Deep Learning Framework Caffe-MPI for KNL Released
Poster: SySAdmin
Posted on June 21, 2016 at 6:56:01 AM
Deep Learning Framework Caffe-MPI for KNL Released
FRANKFURT, Germany, June 21, 2016 /PRNewswire/ --Inspur released the open source deep learning framework Caffe-MPI for KNL at ISC16. Intel announced the official launching of the second generation of the Intel® Xeon Phi(TM) chip "Knights Landing" (KNL), and Inspur has completed the parallel development of Caffe-MPI based on this latest platform - signifying its mark as the first company to do so.
Hu Leijun, the Vice President of Inspur, expressed that the release of Caffe-MPI based on KNL platform not only demonstrates the close partnership between Inspur and Intel, it also allows the opportunity to bring new acceleration solutions to global deep learning users - enabling them to select the acceleration technology that is most in accordance with their own applications. Furthermore, the open-source Caffe-MPI has been generating the attention of many companies and research institutes from China, India and the US, amongst other places.
Caffe-MPI is the first Caffe deep learning computing framework of high-performance MPI cluster version and it adopts matured MPI technology to optimize the data parallel of Caffe - solving the efficiency problem in deep learning computing model training. Caffe is currently the deep convolutional neural network (CNN) architecture and designed for the CNN training to complete the development of single machine with multi cards by UC Berkeley Lab. However, with the increasingly complex training model and increasing training data, the single machine can no longer meet the actual demand of users.
Therefore, Inspur developed the parallelization of multi-node - while still preserving the characteristics of the original caffe architecture. Caffe-MPI is capable of adopting multi-machine with multiply cards to train simultaneously, while deploying to large-scale training platform to achieve the training of large-scale data samples.
The new generation of Intel® Xeon Phi(TM) processor-Knights Landing can not only be used as co-processor, but also can be used as host central processing unit - deemed to be the revolutionary product in the field of high-performance computing and deep learning. With the release of Inspur's Caffe-MPI based on KNL platform, users will quickly recognize the great improvement in deep learning efficiency.
SOURCE Inspur Group Co., Ltd.
Inspur Group Co., Ltd.
CONTACT: Yan Panpan, +86-10-8258-1473, yanpanpan@Inspur.com
FRANKFURT, Germany, June 21, 2016 /PRNewswire/ --Inspur released the open source deep learning framework Caffe-MPI for KNL at ISC16. Intel announced the official launching of the second generation of the Intel® Xeon Phi(TM) chip "Knights Landing" (KNL), and Inspur has completed the parallel development of Caffe-MPI based on this latest platform - signifying its mark as the first company to do so.
Hu Leijun, the Vice President of Inspur, expressed that the release of Caffe-MPI based on KNL platform not only demonstrates the close partnership between Inspur and Intel, it also allows the opportunity to bring new acceleration solutions to global deep learning users - enabling them to select the acceleration technology that is most in accordance with their own applications. Furthermore, the open-source Caffe-MPI has been generating the attention of many companies and research institutes from China, India and the US, amongst other places.
Caffe-MPI is the first Caffe deep learning computing framework of high-performance MPI cluster version and it adopts matured MPI technology to optimize the data parallel of Caffe - solving the efficiency problem in deep learning computing model training. Caffe is currently the deep convolutional neural network (CNN) architecture and designed for the CNN training to complete the development of single machine with multi cards by UC Berkeley Lab. However, with the increasingly complex training model and increasing training data, the single machine can no longer meet the actual demand of users.
Therefore, Inspur developed the parallelization of multi-node - while still preserving the characteristics of the original caffe architecture. Caffe-MPI is capable of adopting multi-machine with multiply cards to train simultaneously, while deploying to large-scale training platform to achieve the training of large-scale data samples.
The new generation of Intel® Xeon Phi(TM) processor-Knights Landing can not only be used as co-processor, but also can be used as host central processing unit - deemed to be the revolutionary product in the field of high-performance computing and deep learning. With the release of Inspur's Caffe-MPI based on KNL platform, users will quickly recognize the great improvement in deep learning efficiency.
SOURCE Inspur Group Co., Ltd.
Inspur Group Co., Ltd.
CONTACT: Yan Panpan, +86-10-8258-1473, yanpanpan@Inspur.com