Amazon Web Services has decided to release the code behind one of its key machine-learning services as an open-source project, as it continues to push back against critics who find its relationship with open-source software out of balance.
SageMaker Neo, which lets customers run machine-learning training models across multiple operating environments, was introduced last November at re:Invent 2018. On Thursday AWS announced that it would be releasing that code, itself derived in part from other open-source projects, as the Neo-AI Project under the Apache Software License.
Neo-AI will allow chip makers to support models built using several popular machine-learning frameworks, including “TensorFlow, MXNet, PyTorch, ONNX, and XGBoost models,” it said in a blog post. Intel, Nvidia, and Arm are on board with the first release, while support from Xilinx, Cadence, and Qualcomm arriving later. Neo AI uses TVM, an open deep-learning compiler that originated in at the University of Washington from the Paul G. Allen School’s SAMPL group.
Machine learning has immense potential to transform application development, but only a handful of companies can afford the expertise required to build, design, and maintain the tools needed to make everything work. AWS is one of those companies, and all the major cloud providers consider machine learning one of the most important emerging technologies in this world.
The release of Neo-AI gives chip makers and software developers a way to validate machine-learning models for lots of different types of devices, which previously required a lot of custom work. “By making optimization easier, Neo-AI allows sophisticated models to run on resource-constrained devices, where they can unlock innovation in areas such as autonomous vehicles, home security, and anomaly detection,” AWS said in the post.
The release is also another sign that AWS increasing involvement with the open-source community, after years of criticism over its tendency to use open-source projects as the foundation for revenue-generating services without contributing much back to the community. Neo-AI joins Firecracker, which was also unveiled at re:Invent 2018, as another fundamental technology advance that the cloud leader has decided to release as an open-source project.