MIC 2018 - Tensorflow optimizations and performance tuning for Intel platforms

Bio:
Ellick Chan is the Head of University Relations and Research at Intel AI. He earned a PhD in Computer Science from the University of Illinois at Urbana

Title:
Tensorflow optimizations and performance tuning for Intel platforms

Abstract:
Tensorflow is a popular software framework used for deep learning. Intel’s introduction of AVX-512 vectorized hardware greatly accelerates neural network inference and training at data center scale. This talk describes the software and hardware optimizations Intel has made to Tensorflow to leverage AVX-512 and how end users can leverage tools such as Tensorflow Timeline and VTune to optimize their AI workloads.