.
Please remember that there can be multiple use-cases and user preferences
to build MXNet.
Bhavin Thaker.
On Tue, Aug 28, 2018 at 10:29 AM Afrooze, Sina wrote:
> +1 on fully automated scripts being more confusing than helpful. It's
> difficult to debug any issues when the entire instru
017
# AVX_to_SSE_instances/100instructions: 0.1687
# SSE_to_AVX_instances/100instructions: 0.1687
Bhavin Thaker.
On Mon, Aug 13, 2018 at 7:00 AM Pedro Larroy
wrote:
> Hi
>
> I think we should explicitly define march to be x86-64 (which is the
> default in Linux) and documented here:
> h
reproduce the
crash and then use Valgrind to get to the culprit portion of the code
quickly.
Bhavin Thaker.
On Thu, May 3, 2018 at 6:49 AM Pedro Larroy
wrote:
> It's very difficult to reproduce, non-deterministic. We were also running
> without signal handlers in CI so there are no s
Have you tried running with valgrind to get some clues on the root-cause?
Bhavin Thaker.
On Wed, May 2, 2018 at 8:55 PM Da Zheng wrote:
> It might also be possible that this isn't an MKLDNN bug.
> I just saw a similar memory error without MKLDNN build.
>
> http://jenkins.mxnet
+1 to the suggestion of testing CUDA8 in few nightly instances and using
CUDA9 for most instances in CI.
Bhavin Thaker.
On Fri, Mar 16, 2018 at 12:37 PM Naveen Swamy wrote:
> I think its best to add support for CUDA 9.0 while retaining existing
> support for CUDA 8, code might regress wh
Done.
On Mon, Mar 12, 2018 at 11:27 AM Chris Olivier
wrote:
> Can someone please give write me access to the MXNet wiki?
>
Multi-part upload with finalization seems like a good approach for this
problem.
Bhavin Thaker.
On Wed, Mar 7, 2018 at 7:45 AM Naveen Swamy wrote:
> Rahul,
> IMO It is not Ok to write to a local file before streaming, you have to
> consider security implications such as:
> 1) wil
ote that this is also efficient because local memory
or disk I/O (assuming SSD or high-grade rotating hard disk) will be faster
than multiple S3 put calls over long-distance network.
Bhavin Thaker.
On Tue, Mar 6, 2018 at 10:34 PM Rahul Huilgol
wrote:
> Hi Chris,
>
> S3 doesn't sup
.
In short, your proposal #3 is good and ideal but seems less practical based
on my observations so far.
Bhavin Thaker.
On Thu, Jan 18, 2018 at 12:58 AM Marco de Abreu <
marco.g.ab...@googlemail.com> wrote:
> Hi Sheng,
>
> very good suggestions! How long do we plan to suppo
+1 to having committers act as role models for descriptive PR title and
descriptions.
I am sorry to say this explicitly because I found a few committers to have
no descriptions in their PRs with PR titles that seemed vague to me as a
contributor of Apache MXNet.
Bhavin Thaker.
On Sun, Jan 14
this: https://pypi.python.org/pypi/flaky in
another email thread. May be we could have a higher rerun count as part of
the nightly test to have better test automation stability.
Bhavin Thaker.
On Sun, Jan 14, 2018 at 12:21 PM, Sheng Zha wrote:
> Hi Bhavin,
>
> Thanks for sharing your
documented here (prepared after the MXNet 1.0 Release):
https://cwiki.apache.org/confluence/display/MXNET/Release+Retrospective+for+Apache+MXNet+%28incubating%29+1.0.0
Any comments or suggestion on the above Release retrospective are most
welcome.
Bhavin Thaker.
On Mon, Nov 13, 2017 at 3:03 PM
instead of doing it
during the upcoming release.
Bhavin Thaker.
On Sat, Jan 13, 2018 at 2:20 PM, Marco de Abreu <
marco.g.ab...@googlemail.com> wrote:
> Hello Sheng,
>
> thanks a lot for leading this task!
>
> +1 for both points. Additionally, I'd propose to add the
MXNet release 1.0.1.
Thanks,
Bhavin Thaker.
On Fri, Jan 12, 2018 at 5:43 PM Chris Olivier wrote:
> i’d be +1 if CODEOWNERS file has a big note at the top saying basically
> it’s just for watching code changes that you’d like to know about (to
> review or just to follow) and that anyon
Ok, Marco. Do all the permitted datasets reside in S3 or is this a todo
item?
Bhavin Thaker.
On Fri, Jan 12, 2018 at 7:28 AM Marco de Abreu
wrote:
> It would make sense, but the license does not permit redistribution of the
> GroupLens-Movie dataset. We already have a few datasets in
Does it make sense to cache the datasets into a (reliable) S3 bucket so the
tests to run reliably?
Does the dataset licensing allow downloading the dataset?
Bhavin Thaker.
On Fri, Jan 12, 2018 at 5:52 AM kellen sunderland <
kellen.sunderl...@gmail.com> wrote:
> Hey all, since this ser
Would it make sense to have a developer best practices section on the
Apache wiki where such guidance can be documented for future reference?
Bhavin Thaker.
On Thu, Jan 11, 2018 at 9:56 AM Anirudh wrote:
> Hi,
>
>
> I have been thinking about exception handling specifically in
to the open-source Apache MXNet.
Bhavin Thaker.
On Sat, Jan 6, 2018 at 12:30 PM Naveen Swamy wrote:
> +1 to that. I think we don't have to run CUDA 8 on every PR.
>
> On Sat, Jan 6, 2018 at 12:26 PM, Marco de Abreu <
> marco.g.ab...@googlemail.com> wrote:
>
>
Hi Marco,
Here are the Years in which the GPU architectures were introduced:
- Tesla: 2008;
- Fermi: 2010;
- Kepler: 2012;
- Maxwell: 2014;
- Pascal:2016;
- Volta: 2017;
I see no need to support the 7+ year old Fermi architecture for fast-moving
Apache MXNet.
Bhavin Thaker
that the MXNet
CI run tests only with latest CUDA9 version and NOT CUDA8.
I am eager to hear alternate viewpoints/corrections from folks other than
Kellen and me.
Bhavin Thaker.
On Sat, Jan 6, 2018 at 8:24 AM kellen sunderland <
kellen.sunderl...@gmail.com> wrote:
> Thanks for the though
/incubator-mxnet/issues/8805
Bhavin Thaker.
On Sat, Jan 6, 2018 at 3:30 AM kellen sunderland <
kellen.sunderl...@gmail.com> wrote:
> Hello all, I'd like to propose that we nail down exactly which versions of
> CUDA we're supporting. We can then ensure that we've got goo
.
Thanks,
Bhavin Thaker.
On Wed, Jan 3, 2018 at 10:39 AM, Hen wrote:
> Thanks Bhavin. Let's definitely get that as an active conversation.
>
> In general I'm okay with the report. I'm having login pains with the wiki
> (yay for non-ldap service), but chatted with Suneel on I
questions in the one question above, I am
looking for more ideas or thoughts.
Cheers,
Bhavin Thaker.
various MXNet talks at
conferences, universities, etc. If anybody has suggestions on how to
increase the contributions to MXNet, please share your ideas.
Bhavin Thaker.
On Tue, