“The convergence of supercomputing and big data analytics is happening now, and the rise of deep learning algorithms is evidence of how customers are increasingly using high performance computing techniques to accelerate analytics applications,” said Cray CTO Steve Scott. “Training problems look very much like classical supercomputing problems. We believe that with our programming environment, toolkits, and processing technologies, we have the right combination of hardware and software to help our customers execute deep learning workloads.”
They are validated on Cray XC and CS-Storm systems, and include Microsoft Cognitive Toolkit (previously CNTK), TensorFlow, Nvidia Digits deep learning GPU training system), Caffe, Torch, and MXNet.
CS-Storm system – which offers 850 GPU Tflop/s in a single rack – now supports the PCIe version of Nvidia’s Tesla P100 data centre accelerator (scroll down the IBM article) and the Nvidia Tesla M40 deep learning training accelerator. Cray XC50 now also has Tesla P100.
Marine geophysical company PGS is running machine learning algorithms on its XC40 for seismic exploration deep under the Gulf of Mexico.
“This class of problems is notoriously hard. It is a multidimensional ill-posed optimisation problem that is far from automated and requires lots of skilled intervention – sometimes more art than science in many cases,” said PGS chief geophysicist Dr Sverre Brandsberg-Dahl. “Our XC40 was able to learn how to best steer refracted and diving waves for deep model updates and how best to reproduce the sharp salt boundaries in the Gulf of Mexico. Machine learning at scale showed dramatic improvement in the quality of the inversion process as compared to current state-of-the-art ‘full waveform inversion’.”