ICML 2019 Expo Talk

June 9, 2019

Expo Schedule »

Optimize/profile deep learning on Apache Spark with the latest Intel DL Boost Technologies and Intel Parallel Studio

Sponsor: Intel AI

Organizers:
Louie Tsai (Speaker)

Abstract:

With Intel DL Boost (VNNI )support in BigDL, the DL inference speed can be improved by 1.8X to 3.2X. We will show case the VGG-16 Fp32/Int8 throughput improvement by using VNNI on BigDL. We will also show case how to use Intel Parallel Studio to profile DL and then optimize the DL workloads accordingly.