Google Git
Sign in
apache / mxnet / 63aea9e0310ed672b8816d332ae9ec25f9d8a893
commit63aea9e0310ed672b8816d332ae9ec25f9d8a893[log] [tgz]
authorAndrzej Kotłowski <Andrzej.Kotlowski@intel.com>Wed Jun 22 08:27:26 2022 +0200
committerGitHub <noreply@github.com>Wed Jun 22 08:27:26 2022 +0200
tree22f8118772108f8e5c402e944f31bd9457434754
parent1ad198d6390d66d2f54dc8820cda6abe754fa0d2 [diff]
[FEATURE] Add quantization for npi_add with oneDNN (#21041)

* [FEATURE] Add quantization for npi_add with oneDNN

+ tests for it,
+ benchmark for FC + npi_add with broadcasted input

* Fix tests in test_fc_subgraph.py

The initialisation of model weights was wrong
as negative numbers couldn't be represented with uint8

* Fix merge

* Fix callibration and scales

+ some rearrangment to make code more readable

* Add tests for full path of quantization of add operators

* Fix test_fc_subgraph.py::test_fc_eltwise[exp-True-True

After previous fix initializing FC weights to only positive values for
uint8, input for exp function was to high.
10 files changed
tree: 22f8118772108f8e5c402e944f31bd9457434754
  1. .github/
  2. 3rdparty/
  3. benchmark/
  4. cd/
  5. ci/
  6. cmake/
  7. config/
  8. contrib/
  9. cpp-package/
  10. docker/
  11. docs/
  12. example/
  13. include/
  14. licenses/
  15. plugin/
  16. python/
  17. src/
  18. tests/
  19. tools/
  20. .asf.yaml
  21. .clang-format
  22. .clang-tidy
  23. .cmakelintrc
  24. .codecov.yml
  25. .git-blame-ignore-revs
  26. .gitattributes
  27. .gitignore
  28. .gitmodules
  29. .licenserc.yaml
  30. .mxnet_root
  31. CMakeLists.txt
  32. CODE_OF_CONDUCT.md
  33. CODEOWNERS
  34. conftest.py
  35. CONTRIBUTORS.md
  36. DISCLAIMER
  37. DNNL_README.md
  38. doap.rdf
  39. LICENSE
  40. NEWS.md
  41. NOTICE
  42. prospector.yaml
  43. pytest.ini
  44. rat-excludes
  45. README.md
  46. readthedocs.yml
  47. SECURITY.md
  48. snap.python
README.md

banner

Apache MXNet (incubating) for Deep Learning

GitHub release (latest SemVer) GitHub stars GitHub forks GitHub contributors GitHub issues good first issue GitHub pull requests by-label GitHub license Twitter Twitter Follow

Apache MXNet is a deep learning framework designed for both efficiency and flexibility. It allows you to mix symbolic and imperative programming to maximize efficiency and productivity. At its core, MXNet contains a dynamic dependency scheduler that automatically parallelizes both symbolic and imperative operations on the fly. A graph optimization layer on top of that makes symbolic execution fast and memory efficient. MXNet is portable and lightweight, scalable to many GPUs and machines.

Apache MXNet is more than a deep learning project. It is a community on a mission of democratizing AI. It is a collection of blue prints and guidelines for building deep learning systems, and interesting insights of DL systems for hackers.

Licensed under an Apache-2.0 license.

BranchBuild Status
masterCentOS CPU Build Status CentOS GPU Build Status Clang Build Status
Edge Build Status Miscellaneous Build Status Sanity Build Status
Unix CPU Build Status Unix GPU Build Status Website Build Status
Windows CPU Build Status Windows GPU Build Status Documentation Status
v1.xCentOS CPU Build Status CentOS GPU Build Status Clang Build Status
Edge Build Status Miscellaneous Build Status Sanity Build Status
Unix CPU Build Status Unix GPU Build Status Website Build Status
Windows CPU Build Status Windows GPU Build Status Documentation Status

Features

  • NumPy-like programming interface, and is integrated with the new, easy-to-use Gluon 2.0 interface. NumPy users can easily adopt MXNet and start in deep learning.
  • Automatic hybridization provides imperative programming with the performance of traditional symbolic programming.
  • Lightweight, memory-efficient, and portable to smart devices through native cross-compilation support on ARM, and through ecosystem projects such as TVM, TensorRT, OpenVINO.
  • Scales up to multi GPUs and distributed setting with auto parallelism through ps-lite, Horovod, and BytePS.
  • Extensible backend that supports full customization, allowing integration with custom accelerator libraries and in-house hardware without the need to maintain a fork.
  • Support for Python, Java, C++, R, Scala, Clojure, Go, Javascript, Perl, and Julia.
  • Cloud-friendly and directly compatible with AWS and Azure.

Contents

What's New

Ecosystem News

Stay Connected

ChannelPurpose
Follow MXNet Development on GithubSee what's going on in the MXNet project.
MXNet Confluence Wiki for Developers MXNet developer wiki for information related to project development, maintained by contributors and developers. To request write access, send an email to send request to the dev list .
dev@mxnet.apache.org mailing listThe “dev list”. Discussions about the development of MXNet. To subscribe, send an email to dev-subscribe@mxnet.apache.org .
discuss.mxnet.io Asking & answering MXNet usage questions.
Apache Slack #mxnet Channel Connect with MXNet and other Apache developers. To join the MXNet slack channel send request to the dev list .
Follow MXNet on Social MediaGet updates about new features and events.

Social Media

Keep connected with the latest MXNet news and updates.

History

MXNet emerged from a collaboration by the authors of cxxnet, minerva, and purine2. The project reflects what we have learned from the past projects. MXNet combines aspects of each of these projects to achieve flexibility, speed, and memory efficiency.

Tianqi Chen, Mu Li, Yutian Li, Min Lin, Naiyan Wang, Minjie Wang, Tianjun Xiao, Bing Xu, Chiyuan Zhang, and Zheng Zhang. MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems. In Neural Information Processing Systems, Workshop on Machine Learning Systems, 2015