Re: Embedded World 2019 Robotics Demo

[Joshua Z. Zhang]

Played with the demo closely today, and it looks awesome with the real robot 
arms!

-Zhi

> On Feb 27, 2019, at 11:03 PM, Thomas DELTEIL  
> wrote:
> 
> Just tweeted two videos about it:
> https://twitter.com/thdelteil/status/1101012275991764993
> https://twitter.com/thdelteil/status/1101012034051727360
> We will write a blog post or two about the project when Anton, Pavel and
> the others are back from the conference to give more details :)
> 
> To summarize and as can be inferred from the videos, two robotic arms are
> moving according to the position of the wrists of the player. The pose of
> the player is detected using a pose estimation neural network based on the
> implementation of Simple Baselines for Pose Estimation paper
> that
> is going to be released in the next version of gluon-cv, shout out to Zhi
> and He Tong for providing the training code. Another classifier detects the
> hand as open or closed and that decides the position of the grips. The goal
> of the game is to take a book from one side and drop it in the bucket on
> the other side by passing it between the two arms. The robots are from
> Nyrio, a French startup.
> 
> To answer your question Isabel, this project was a joint cooperation
> between a few MXNet team members at AWS, including Anton, Pavel and myself
> and some employees at the QT (the C++ library) company, in their industrial
> automation department.
> 
> All the best,
> 
> Thomas Delteil
> 
> Le mer. 27 févr. 2019 à 22:27, Junru Shao  a
> écrit :
> 
>> Hi Anton,
>> 
>> Would love to know more about this great demo! Will you guys share video
>> clips?
>> 
>> Cheers,
>> Junru
>> 
>> On Wed, Feb 27, 2019 at 8:11 PM Isabel Drost-Fromm 
>> wrote:
>> 
>>> Hi,
>>> 
>>> Sounds interesting. Can you share more on what you are showing and what
>>> role mxnet plays?
>>> 
>>> Also, who's that "we" behind "our booth"?
>>> 
>>> 
>>> Have fun in Nürnberg,
>>> Isabel
>>> 
>>> Am 27. Februar 2019 11:38:57 MEZ schrieb Anton Chernov <
>>> mecher...@gmail.com>:
 Dear MXNet Community,
 
 If you happens to be at the Embedded World exhibition in Nürnberg drop
 by
 our booth at the Qt stand in hall 4 to see a MXNet robotics demo.
 
 Looking forward to see you!
 
 Best
 Anton
>>> 
>>> --
>>> Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.
>>> 
>> 

Re: Embedded World 2019 Robotics Demo

[Thomas DELTEIL]

Just tweeted two videos about it:
https://twitter.com/thdelteil/status/1101012275991764993
https://twitter.com/thdelteil/status/1101012034051727360
We will write a blog post or two about the project when Anton, Pavel and
the others are back from the conference to give more details :)

To summarize and as can be inferred from the videos, two robotic arms are
moving according to the position of the wrists of the player. The pose of
the player is detected using a pose estimation neural network based on the
implementation of Simple Baselines for Pose Estimation paper
that
is going to be released in the next version of gluon-cv, shout out to Zhi
and He Tong for providing the training code. Another classifier detects the
hand as open or closed and that decides the position of the grips. The goal
of the game is to take a book from one side and drop it in the bucket on
the other side by passing it between the two arms. The robots are from
Nyrio, a French startup.

To answer your question Isabel, this project was a joint cooperation
between a few MXNet team members at AWS, including Anton, Pavel and myself
and some employees at the QT (the C++ library) company, in their industrial
automation department.

