Apache MXNet 1.7.0 is a minor release that includes backward compatible changes.
As MXNet follows semver, this release includes backwards-compatible fixes and new features only.

Discussion on the roadmap: https://github.com/apache/incubator-mxnet/issues/16864

Release Manager: Ciyong Chen (https://github.com/ciyongch)
Release shepherd: Tao Lv (https://github.com/TaoLv)

Scope

Reference:  MXNet 1.7.0 Roadmap

Issue

PRs

Contributor(s)

Notes

MKLDNN's branding changed to DNNL from v1.1https://github.com/apache/incubator-mxnet/pull/16823Done
AMP to support bfloat16 traininghttps://github.com/apache/incubator-mxnet/pull/17265@rongzha1Done
MKLDNN as the default CPU binary distribution

Already enabled in nightly build. It's binary release scope.
Prebuilt scala package for 2.12https://github.com/apache/incubator-mxnet/issues/16438@clouds56Not start yet

MXNet Extensions

dynamic custom operator: Sparse, RNG

dynamic subgraph: fixes, compile

https://github.com/apache/incubator-mxnet/pull/15921

https://github.com/apache/incubator-mxnet/pull/18063

Done
Saving and loading cudNN autotune and graph optimizationhttps://github.com/apache/incubator-mxnet/issues/16173@QueensGambitNo PR yet
Need a better build script for CMakehttps://github.com/apache/incubator-mxnet/issues/17180No PR yet
Implement remaining nn_conv ops in opperfhttps://github.com/apache/incubator-mxnet/pull/17500@connorgogginsDone
Vectorization for GPUshttps://github.com/apache/incubator-mxnet/pull/17767Przemysław TrędakReverted due to GPU memory usage regression
[mkldnn]Mkldnn bn opt backport from master to 1.7xhttps://github.com/apache/incubator-mxnet/pull/18009@rongzha1Done
Fix for handling negative indices in the fusion of slicehttps://github.com/apache/incubator-mxnet/pull/18041Done
Fix ElemwiseSum for more than 4 inputshttps://github.com/apache/incubator-mxnet/pull/18041Done
GPU gemms true fp16https://github.com/apache/incubator-mxnet/pull/18023@MoisesHerDone
Support projection feature for LSTM on CPU (Only Inference)https://github.com/apache/incubator-mxnet/pull/18038@zixuanweeeiDone
Fix issue of zeros gradients w.r.t. RNN bias when num_layers > 1https://github.com/apache/incubator-mxnet/pull/18038@zixuanweeeiDone
Fix gnu_tls handshake error on Ubuntu 14.04 Nvidia Dockerhttps://github.com/apache/incubator-mxnet/pull/18044@ChaiBapchyaDone
Remove overhead of sg_mkldnn_fullyconnected ophttps://github.com/apache/incubator-mxnet/pull/18039@ciyongchDone








Release Timeline

Following timeline is based on everything goes well.(Added some buffer time)

Step

Tasks

Goal

Actual

Comments

Code Freeze and release startTrack ongoing PRsApril 20th






Cut the release branchCheck license headers



Make code changes with necessary version updatesApril 21th


Cut the release branchApril 20th


Update the version on masterApril 21th

Test the release and tag the releaseNightly test, Jenkins CI



RAT check



Tag RC0July 6th


Tag RC1July 16th






Package artifacts and validateCreate release artifactsJuly 6th


Validate release packageJuly 6th


Test release packageJuly 6th


Scala release process







Begin apache votingStart vote on dev@July 6th


Start vote on general@July 25th






Finalizing and posting the releaseCreate the final release tag on github



Rename, resign and upload the src tar to final dir



Update the website using tag



Release the official pip package



Release the official docker images



After 24 hrs, validate the packages are uploaded



Draft the offical announce email and review



Send out the email on announce@



Update the apache blog



update the aws blog



send internal announcement







CI status: 

Release Note: 1.7.0 Release Notes

  • No labels