Huawei Solutions for AI Computing (MindSpore framework + Hardware Atlas 800)

Вторник, 28 сентября 2021 г., 11:30 - 13:30
Зал "Международная"

1. Prof. Chen Lei – Huawei Chief Scientist of MindSpore: “MindSpore – an end to end deep learning computing framework”, (online) – 30 min
2. Dmitry Yudin – PhD, Head of Intelligent Transport Lab, Deputy-Head of the MIPT Center for Cognitive Modeling, expert from AI NTI center (MIPT): “Training of neural networks for computer vision based on MindSpore using GPUs”, (offline) – 30 min
3. Oleg Sirotyuk – Head of Huawei AI presales team: “Huawei Atlas 800 for AI supercomputer centers”, (offline) – 30 min
4. Q+A session – 10-15 min       

Format

Mixed online & offline workshop

Description

• Prof. Chen Lei will briefly introduce MindSpore, a Huawei self-developed end to end deep learning framework.  He will share the motivation of designing such a new framework, then illustrate the unique features of Mindspore, such as automatic parallel execution, deep graph and operator fusion, edge-device-cloud collaborative training and AI-supported scientific computing. Professor will also highlight some future works and projects of MindSpore;
• Dmitry Yudin will focus on practical cases and technical features of training and inference of neural network models of computer vision in Mindspore;
• Oleg Sirotyuk will focus on an issue that China has launched the construction of national supercomputing centers for AI. The main technological component of the centers is Huawei Atlas 800 model 9000 server. During presentation he will present the features of the server, as well as the HW/SW stack of supercomputing centers in China and examples of constructed centers.