All the best,

Thomas Delteil

Le mer. 27 févr. 2019 à 22:27, Junru Shao  a
écrit :

> Hi Anton,
>
> Would love to know more about this great demo! Will you guys share video
> clips?
>
> Cheers,
> Junru
>
> On Wed, Feb 27, 2019 at 8:11 PM Isabel Drost-Fromm 
> wrote:
>
> > Hi,
> >
> > Sounds interesting. Can you share more on what you are showing and what
> > role mxnet plays?
> >
> > Also, who's that "we" behind "our booth"?
> >
> >
> > Have fun in Nürnberg,
> > Isabel
> >
> > Am 27. Februar 2019 11:38:57 MEZ schrieb Anton Chernov <
> > mecher...@gmail.com>:
> > >Dear MXNet Community,
> > >
> > >If you happens to be at the Embedded World exhibition in Nürnberg drop
> > >by
> > >our booth at the Qt stand in hall 4 to see a MXNet robotics demo.
> > >
> > >Looking forward to see you!
> > >
> > >Best
> > >Anton
> >
> > --
> > Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.
> >
>

Re: [DISCUSS] Process to remove deprecated operators

[Lin Yuan]

Agreed. When we deprecate an operator, we should add in the log message
something like "This operator X is deprecate and will be removed in the
next release. Please use operator Y instead."

Lin

On Wed, Feb 27, 2019 at 10:23 PM Junru Shao  wrote:

> Hi Lin,
>
> I would love to share some immature ideas about deprecating operators. Not
> only adopting semantic versioning, but also should we provide enough
> informative error message for customers to understand how to replace
> deprecated operators with new ones.
>
> Thanks,
> Junru
>
> On Wed, Feb 27, 2019 at 9:30 PM Lin Yuan  wrote:
>
> > Sheng,
> >
> > Thanks for your quick response.
> > If that's the case, we will wait till 2.0 release to remove the
> deprecated
> > operators from code.
> >
> > Best,
> > Lin
> >
> > On Wed, Feb 27, 2019 at 9:06 PM Sheng Zha  wrote:
> >
> > > MXNet follows semantic versioning so we will be able to delete them in
> > the
> > > next major release.
> > >
> > > -sz
> > >
> > > On Wed, Feb 27, 2019 at 8:53 PM Lin Yuan  wrote:
> > >
> > > > Dear Community,
> > > >
> > > > In MXNet there are many legacy operators such as this
> > > > <
> > > >
> > >
> >
> http://mxnet.incubator.apache.org/versions/master/api/python/symbol/symbol.html?highlight=convolution_v1#mxnet.symbol.Convolution_v1
> > > > >
> > > > that has been marked DEPRECATE for several releases. However, these
> > > > operators still exist in our code. This caused a few problems:
> > > >
> > > > 1) Make the codebase bloated and reduce readability
> > > > 2) Increase unnecessary maintanence effort
> > > > 3) Bug prone as some people will look up these legacy code as example
> > > > 4) Cause confusion to end users and make documentation page lengthy
> > > >
> > > > I would like to propose the following process (if there is no
> existing
> > > one)
> > > > to remove deprecate operators from our code base.
> > > >
> > > > 1. Documnent the deprecate operators/environment variables in the
> > release
> > > > note as well as man pages.
> > > > 2. Limit the life cycle of deprecate operators/argument to two minor
> > > > release. For example, if one operator is marked deprecate in 1.4
> > release,
> > > > it will be removed in 1.6 release.
> > > > 3. If there is some concern raised from customers during 1.4 and 1.5
> > > > release, we can convert the deprecated operator back to current and
> it
> > > will
> > > > be treated as new operator.
> > > > 4. PRs that remove deprecate operators should contain [Cleanup] in
> > title.
> > > >
> > > > Any comment is appreciated.
> > > >
> > > > Lin
> > > >
> > >
> >
>

Re: Embedded World 2019 Robotics Demo

[Junru Shao]

Hi Anton,

Would love to know more about this great demo! Will you guys share video
clips?

