This talk will review results several large scale data analytics projects at NERSC. We will review a range of technologies (Julia, Caffe, TensorFlow, Python) and methods (clustering, 3-pt correlation, variational inference, deep learning) that have been scaled out to the entire Cori system. All these projects have been conducted in the context of real science drivers from cosmology, astronomy, climate science and high-energy physics, and have obtained O(1)-O(10) PF performance. Finally, we will comment on our first ExaOp application involving scaling Deep Learning on the OLCF Summit platform.
Back to Workshop III: HPC for Computationally and Data-Intensive Problems