OneAPI Deep Neural Network Library: Difference between revisions

oneAPI Deep Neural Network Library
Repository	github.com/oneapi-src/oneDNN
Engine
Operating system	Microsoft Windows, Linux, macOS
Platform	Cross-platform
Type	Open-source Library
Website	spec.oneapi.io/versions/latest/elements/oneDNN/source/index.html

Newer edit →

VisualWikitext

Revision as of 20:16, 25 October 2021

Search OneAPI Deep Neural Network Library on Amazon.

oneAPI Deep Neural Network Library (oneDNN) is a performance library of basic building blocks for deep learning applications.^[1] oneDNN is an open-source and cross-platform library element of oneAPI^[2]^[3].

oneDNN is useful for deep learning application and framework developers to improve application performance ^[4]^[5]^[6]using the same API for both CPUs and GPUs. It abstracts out the specific instruction set and other complexities of performance optimizations. The oneDNN library is available for Windows, Linux and macOS operating systems.

oneDNN provides optimizations for popular deep learning frameworks including:

TensorFlow*^[7]
PyTorch*^[8]

History

oneDNN started as the Deep Neural Networks (DNN) component of Intel MKL library first released in Intel MKL 2017.
Considering fast development in the field and opportunity for industry collaboration Intel launched an open source project named Intel MKL-DNN on August 28, 2016.
The Intel MKL DNN component was deprecated in Intel MKL 2019 and removed in Intel MKL 2020.
MKL-DNN reached production maturity with MKL-DNN v1.0 on Jul 12, 2019.
Arm contributed support for AArch64 processors in oneDNN v1.5
The library became a part of oneAPI initiative and renamed to oneDNN in oneDNN v2.0 on Dec 8, 2020.
Intel branded the binary distribution Intel® oneAPI Deep Neural Network Library (oneDNN) launched as part of Intel® oneAPI 2021 release.
Codeplay contributed support for NVIDIA GPUs in oneDNN v2.1

License

Apache License 2.0

References

↑ "oneAPI Deep Neural Network Library". oneAPI.
↑ "oneAPI".
↑ Deep Learning Systems. Google Books: Morgan & Claypool Publishers. ISBN 9781681739670. Search this book on
↑ "oneDNN benchmarks". openbenchmarking.
↑ "Developing Deep Learning Frameworks for Exascale". HPC Wire.
↑ "Optimizing Inference Performance of Transformers on CPUs" (PDF). arXiv. Cornell University.
↑ "TensorFlow and oneDNN in Partnership" (PDF). oneAPI.
↑ "Optimize the Latest Deep Learning Workloads using Intel-optimized PyTorch". Argonne Leadership Computing Facility.

External links

oneAPI industry specification
oneAPI oneDNN specification
developer guide explains the programming model, functionality, and implementation details, and includes examples
oneDNN API reference provides a comprehensive reference of the library API.
Intel® oneAPI Deep Neural Network Library

This article "OneAPI Deep Neural Network Library" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:OneAPI Deep Neural Network Library. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.

[1] "oneAPI Deep Neural Network Library". oneAPI.

[2] "oneAPI".

[3] Deep Learning Systems. Google Books: Morgan & Claypool Publishers. ISBN 9781681739670. Search this book on

[4] "oneDNN benchmarks". openbenchmarking.

[5] "Developing Deep Learning Frameworks for Exascale". HPC Wire.

[6] "Optimizing Inference Performance of Transformers on CPUs" (PDF). arXiv. Cornell University.

[7] "TensorFlow and oneDNN in Partnership" (PDF). oneAPI.

[8] "Optimize the Latest Deep Learning Workloads using Intel-optimized PyTorch". Argonne Leadership Computing Facility.

