to enable Automatic Mixed Precision in TensorFlow. Learn more.