Intel Optimized Libraries for AI: Our Fast Track to oneAPI
Aible provides developers, data scientists, ML engineers, and their business teams an end-to-end data analytics and autoML platform with the goal of going from raw data to valuable insights in hours and to measurable business impact in one month or less – guaranteed. As such our development investment needs to cover a wide range of areas including User Experience, data transform and feature engineering, business-driven model optimization, user collaboration, and model deployment and monitoring together with portability across cloud platforms. An overarching requirement is to provide great performance if we are to make good on our commitment to “valuable insights in hours” and “business impact in one month” and leveraging all that we can from Intel Architecture is key. In this lightning talk we will cover how we leveraged the Intel optimizations in Tensorflow, Numpy, Scipy to get the performance benefits of OneAPI on Intel Architecture, the (minimal) code impact and the resulting performance gains. We will also cover ongoing and current results as we continue to incorporate the latest improvements available in these OneAPI optimized libraries.