Posts Taged cluster

Preferred Networks develops a custom deep learning processor MN-Core for use in MN-3, a new large-scale cluster, in spring 2020

Dec. 12, 2018, Tokyo Japan – Preferred Networks, Inc. (“PFN”, Head Office: Tokyo, President & CEO: Toru Nishikawa) announces that it is developing MN-Core (TM), a processor dedicated to deep learning and will exhibit this independently developed hardware for deep learning, including the MN-Core chip, board, and server, at the SEMICON Japan 2018, held at Tokyo Big Site.  


With the aim of applying deep learning in the real world, PFN has developed the Chainer (TM) open source deep learning framework and built powerful GPU clusters MN-1 and MN-1b, which support its research and development activities. By using these clusters with the innovative software to conduct large-scale distributed deep learning, PFN is accelerating R&D in various areas, such as autonomous driving, intelligent robots, and cancer diagnosis and increasing efforts to put these R&D results to practical use.

To speed up the training phase in deep learning, PFN is currently developing the MN-Core chip, which is dedicated and optimized for performing matrix operations, a process characteristic of deep learning. MN-Core is expected to achieve a world top-class performance per watt of 1 TFLOPS/W (half precision). Today, floating-point operations per second per watt is one of the most important benchmarks to consider when developing a chip. By focusing on minimal functionalities, the dedicated chip can boost effective performance in deep learning as well as bringing down costs.

  • Specifications of the MN-Core chip
    • Fabrication Process : TSMC 12nm
    • Estimated power consumption (W) : 500
    • Peak performance (TFLOPS) :   32.8(DP) / 131(SP) / 524 (HP)
    • Estimated performance per watt (TFLOPS / W) : 0.066 (DP)/ 0.26(SP) / 1.0(HP)

(Notes) DP: double precision, SP: single precision, HP: half precision

https://projects.preferred.jp/mn-core/en/

 

Further improvement in the accuracy and computation speed of pre-trained deep learning models is an essential prerequisite for PFN to work on more complex problems that remain unsolved. It is therefore important to make continued efforts to increase computing resources and make them more efficient. PFN plans to build a new large-scale cluster loaded with MN-Cores, named MN-3, with plans to operate it in the spring of 2020. MN-3 comprises more than 1,000 dedicated server nodes, and PFN intends to increase its computation speed to a target of 2 EFLOPS eventually.

For MN-3 and subsequent clusters, PFN aims to build more efficient computing environments by making use of MN-Core and GPGPU (general-purpose computing on GPU) according to their respective fields of specialty.   

Furthermore, PFN will advance the development of the Chainer deep learning framework so that MN-Core can be selected as a backend, thus utilizing both software and hardware approaches to drive innovations based on deep learning.

 

PFN’s self-developed hardware for deep learning, including MN-Core, will be showcased at its exhibition booth at the SEMICON Japan 2018.

  • PFN exhibition booth at SEMICON Japan 2018
    • Dates/Time: 10:00 to 17:00 Dec. 12 – 14, 2018
    • Venue: Booth #3538, Smart Applications Zone, East Hall 3 at Tokyo Big Site
    • Exhibits:
      (1)  Deep Learning Processor MN-Core, Board, Server
      (2) Preferred Networks Visual Inspection
      (3) Preferred Networks plug&pick robot

 

*MN-Core (TM) and Chainer (TM) are the trademarks or the registered trademarks of Preferred Networks, Inc. in Japan and elsewhere.

Preferred Networks wins second place in the Google AI Open Images – Object Detection Track, competed with 454 teams

Sept. 7, 2018, Tokyo Japan – Preferred Networks, Inc. (PFN, Headquarters: Chiyoda-ku, Tokyo, President and CEO: Toru Nishikawa) participated in the Google AI Open Images – Object Detection Track, an object detection challenge hosted by Kaggle*1, and won second place in the competition among 454 teams from around the world.

 

Object detection, which is one of the major research subjects in computer vision, is a basic technology that is critical for autonomous driving and robotics. Challenges in using large-scale datasets, such as ImageNet and MS COCO, to achieve better accuracy in object detection have been the unifying force of the research community, contributing to the rapid improvement of detection techniques and algorithms.

 

The Google AI Open Images – Object Detection Track held between July 3, 2018 and August 30, 2018 was a competition of an unprecedented scale that used Open Images V4*2, a large and complex dataset released by Google this year. As a result, the event attracted the attention of many researchers. A total of 454 teams from around the world participated in the competition.

PFN entered the competition as team “PFDet”, comprising interested members, mainly developers of ChainerMN and ChainerCV, PFN’s distributed deep learning library and computer vision library based on deep learning, respectively, as well as specialists in the fields of autonomous driving and robotics. During the competition, PFN’s large-scale cluster MN-1b that has 512 NVIDIA (R) Tesla(R) V100 32GB GPUs was in full operation for the first time since its launch in July this year. In addition, the team utilized a parallel deep learning technique to speed up training with a large-scale dataset and made full use of research results PFN had accumulated over the years in the fields of autonomous driving and robotics.  These efforts resulted in the team finishing in a close second place by a narrow margin of 0.023% behind the team who won first place.

 

We have published a paper, entitled “PFDet: 2nd Place Solution to Open Images Challenge 2018 Object Detection Track,” regarding our solution method in this competition, at https://arxiv.org/abs/1809.00778

We also plan to present the content of the paper at a workshop at the European Conference on Computer Vision (ECCV)2018.

 

A part of the techniques developed for this competition will be released as additional functionality to ChainerMN and ChainerCV.

 

PFN will continue to work on research and development of image analysis and object detection technologies, and promote their practical applications in our three primary business domains, namely, transportation, manufacturing, and bio/healthcare.

 

*1:A platform for machine learning competitions

*2:A very large training dataset comprised of 1.7 million images (including 12 million objects of 500 classes)