Cheers,
Junru

On Wed, Feb 27, 2019 at 8:11 PM Isabel Drost-Fromm 
wrote:

> Hi,
>
> Sounds interesting. Can you share more on what you are showing and what
> role mxnet plays?
>
> Also, who's that "we" behind "our booth"?
>
>
> Have fun in Nürnberg,
> Isabel
>
> Am 27. Februar 2019 11:38:57 MEZ schrieb Anton Chernov <
> mecher...@gmail.com>:
> >Dear MXNet Community,
> >
> >If you happens to be at the Embedded World exhibition in Nürnberg drop
> >by
> >our booth at the Qt stand in hall 4 to see a MXNet robotics demo.
> >
> >Looking forward to see you!
> >
> >Best
> >Anton
>
> --
> Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.
>

Re: [DISCUSS] Process to remove deprecated operators

[Junru Shao]

Hi Lin,

I would love to share some immature ideas about deprecating operators. Not
only adopting semantic versioning, but also should we provide enough
informative error message for customers to understand how to replace
deprecated operators with new ones.

Thanks,
Junru

On Wed, Feb 27, 2019 at 9:30 PM Lin Yuan  wrote:

> Sheng,
>
> Thanks for your quick response.
> If that's the case, we will wait till 2.0 release to remove the deprecated
> operators from code.
>
> Best,
> Lin
>
> On Wed, Feb 27, 2019 at 9:06 PM Sheng Zha  wrote:
>
> > MXNet follows semantic versioning so we will be able to delete them in
> the
> > next major release.
> >
> > -sz
> >
> > On Wed, Feb 27, 2019 at 8:53 PM Lin Yuan  wrote:
> >
> > > Dear Community,
> > >
> > > In MXNet there are many legacy operators such as this
> > > <
> > >
> >
> http://mxnet.incubator.apache.org/versions/master/api/python/symbol/symbol.html?highlight=convolution_v1#mxnet.symbol.Convolution_v1
> > > >
> > > that has been marked DEPRECATE for several releases. However, these
> > > operators still exist in our code. This caused a few problems:
> > >
> > > 1) Make the codebase bloated and reduce readability
> > > 2) Increase unnecessary maintanence effort
> > > 3) Bug prone as some people will look up these legacy code as example
> > > 4) Cause confusion to end users and make documentation page lengthy
> > >
> > > I would like to propose the following process (if there is no existing
> > one)
> > > to remove deprecate operators from our code base.
> > >
> > > 1. Documnent the deprecate operators/environment variables in the
> release
> > > note as well as man pages.
> > > 2. Limit the life cycle of deprecate operators/argument to two minor
> > > release. For example, if one operator is marked deprecate in 1.4
> release,
> > > it will be removed in 1.6 release.
> > > 3. If there is some concern raised from customers during 1.4 and 1.5
> > > release, we can convert the deprecated operator back to current and it
> > will
> > > be treated as new operator.
> > > 4. PRs that remove deprecate operators should contain [Cleanup] in
> title.
> > >
> > > Any comment is appreciated.
> > >
> > > Lin
> > >
> >
>

Re: [DISCUSS] Process to remove deprecated operators

[Lin Yuan]

Sheng,

Thanks for your quick response.
If that's the case, we will wait till 2.0 release to remove the deprecated
operators from code.

Best,
Lin

On Wed, Feb 27, 2019 at 9:06 PM Sheng Zha  wrote:

> MXNet follows semantic versioning so we will be able to delete them in the
> next major release.
>
> -sz
>
> On Wed, Feb 27, 2019 at 8:53 PM Lin Yuan  wrote:
>
> > Dear Community,
> >
> > In MXNet there are many legacy operators such as this
> > <
> >
> http://mxnet.incubator.apache.org/versions/master/api/python/symbol/symbol.html?highlight=convolution_v1#mxnet.symbol.Convolution_v1
> > >
> > that has been marked DEPRECATE for several releases. However, these
> > operators still exist in our code. This caused a few problems:
> >
> > 1) Make the codebase bloated and reduce readability
> > 2) Increase unnecessary maintanence effort
> > 3) Bug prone as some people will look up these legacy code as example
> > 4) Cause confusion to end users and make documentation page lengthy
> >
> > I would like to propose the following process (if there is no existing
> one)
> > to remove deprecate operators from our code base.
> >
> > 1. Documnent the deprecate operators/environment variables in the release
> > note as well as man pages.
> > 2. Limit the life cycle of deprecate operators/argument to two minor
> > release. For example, if one operator is marked deprecate in 1.4 release,
> > it will be removed in 1.6 release.
> > 3. If there is some concern raised from customers during 1.4 and 1.5
> > release, we can convert the deprecated operator back to current and it
> will
> > be treated as new operator.
> > 4. PRs that remove deprecate operators should contain [Cleanup] in title.
> >
> > Any comment is appreciated.
> >
> > Lin
> >
>

Re: [DISCUSS] Process to remove deprecated operators

[Sheng Zha]

MXNet follows semantic versioning so we will be able to delete them in the
next major release.

-sz

On Wed, Feb 27, 2019 at 8:53 PM Lin Yuan  wrote:

> Dear Community,
>
> In MXNet there are many legacy operators such as this
> <
> http://mxnet.incubator.apache.org/versions/master/api/python/symbol/symbol.html?highlight=convolution_v1#mxnet.symbol.Convolution_v1
> >
> that has been marked DEPRECATE for several releases. However, these
> operators still exist in our code. This caused a few problems:
>
> 1) Make the codebase bloated and reduce readability
> 2) Increase unnecessary maintanence effort
> 3) Bug prone as some people will look up these legacy code as example
> 4) Cause confusion to end users and make documentation page lengthy
>
> I would like to propose the following process (if there is no existing one)
> to remove deprecate operators from our code base.
>
> 1. Documnent the deprecate operators/environment variables in the release
> note as well as man pages.
> 2. Limit the life cycle of deprecate operators/argument to two minor
> release. For example, if one operator is marked deprecate in 1.4 release,
> it will be removed in 1.6 release.
> 3. If there is some concern raised from customers during 1.4 and 1.5
> release, we can convert the deprecated operator back to current and it will
> be treated as new operator.
> 4. PRs that remove deprecate operators should contain [Cleanup] in title.
>
> Any comment is appreciated.
>
> Lin
>

[DISCUSS] Process to remove deprecated operators

[Lin Yuan]

Dear Community,

In MXNet there are many legacy operators such as this

that has been marked DEPRECATE for several releases. However, these
operators still exist in our code. This caused a few problems:

1) Make the codebase bloated and reduce readability
2) Increase unnecessary maintanence effort
3) Bug prone as some people will look up these legacy code as example
4) Cause confusion to end users and make documentation page lengthy

I would like to propose the following process (if there is no existing one)
to remove deprecate operators from our code base.

1. Documnent the deprecate operators/environment variables in the release
note as well as man pages.
2. Limit the life cycle of deprecate operators/argument to two minor
release. For example, if one operator is marked deprecate in 1.4 release,
it will be removed in 1.6 release.
3. If there is some concern raised from customers during 1.4 and 1.5
release, we can convert the deprecated operator back to current and it will
be treated as new operator.
4. PRs that remove deprecate operators should contain [Cleanup] in title.

Any comment is appreciated.

Lin

Re: Embedded World 2019 Robotics Demo

[Isabel Drost-Fromm]

Hi,

Sounds interesting. Can you share more on what you are showing and what role 
mxnet plays?

Also, who's that "we" behind "our booth"?


Have fun in Nürnberg,
Isabel

Am 27. Februar 2019 11:38:57 MEZ schrieb Anton Chernov :
>Dear MXNet Community,
>
>If you happens to be at the Embedded World exhibition in Nürnberg drop
>by
>our booth at the Qt stand in hall 4 to see a MXNet robotics demo.
>
>Looking forward to see you!
>
>Best
>Anton

-- 
Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.

Re: CI woes pt.2

[Junru Shao]

Gotcha! Thanks!

On Wed, Feb 27, 2019 at 1:44 PM Gavin M. Bell 
wrote:

> Per, Chance and team - Nice work!
>
>
> On Wed, Feb 27, 2019 at 10:43 PM Per da Silva 
> wrote:
>
> > Hey Junru,
> >
> >
> > Please rebase and try again. Although, I'm finding this a bit
> > strangelet me know how it goes and tomorrow morning I'll follow it
> up!
> >
> > Cheers,
> >
> > Per
> >
> > On Wed., 27 Feb. 2019, 10:40 pm Junru Shao, 
> > wrote:
> >
> > > Hey Per,
> > >
> > > CI for my PR still fails (
> > >
> > >
> >
> http://jenkins.mxnet-ci.amazon-ml.com/blue/organizations/jenkins/mxnet-validation%2Fwindows-gpu/detail/PR-14217/13/pipeline
> > > ),
> > > saying "ERROR: Queue task was cancelled". Should I rebase?
> > >
> > > Thanks,
> > > Junru
> > >
> > > On Wed, Feb 27, 2019 at 10:36 AM Junru Shao 
> > > wrote:
> > >
> > > > Finally! Thanks Per for the hard work!
> > > >
> > > > On Wed, Feb 27, 2019 at 6:35 AM Per da Silva 
> > > wrote:
> > > >
> > > >> Hi everyone,
> > > >>
> > > >> The PR temporarily disabling windows tests has been merged, so the
> > > windows
> > > >> checks shouldn't block progress for the time being. Please retrigger
> > any
> > > >> PRs you have open. We are still working hard on solving the windows
> > > >> instance issues. Once that is sorted out, we'll come back with more
> > > news.
> > > >>
> > > >> Cheers,
> > > >>
> > > >> Per
> > > >>
> > > >
> > >
> >
>
>
> --
> Sincerely,
> Gavin M. Bell
>
>  "Never mistake a clear view for a short distance."
>   -Paul Saffo
>

Re: CI woes pt.2

[Gavin M. Bell]

Per, Chance and team - Nice work!


On Wed, Feb 27, 2019 at 10:43 PM Per da Silva  wrote:

> Hey Junru,
>
>
> Please rebase and try again. Although, I'm finding this a bit
> strangelet me know how it goes and tomorrow morning I'll follow it up!
>
> Cheers,
>
> Per
>
> On Wed., 27 Feb. 2019, 10:40 pm Junru Shao, 
> wrote:
>
> > Hey Per,
> >
> > CI for my PR still fails (
> >
> >
> http://jenkins.mxnet-ci.amazon-ml.com/blue/organizations/jenkins/mxnet-validation%2Fwindows-gpu/detail/PR-14217/13/pipeline
> > ),
> > saying "ERROR: Queue task was cancelled". Should I rebase?
> >
> > Thanks,
> > Junru
> >
> > On Wed, Feb 27, 2019 at 10:36 AM Junru Shao 
> > wrote:
> >
> > > Finally! Thanks Per for the hard work!
> > >
> > > On Wed, Feb 27, 2019 at 6:35 AM Per da Silva 
> > wrote:
> > >
> > >> Hi everyone,
> > >>
> > >> The PR temporarily disabling windows tests has been merged, so the
> > windows
> > >> checks shouldn't block progress for the time being. Please retrigger
> any
> > >> PRs you have open. We are still working hard on solving the windows
> > >> instance issues. Once that is sorted out, we'll come back with more
> > news.
> > >>
> > >> Cheers,
> > >>
> > >> Per
> > >>
> > >
> >
>


-- 
Sincerely,
Gavin M. Bell

 "Never mistake a clear view for a short distance."
  -Paul Saffo

Re: CI woes pt.2

[Per da Silva]

Hey Junru,


Please rebase and try again. Although, I'm finding this a bit
strangelet me know how it goes and tomorrow morning I'll follow it up!

