Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

Issue

Proposer

PRs

Contributor(s)

Notes

Update to MKLDNN 1.0@pengzhao-intelhttps://github.com/apache/incubator-mxnet/pull/16555@TaoLvDone 10/31
Enabling MKLDNN Subgraph fusion by default@pengzhao-intelhttps://github.com/apache/incubator-mxnet/pull/15518@ZhennanQinDone 7/26
Quantization API improvement@pengzhao-intelhttps://github.com/apache/incubator-mxnet/pull/15448@zinyu-intelDone 8/1
Enhancements to INT64 support@sandeep-krishnamurthy


Improvements to Estimator API@sandeep-krishnamurthy


BERT optimizations for GPU@ptrendx

https://github.com/apache/incubator-mxnet/pull/15545

https://github.com/apache/incubator-mxnet/pull/15167

https://github.com/apache/incubator-mxnet/pull/15657

https://github.com/apache/incubator-mxnet/pull/16039

https://github.com/apache/incubator-mxnet/pull/16398

https://github.com/apache/incubator-mxnet/pull/16122

https://github.com/apache/incubator-mxnet/pull/16355

https://github.com/apache/incubator-mxnet/pull/16408

https://github.com/apache/incubator-mxnet/pull/16615

@ptrendx

@MoisesHer

@drivanov

@DickJC123

@Caenorst

@nvchai

@blchu

15545 -  done 8/21

15167 - Done 10/31

15657 - Done 11/1

16039 - done 9/24

16398 - done 10/19

16122 - done 9/30

16355 - done 10/5

16408 - Under review

16615 - done 10/25

Fixing nightly tests

https://github.com/apache/incubator-mxnet/issues/15374

@roywei


NumPy compatibility@reminisce


Better TensorRT documentation@KellenSunderlandhttps://github.com/apache/incubator-mxnet/pull/14860
Under review

Multithreaded inference

https://github.com/apache/incubator-mxnet/issues/16431

@anirudh2290

The PR is expected Friday 25th Oct.
Faster implementation of NMS on GPU@ptrendxhttps://github.com/apache/incubator-mxnet/pull/16542@ptrendxUnder review
Fix large tensor for log_softmaxChaitanya Prakash Bapat https://github.com/apache/incubator-mxnet/pull/16711@haojin2Done 11/02








































Release Timeline

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

...