mxnet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Barber, Christopher" <Christopher.Bar...@analog.com>
Subject Re: Display the master branch website in default for a period
Date Fri, 02 Mar 2018 03:54:46 GMT
I was thinking more along the lines of benchmarks of MXNet vs TensorFlow, PyTorch, and Caffe2.
Benchmarks of edge devices would definitely be interesting, but I would also want to see benchmarks
of training time and memory use and accuracy on large models. Obviously this would be a non-trivial
amount of work, which is why no one else is doing it, but there would be a lot of interest
in this. Also would like to see benchmarks of ndarray, vs symbol vs gluon.

But yes, if you want to drive traffic to the website you should have content that changes
frequently. I have to say I find it really strange to have the entire website change when
I select a different version from the top tab. The design of the website should be independent
of the code version.

- Christopher

On 3/1/18, 4:33 PM, "Marco de Abreu" <marco.g.abreu@googlemail.com> wrote:

    As far as I know, there are plans to make regular benchmarks and generate
    statistics. We could use that data. My personal task after CI is creating
    an infrastructure to automatically perform performance and power
    consumption benchmarks on edge devices (raspberry and Nvidia Jetson). It
    would definitely be a good idea to share this data with the community
    (especially considering the impressive performance of MXNet).
    
    Aaron is currently gathering requirements for recreating the website build
    and publish process, so input like this is definitely helpful. This could
    basically be summarized as a requirement to make the website contain static
    parts (e.g. APIs and documentation) as well as dynamic parts (e.g. news,
    statistics, recent papers etc).
    
    How does that sound?
    
    -Marco
    
    Aaron Markham <aaron.s.markham@gmail.com> schrieb am Do., 1. März 2018,
    22:11:
    
    > Do you have specific public benchmarks in mind?
    >
    > On Mar 1, 2018 10:13, "Barber, Christopher" <Christopher.Barber@analog.com
    > >
    > wrote:
    >
    > > I think one thing that could draw more users would be published
    > benchmarks
    > > that show that networks implemented using MXNet perform better than
    > > comparable ones using other platforms using the same hardware. If you can
    > > definitively show that MXNet is much faster and/or uses much less memory,
    > > you will attract much more interest.
    > >
    > > - Christopher
    > >
    > > On 2/25/18, 11:53 PM, "Li, Mu" <mli@amazon.com> wrote:
    > >
    > >     That's a great idea. The thing Simon's team is doing is publishing
    > > more tutorials on both MXNet website and AWS blog, which may attract a
    > lot
    > > of traffics. Also Sukwon is tracking the progress of publishing news more
    > > frequently on the homepage.
    > >
    > >     Best,
    > >     Mu
    > >
    > >     > On Feb 25, 2018, at 8:48 PM, Chris Olivier <cjolivier01@gmail.com>
    > > wrote:
    > >     >
    > >     > My (probably less-than-$0.02):
    > >     >
    > >     > I have as my home page on my
    > >     > phone, the Google Research Blog, and they frequently release stuff
    > > like
    > >     > data sets and models do this or that.  Usually it seems pretty
    > > interesting
    > >     > and I am compelled to try it.
    > >     >
    > >     > Maybe we do something similar, but I’m not aware of it. I know we
    > > have all
    > >     > sorts of examples and whatnot, but it doesn’t seem the same as what
    > > at
    > >     > least appears to be something new to play with scrolling past every
    > > couple
    > >     > of weeks:
    > >     >
    > >     > For example, a few days ago:
    > >     >
    > >     > “Introducing the HDR+ Burst Photography Dataset”.
    > >     >
    > >     > https://research.googleblog.com/2018/02/introducing-hdr-
    > > burst-photography.html?m=1
    > >     >
    > >     > Reading that makes me want to download it and play around.
    > > Obviously I
    > >     > would use Tensorflow by default because it’s ready to roll as-is
    > > with this
    > >     > dataset/model.
    > >     >
    > >     >
    > >     > -Chris
    > >     >
    > >     >
    > >     >
    > >     >> On Sun, Feb 25, 2018 at 8:34 PM Mu Li <limu.cn@gmail.com>
wrote:
    > >     >>
    > >     >> Not sure why the screenshot of the page view is not there, attach
    > > it again:
    > >     >>
    > >     >>
    > >     >> Best,
    > >     >> Mu
    > >     >>
    > >     >>> On Feb 25, 2018, at 8:32 PM, Li, Mu <mli@amazon.com>
wrote:
    > >     >>>
    > >     >>> Hi Team,
    > >     >>>
    > >     >>> We are in trouble attracting new users. According to Google
    > > analytics,
    > >     >> there is almost no increase in the number of paper views for the
    > > document
    > >     >> site mxnet.io.
    > >     >>>
    > >     >>>
    > >     >>>
    > >     >>>
    > >     >>> The number of paper views is an important metric for the
    > adoption,
    > > I
    > >     >> would like to take actions to improve this number. It includes
    > > improving
    > >     >> the website so that users can get information easier. However,
the
    > > current
    > >     >> website displays the last stable version instead of the master
    > > branch. Then
    > >     >> the effect of a modification, namely the user behaviors, may need
    > a
    > > few
    > >     >> months to observe, which is definitely not effective.
    > >     >>>
    > >     >>> One aggressive idea is just showing the master branch in default
    > > during
    > >     >> the website improvement period (may take 3 months). Another way
is
    > >     >> releasing more frequently, e.g a new release per 2 weeks.
    > >     >>>
    > >     >>> What's your thoughts?
    > >     >>>
    > >     >>> Best,
    > >     >>> Mu
    > >     >>
    > >
    > >
    > >
    > >
    >
    

Mime
View raw message