Cheers,

Per

On Wed., 27 Feb. 2019, 10:40 pm Junru Shao,  wrote:

> Hey Per,
>
> CI for my PR still fails (
>
> http://jenkins.mxnet-ci.amazon-ml.com/blue/organizations/jenkins/mxnet-validation%2Fwindows-gpu/detail/PR-14217/13/pipeline
> ),
> saying "ERROR: Queue task was cancelled". Should I rebase?
>
> Thanks,
> Junru
>
> On Wed, Feb 27, 2019 at 10:36 AM Junru Shao 
> wrote:
>
> > Finally! Thanks Per for the hard work!
> >
> > On Wed, Feb 27, 2019 at 6:35 AM Per da Silva 
> wrote:
> >
> >> Hi everyone,
> >>
> >> The PR temporarily disabling windows tests has been merged, so the
> windows
> >> checks shouldn't block progress for the time being. Please retrigger any
> >> PRs you have open. We are still working hard on solving the windows
> >> instance issues. Once that is sorted out, we'll come back with more
> news.
> >>
> >> Cheers,
> >>
> >> Per
> >>
> >
>

Re: CI woes pt.2

[Junru Shao]

Hey Per,

CI for my PR still fails (
http://jenkins.mxnet-ci.amazon-ml.com/blue/organizations/jenkins/mxnet-validation%2Fwindows-gpu/detail/PR-14217/13/pipeline),
saying "ERROR: Queue task was cancelled". Should I rebase?

Thanks,
Junru

On Wed, Feb 27, 2019 at 10:36 AM Junru Shao  wrote:

> Finally! Thanks Per for the hard work!
>
> On Wed, Feb 27, 2019 at 6:35 AM Per da Silva  wrote:
>
>> Hi everyone,
>>
>> The PR temporarily disabling windows tests has been merged, so the windows
>> checks shouldn't block progress for the time being. Please retrigger any
>> PRs you have open. We are still working hard on solving the windows
>> instance issues. Once that is sorted out, we'll come back with more news.
>>
>> Cheers,
>>
>> Per
>>
>

Re: Slack invite

[Steffen Rochel]

Welcome Arthur and thanks for your contributions.
Steffen

On Wed, Feb 27, 2019 at 1:13 PM Arthur Caillau  wrote:

> Hi there,
>
> I'm helping work on the Clojure MXNet package, could I please have an
> invite?
>
> Cheers!

Slack invite

[Arthur Caillau]

Hi there,

I'm helping work on the Clojure MXNet package, could I please have an invite?

Cheers!

Re: CI woes pt.2

[Junru Shao]

Finally! Thanks Per for the hard work!

On Wed, Feb 27, 2019 at 6:35 AM Per da Silva  wrote:

> Hi everyone,
>
> The PR temporarily disabling windows tests has been merged, so the windows
> checks shouldn't block progress for the time being. Please retrigger any
> PRs you have open. We are still working hard on solving the windows
> instance issues. Once that is sorted out, we'll come back with more news.
>
> Cheers,
>
> Per
>

CI woes pt.2

[Per da Silva]

Hi everyone,

The PR temporarily disabling windows tests has been merged, so the windows
checks shouldn't block progress for the time being. Please retrigger any
PRs you have open. We are still working hard on solving the windows
instance issues. Once that is sorted out, we'll come back with more news.

Cheers,

Per

CI woes

[Per da Silva]

Hi everyone,

The recent issues with the CI windows instances have created a massive
queue on Jenkins. This brought it down momentarily, and it's now back up. I
have created a PR that comments out the steps in the windows pipelines (
https://github.com/apache/incubator-mxnet/pull/14261). I would appreciated
if a committer out there could please merge it. This is a temporary
solution to keep Jenkins stable while we work on fixing the windows
instances.

Your help in this matter is greatly appreciated.

We'll endeavour to fix this issue ASAP and keep you in the loop.

Cheers,

Per

Embedded World 2019 Robotics Demo

[Anton Chernov]

Dear MXNet Community,

If you happens to be at the Embedded World exhibition in Nürnberg drop by
our booth at the Qt stand in hall 4 to see a MXNet robotics demo.

Looking forward to see you!

Best
Anton