Scaling AI at Cerebras

Urs Köster (Cerebras Systems)

11:05am–11:45am Wednesday, September 11, 2019

Location: 230 C

Implementing AI

Secondary topics: Deep Learning, Deep Learning tools, Hardware

Average rating:

(5.00, 1 rating)

Download slides (PDF)

Long training times are the single biggest factor slowing down innovation in deep learning. Today’s common approach of scaling large workloads out over many small processors is inefficient and requires extensive model tuning. With increasing model and dataset sizes, new ideas are needed to reduce training times.

Urs Köster explores trends in the computer vision and natural language processing domains and techniques for scaling with the Cerebras wafer scale engine—the largest chip in the world. Cerebras’s unique, purpose-built processor allows you to leverage sparsity for building larger models and enables model-parallel training as an efficient alternative to data-parallel training.

What you'll learn

Discover new ideas for reducing training times

Urs Köster

Cerebras Systems

Urs Köster is the head of machine learning at Cerebras Systems, where he develops novel deep learning algorithms to enable the next generation of AI. He has 15 years of experience in neural networks and computational neuroscience, contributed to machine learning frameworks, developed low-precision numerical formats, and led data science engagements. Previously, he was head of algorithms R&D at Intel Nervana and a researcher at UC Berkeley.