lars_multi_mp_sgd_update kernel used in MLPERF v0.7 not found in v1.6 or v1.7

[Eric-seu]

Description

In the submission of mxnet in mlperf training v0.7 benchmark, mxnet used a optimizer kernel call "lars_multi_mp_sgd_update" or "lars_multi_mp_sgd_mom_update".

Mxnet used this optimizer and got great results in mlperf v0.7, and I would like to try it in my project.
BUT i could not find this kernel in mxnet v1.6 or v1.7.

Error Message

What have you tried to solve it?

i had tried with the master branches of v1.6 and v1.7

[github-actions[bot]]

Welcome to Apache MXNet (incubating)! We are on a mission to democratize AI, and we are glad that you are contributing to it by opening this issue.
Please make sure to include all the relevant context, and one of the @apache/mxnet-committers will be here shortly.
If you are interested in contributing to our project, let us know! Also, be sure to check out our guide on contributing to MXNet and our development guides wiki.

[ptrendx]

You can use LARS optimizer - it uses the same kernels as the MLPerf submission (they are named slightly differently - preloaded_multi_mp_sgd_update). The NVIDIA NGC MXNet container, which was used for that submission, had the LARS optimizer before the upstream and the names were changed as part of the upstreaming process.