| The name of the base system (platform): | Amazon Web Services (AWS) |
| Developers: | Amazon |
| Date of the premiere of the system: | November, 2018 |
| Last Release Date: | January, 2019 |
| Branches: | Information technologies |
The Amazon SageMaker Neo service is intended for optimization of the constructed models for the purpose of acceleration of work and decrease in resource consumption. The solution allows to train model once, and then to unroll it anywhere. Using machine learning technologies of SageMaker Neo automatically optimizes the trained model created in AWS for a specific hardware platform. After that the model can be unrolled in instansa of EC2 and SageMaker or on any peripheral device with Neo runtime environment, including on AWS Greengrass devices.
2019: Disclosure of source codes
In January, 2019 Amazon opened source codes of the Amazon SageMaker Neo service allowing to construct and train once of model of machine learning for PyTorch, and then to unroll them in a cloud of AWS with a double performance.
The platform is issued within the Neo-AI project under the license Apache Software License and after disclosure of source codes will allow producers of chips and devices and also software developers to optimize models of machine learning for a broad spectrum of hardware platforms.
According to the statement of Amazon, Neo-AI automatically optimizes models for the TensorFlow, MXNet, PyTorch, ONNX and XGBoost platforms and can double the speed of work of these models without loss of accuracy. Service will also transform models to the general format to reduce compatibility issues.
 | As a rule, optimization of model of machine learning for several hardware platforms is a difficult task as developers need to configure models manually for configuring hardware and the software of each platform — Sukvon Kim and Vin Sharma from Amazon Web Services (AWS) write in the blog. — It is especially difficult for peripheral devices which are usually limited in the computing power and memory". |  |
The first version of Neo-AI was supported by the companies Intel, Nvidia and Arm, and chipmakers Xilinx, Cadence and Qualcomm are going to join them.[1]
Notes
| The site content is translated by machine translation software powered by PROMT. The machine-translated articles are not always perfect and may contain errors in vocabulary, syntax or grammar. Read original article If you find inaccuracies or errors in the results of machine translation, please write to editor@tadviser.ru. We will make every effort to correct them as soon as possible. |