Home > News content

Amazon announced MXNet as the AWS deep learning framework

via:博客园     time:2016/12/9 22:30:40     readed:2233

English Original:Amazon Announces MXNet as Deep Learning Framework of Choice at AWS

AmazonWerner VogelsAnnounced last week that the Amazon Deep Learning Framework will be availableformalSelectionMXNet, And AWS will contribute to the long-term success of MXNet by increasing source code contributions, improving documentation, and supporting visualization, development, and migration tools from other frameworks.

Vogles pointed out that in the field of fraud detection, recommendation pipeline, inventory and product inspection audit, there are a series of computing tasks that can not be realized by writing explicit algorithms,Depth learningofrobotic leanringMethods are increasingly playing an important role, in addition, in content search, autonomous UAV, order fulfillment center robot, text and speech recognition and other fields also widely used machine learning method. Vogels gives the three factors that Amazon considers in the selection of the depth learning framework: scalability, development speed and portability.

Deep Learning LibraryCaffe, & Lt;CNTK, MXNet,TensorFlow, & Lt;TheanoandTorchHave been evaluated by Amazon and supported by AWS. Now AWS has selected MXNet as an extensible framework, and called on the open source community to put more effort into MXNet. In the development of machine learning platform services, AWS will take the same approach as in RDS, and:

"We will support all of the popular depth learning frameworks in the depth learning framework by providing the best EC2 instance groups and applicable software tools." & Rdquo;

Vogels mentioned the depth of learning that was released earlier this yearAMIAnd incidental cloud informationtemplate. The AMI toolset is a 64-bit release of Amazon Linux with pre-installed support for CNTK, as well as support for MXNet,Graphviz, & Lt;PygalAnd PythonPandasOf the update package. The release builds on six advanced learning architectures: NXNet, Caffe, TensorFlow, Theano, Torch, and CNTK. AMI also includes NVIDIA CUDAToolkitandCuDNNLibrary installer,Anaconda, Python2and3. Commentary shows that this is still the first generation of AMI, it has been able to start using the GPU architecture analysis engineers to provide a good working basis.

MXNet was originally developed byWashington UniversityAnd Carnegie Mellon UniversityCMU) Was developed to support convolutional neural networks (CNN) And long-short-term memory networksLSTM) Operation. CMU Department of Computer Science Andrew Moore pointed out:

MXNet was born in CMU, and in CMU development and growth. It is the most extensible depth learning framework I have ever seen, and it is also a representative achievement of brilliant research in computer science. A number of different disciplines intersect in which and work together to achieve the linear algebra will be creatively introduced into large-scale distributed computing, leading to a new situation in-depth learning. "We are excited to see Amazon investing in MXNet and look forward to the continued strength of MXNet."

The CMU trained the Inception V3 algorithm on MXNet. The training runs on a cluster of P2 instances and gradually increases the number of GPUs. When running up to 1000 layers of deep network, MXNet only takes up to 4GB of memory. MXNet supports multi-language API interfaces including Python, C ++ (and supports compilation on Android and iOS), R, Scala, Julia, Matlab, and JavaScript.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments