Amazon Web Services has released relevant code from its SageMaker Neo machine learning, a project licensed by Apache software. SageMaker Neo was first launched on AWS Re: Invent in November 2018 to help developers optimize existing machine learning models for hardware platforms.
Similarly, Neo-AI will work with chip manufacturers, device manufacturers and developers to optimize machine learning models for various hardware platforms.
AWS says that Neo-AI will automatically optimize the models of TensorFlow, MXNet, PyTorch, ONNX and XGBoost frameworks, and may double the speed of these models without losing accuracy.
The platform also supports the conversion of different models into common formats to reduce compatibility issues.
At present, Intel, NVIDIA and ARM-related hardware currently support Neo-AI, Xilinx, Cadence and Qualcomm hardware will be launched later.