Deep Learning at Scale with Horovod

September 21, Thuesday, 14:00 - 16:00 (Samara, GMT +4)

Speaker: Dmitry Mironov


Duration: 2 hours

Prerequisites: Competency in the Python programming language and experience training deep learning models in Python

Workshop Brief: Modern deep learning challenges leverage increasingly larger datasets and more complex models. As a result, significant computational power is required to train models effectively and efficiently. In the workshop, you will learn how to scale deep learning training to multiple GPUs with Horovod, the open-source distributed training framework originally built by Uber. Upon completion, you'll be able to use Horovod to effectively scale deep learning training in new or existing code bases.

Speaker Information: Dmitry Mironov is an expert in machine learning and inference optimization. Since 2019, Dmitry has been working at Solutions Architect (NVIDIA) and helps various teams to squeeze maximum performance and accuracy from their computing resources with minimal effort. Prior to that, he was CTO of a computer vision startup, leading several successful projects. Prior to his career in the industry, Dmitry graduated from Moscow Institute of Physics and Technology and was a PhD student at Skoltech, in the robotics laboratory.

Efficient pattern recognition on domestic VLIW-platforms

September 22, Wednesday, 14:00 - 15:30 (Samara, GMT +4)

Speakers: Elena Limonova, Anton Trusov

Limonova   Trusov

Duration: 1,5 hours

Prerequisites: Fundamentals of C/C++

Workshop Brief: In the workshop, we will talk about using the Elbrus platform to solve recognition problems. Elbrus processors are Russian domestic processors with VLIW architecture, which provides ample opportunities for implementing internal program parallelism. We will describe the key features of this platform, introduce the listeners to the available computational optimization and effective programming capabilities using examples from the field of image processing and neural network recognition. Finally, a comparison of the execution of recognition applications on Elbrus processors of different generations will be presented.

Speaker Information: Elena Limonova – PhD student at Federal Research Center Computer Science and Control, RAS. Major fields of scientific research: pattern recognition, mobile devices, hardware-oriented neural networks. Anton Trusov – Master’s student at Moscow Institute of Physics and Technology (Department of Radio-Engineering). Major fields of scientific research: pattern recognition, artificial intelligence.

Back to Top