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

@@ Line 16: / Line 16: @@
 }}
-'''oneAPI Deep Neural Network Library''' ('''oneDNN''') is a performance [[Library (computing)|library]] of basic building blocks for [[Deep_learning|deep learning]] applications.<ref>{{cite web |url=https://github.com/oneapi-src/oneDNN |title=github oneDNN}}</ref><ref>{{cite web |title=oneDNN 2.2 Released With More Optimizations For Alder Lake, Sapphire Rapids |url=https://www.phoronix.com/scan.php?page=news_item&px=Intel-oneDNN-2.2 |website=Phoronix}}</ref> oneDNN is an open-source and cross-platform library element of [[OneAPI_(compute_acceleration)|oneAPI]]<ref>{{cite web |url=https://www.oneapi.io/ |title=oneAPI}}</ref>.
+'''oneAPI Deep Neural Network Library''' ('''oneDNN''') is a performance [[Library (computing)|library]] of basic building blocks for [[Deep_learning|deep learning]] applications.<ref>{{cite web |title=oneAPI Deep Neural Network Library |url=https://spec.oneapi.io/versions/0.5.0/oneAPI/Elements/onednn/onednn_root.html |website=oneAPI}}</ref> oneDNN is an open-source and cross-platform library element of [[OneAPI_(compute_acceleration)|oneAPI]]<ref>{{cite web |url=https://www.oneapi.io/ |title=oneAPI}}</ref><ref>{{cite book |title=Deep Learning Systems |publisher=Morgan & Claypool Publishers |location=Google Books |isbn=9781681739670 |url=https://www.google.com/books/edition/Deep_Learning_Systems/OKAFEAAAQBAJ}}</ref>.
-oneDNN is useful for deep learning application and framework developers to improve application performance<ref>{{cite web |title=Software AI accelerators: AI performance boost for free |url=https://venturebeat.com/2021/09/22/software-ai-accelerators-ai-performance-boost-for-free/ |website=Venture Beat}}</ref><ref>{{cite web |title=oneDNN hits v1.4 |url=https://devclass.com/2020/04/20/intel-onednn-hits-v1-4-with-drumroll-better-intel-support/ |website=Dev Class}}</ref> using the same API for both CPUs and GPUs. It abstracts out the specific instruction set and other complexities of performance optimizations.  The oneDNN library is available for [[Windows]], [[Linux]] and [[macOS]] [[operating system]]s.
+oneDNN is useful for deep learning application and framework developers to improve application performance <ref>{{cite web |title=oneDNN benchmarks |url=https://openbenchmarking.org/test/pts/onednn |website=openbenchmarking}}</ref><ref>{{cite web |title=Developing Deep Learning Frameworks for Exascale |url=https://www.hpcwire.com/off-the-wire/aurora-software-development-developing-deep-learning-frameworks-for-exascale/ |website=HPC Wire}}</ref><ref>{{cite web |title=Optimizing Inference Performance of Transformers on CPUs |url=https://arxiv.org/pdf/2102.06621.pdf |website=arXiv |publisher=Cornell University}}</ref>using the same API for both CPUs and GPUs. It abstracts out the specific instruction set and other complexities of performance optimizations.  The oneDNN library is available for [[Windows]], [[Linux]] and [[macOS]] [[operating system]]s.
-oneDNN provides optimizations for popular frameworks including:
+oneDNN provides optimizations for popular deep learning frameworks including:
-* TensorFlow*<ref>{{cite web |title=Leverage Intel Deep Learning Optimizations in TensorFlow |url=https://medium.com/intel-analytics-software/leverage-intel-deep-learning-optimizations-in-tensorflow-129faa80ee07 |website=Medium}}</ref><ref>{{cite web |title=TensorFlow 2.5.0 Released |url=https://analyticsindiamag.com/tensorflow-2-5-0-released-all-major-updates-features/ |website=Analytics India Magazine}}</ref>
+* TensorFlow*<ref>{{cite web |title=TensorFlow and oneDNN in Partnership |url=https://www.oneapi.io/wp-content/uploads/sites/74/Penporn-Koanantakool-TensorFlow-and-oneDNN-in-Partnership.pdf |website=oneAPI}}</ref>
-* PyTorch*<ref>{{cite web |title=Accelerate PyTorch with IPEX and oneDNN |url=https://medium.com/pytorch/accelerate-pytorch-with-ipex-and-onednn-using-intel-bf16-technology-dca5b8e6b58f |website=Medium}}</ref><ref>{{cite web |title=Optimize the Latest Deep Learning Workloads using Intel-optimized PyTorch |url=https://www.alcf.anl.gov/events/optimize-latest-deep-learning-workloads-using-intel-optimized-pytorch |website=Argonne Leadership Computing Facility}}</ref>
+* PyTorch*<ref>{{cite web |title=Optimize the Latest Deep Learning Workloads using Intel-optimized PyTorch |url=https://www.alcf.anl.gov/events/optimize-latest-deep-learning-workloads-using-intel-optimized-pytorch |website=Argonne Leadership Computing Facility}}</ref>
 ==History==
@@ Line 46: / Line 46: @@
 * [https://oneapi-src.github.io/oneDNN/group_dnnl_api.html oneDNN API reference] provides a comprehensive reference of the library API.
 * [https://software.intel.com/content/www/us/en/develop/tools/oneapi/components/onednn.html Intel® oneAPI Deep Neural Network Library]
+{{AfC submission|||ts=20211020202327|u=Python4all|ns=118}}
 {{⚠️🚨COPIED from EverybodyWiki ❗❕⚠️😡😤Please respect Licence CC-BY-SA ❗}}
 {{Source Wikipedia}}