oneAPI Deep Neural Network Library (oneDNN) is an open-source performance library for deep learning applications. The library includes basic building blocks for neural networks optimized for Intel® Architecture Processors and Intel® Processor Graphics. oneDNN is intended for deep learning applications and framework developers interested in improving application performance on Intel® CPUs and GPUs
Github: https://github.com/oneapi-src/oneDNN
Code samples are licensed under the MIT license. See License.txt for details.
Third party program Licenses can be found here: third-party-programs.txt
| Type | Name | Description |
|---|---|---|
| Component | getting_started | The sample also includes a Jupyter notebook with step by step instructions on building code with different compilers and runtime configurations oneDNN support. |
| Component | verbose_jitdump | This Jupyter Notebook demonstrates how to use Verbose Mode and JIT Dump to profile oneDNN samples. |
| Component | analyze_isa_with_dispatcher_control | This Jupyter Notebook demonstrates how to use CPU Dispatch Control to generate JIT codes among different ISA on CPU and also analyze JIT kernels among ISAs. |
| Component | Intel® VTune™ Profiler | This Jupyter Notebook demonstrates how to use VTune™ Profiler to profile oneDNN samples and find out performance bottlenecks. |
| Component | benchdnn_tutorial | This Jupyter Notebook demonstrates how to use the benchDNN tool to validate and test oneDNN primitive executions |
Notice : Please use Intel® oneAPI DevCloud as the environment for jupyter notebook samples.
Users can refer to DevCloud Getting Started for using DevCloud
Users can use JupyterLab from DevCloud via "One-click Login in", and download samples via "git clone" or the "oneapi-cli" tool
Once users are in the JupyterLab with downloaded jupyter notebook samples, they can start following the steps without further installation needed.