oneDNN
An open-source cross-platform performance library of basic building blocks for deep learning applications, optimized for CPUs and GPUs.
What is oneDNN?
oneDNN is an open-source, cross-platform performance library that provides fundamental building blocks for deep learning applications. It implements the oneAPI specification for deep neural network operations, enabling framework developers to accelerate their applications across a wide range of CPU and GPU hardware. The library abstracts hardware-specific optimizations to deliver efficient execution on diverse architectures.
Deep learning framework and application developers who need to optimize neural network performance on CPUs and GPUs, particularly those integrating with major frameworks like PyTorch, TensorFlow, ONNX Runtime, and OpenVINO.
Developers choose oneDNN for its vendor-agnostic, high-performance design that leverages just-in-time (JIT) code generation and supports multiple runtimes (OpenMP, TBB, SYCL) and hardware platforms. Its unique selling point is providing optimized deep learning primitives that allow frameworks to focus on higher-level logic while ensuring efficient execution across Intel, AMD, Arm, NVIDIA, and experimental architectures.
Overview
oneAPI Deep Neural Network Library (oneDNN)
Use Cases
Best For
- Accelerating deep learning frameworks like PyTorch and TensorFlow on Intel and AMD CPUs with JIT optimization.
Related Projects
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
