Update NEWS, README and website for 1.7.0 #19047
Conversation
|
Hey @ciyongch , Thanks for submitting the PR
CI supported jobs: [unix-cpu, centos-gpu, miscellaneous, windows-cpu, unix-gpu, windows-gpu, edge, clang, website, sanity, centos-cpu] Note: |
|
It's strange. I remember I changed the download table to use mirror links in #16501. |
There was a problem hiding this comment.
Must use archive.apache.org for all but the latest release. See https://infra.apache.org/release-publishing.html and https://infra.apache.org/release-distribution.html#download-links
Sure, I will update the download link according to this spec. Thanks! |
|
Hi @leezu , I've updated the download link, please help to take a review. |
| - [1.7.0](#170) | ||
| - [New features](#new-features) | ||
| - [MXNet Extensions: custom operators, partitioning, and graph passes](#mxnet-extensions-custom-operators-partitioning-and-graph-passes) | ||
| - [OpPerf utility enabled in the binary distribution](#opperf-utility-enabled-in-the-binary-distribution) |
There was a problem hiding this comment.
Hi @ChaiBapchya , probably it's not drived by any PR in the current scope, but included in the binary release process. cc @szha to see if any comments.
When trying to install mxnet release (1.7.0) wheel via pip, the benchmark/opperf is also installed now. And the original proposal from you is listed in 1.7.0 roadmap
|
Hi @leezu @szha @marcoabreu @ChaiBapchya @TaoLv @pengzhao-intel , please help to take a review and comment if anything needs to be updated, thanks! |
| {% highlight bash %} | ||
| pip install mxnet | ||
| {% endhighlight %} | ||
|
|
||
| Start from 1.7.0 release, MKL-DNN is enabled in pip packages by default. Which are | ||
| optimized for Intel hardware. You can find performance numbers | ||
| in the <a href="https://mxnet.io/api/faq/perf#intel-cpu">MXNet tuning guide</a>. |
There was a problem hiding this comment.
The perf link under mxnet.apache.org will contain the version number like "https://mxnet.apache.org/versions/1.6/api/faq/perf.html#intel-cpu". Is there any basic link without version number which can redirect to the latest version of the page?
Please check my latest update.
| {% highlight bash %} | ||
| pip install mxnet | ||
| {% endhighlight %} | ||
|
|
||
| Start from 1.7.0 release, MKL-DNN is enabled in pip packages by default. Which are |
| {% highlight bash %} | ||
| pip install mxnet | ||
| {% endhighlight %} | ||
|
|
||
| Start from 1.7.0 release, MKL-DNN is enabled in pip packages by default. Which are | ||
| optimized for Intel hardware. You can find performance numbers |
There was a problem hiding this comment.
Elaborate on the "optimization? For example "oneAPI Deep Neural Network Library (oneDNN) is an open-source cross-platform performance library of basic building blocks for deep learning applications. The library is optimized for Intel Architecture Processors, Intel Processor Graphics and Xe architecture-based Graphics. Support for other architectures such as Arm* 64-bit Architecture (AArch64) and OpenPOWER* Power ISA (PPC64) is experimental." (from the oneDNN repo).
Please mention that the mxnet-native release is available which does come without oneDNN
There was a problem hiding this comment.
Thank you for the valuable inputs, leezu, I will update the description here.
There was a problem hiding this comment.
BTW, I didn't find available mxnet-native pip wheel via pip install mxnet-native, is it still WIP or using the different name?
|
@mxnet-bot run ci [windows-gpu] |
|
Jenkins CI successfully triggered : [windows-gpu] |
Update NEWS README and website for 1.7.0 release.
@szha @TaoLv @leezu