mxnet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sandeep krishnamurthy <sandeep.krishn...@gmail.com>
Subject Re: MKLDNN Integration Stable Release
Date Wed, 04 Jul 2018 03:26:49 GMT
* If all existing RNN integration tests pass with MKL-DNN build, this
should give enough confidence?
* Also, I remember one of the community member saying "mxnet-mkl" pypi
package is not compiled with MKLDNN. Not sure about this, but, can we
please confirm?

Best,
Sandeep

On Tue, Jul 3, 2018 at 7:37 PM Zhao, Patric <patric.zhao@intel.com> wrote:

> Hi Alex,
>
> Regarding RNN, the first version of MKL-DNN RNN API is available in the
> MKL-DNN master branch.
> We have integrated it in our local branch and you can try our code (still
> in developments).
>
>
> https://github.com/lihaofd/incubator-mxnet/blob/mkldnn-rnn/src/operator/nn/mkldnn/mkldnn_rnn_impl.h
>
> We plan to PR our integration into MXNET master when both functionality
> and performance are qualified.
>
> Thanks,
>
> --Patric
>
> > -----Original Message-----
> > From: Alex Zai [mailto:azai91@gmail.com]
> > Sent: Wednesday, July 4, 2018 1:17 AM
> > To: dev@mxnet.incubator.apache.org
> > Subject: MKLDNN Integration Stable Release
> >
> > We are preparing a stable release of MKL-DNN integration in 1.3.0
> > (experimental since 1.2.0), which supports acceleration of operations
> such as
> > Convolution, Deconvolution, FullyConnected, Pooling, Batch Normalization,
> > Activation, LRN, Softmax. Currently the RNN operator is not supported as
> the
> > MKL-DNN API is still experimental; however, they hope to release a more
> > stable version RNN API this or next week in MKL-DNN 0.15.
> >
> > We will have CPP unit test support on these operators and I am planning
> to
> > write python unit tests to compare a RNN network's results from the
> MKLDNN
> > backend with that of the GPU to test accuracy. Is there any additional
> coverage
> > that you think we should cover in the next two weeks?
> >
> > Alex
>


-- 
Sandeep Krishnamurthy

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message