The Habana Labs team looks forward to meeting up with you—researchers, academics, recruitment prospects, and those who are bringing AI to life in the datacenter and cloud—at this year’s Neural Information Processing Systems Conference, aka NeurIPS, which takes place from December 8 to 14 in Vancouver. There we’ll demonstrate our AI Training and Inference technologies, provide “how to” hands-on guidance in our Inference implementation workshop, and present fresh re-thinking on training CNNs.
Here’s how you can participate with the Habana team to benefit from our latest developments in the practical, real-world implementation of Inference and Training AI systems and applications.
ATTEND AND LEARN
EXPO: HABANA WORKSHOP – Sunday, December 8, 2 – 6pm
What: Hands-on Workshop: Implementing High-Performance AI Workloads with Habana AI Processors
Where: Rooms West 208 and 209, Vancouver Convention Centre
In this workshop, Leon Goldgeisser will provide details on how to implement an inference system with the Habana Goya Inference Processor. He’ll share foundations of the processor’s hardware and software and demonstrate how to use Goya to solve the most common and computationally extensive inference tasks.
An additional real world implementation feature of the workshop will be a talk by Whitney Zhao, Hardware Engineer of Facebook, who will discuss industry advances in AI hardware standardization to address the explosion of new types of hardware accelerators for Machine Learning, Deep Learning, and High-Performance Computing.
See the workshop abstract here. We hope you’ll join us at our EXPO workshop that bridges academia and industry application.
COME BY BOOTH 216 AND SEE GAUDI TRAINING AND GOYA INFERENCE IN ACTION
DEMONSTRATION & DISPLAY: Habana Booth – Sunday, December 8 – Wednesday, December 11, 9 – 5pm
What: Habana Training and Inference Solutions Demonstrated Live
Where: Booth 216, Convention Centre East
In our booth you will be able to see Goya Inference and Gaudi Training hardware and and real-time performance demonstrations of both sets of working AI solutions.
Come by booth 216 to meet the Habana team and see the live demos or contact our team at [email protected] for a private meeting.
ATTEND AND HEAR…
POSTER PRESENTATION: Saturday, December 14, noon to 2pm at the “Science Meets Engineering of Deep Learning” Workshop
What: Poster presentation – MIX & MATCH: TRAINING CONVNETS WITH MIXED IMAGE SIZES FOR IMPROVED ACCURACY, SPEED AND SCALE RESILIENCY
Where: Rooms 121 & 122, Convention Centre West
The authors will share yet-to-be explored trade-offs in batch and spatial input size used to train modern convolutional networks. We will also introduce novel methods to exploit these trade-offs, improving both training and inference time, as well as accuracy of convolutional models.
Here is the paper authored by Elad Hoffer, Berry Weinstein and Itay Hubara of Habana Labs Research; Tal Ben-Nun and Torsten Hoefler, Department of Computer Science, ETH Zurich University; and Daniel Soudry, Department of Electrical Engineering Technion, Haifa, Israel.
For more information on the poster presentations, please see details here.
Thanks for your interest in Habana Labs at NeurIPS, and our inference and training product lines, Goya and Gaudi. If you have any questions or wish to arrange for a private meeting during the NeurIPS conference, please contact: Eitan Medina, Chief Business Officer: [email protected]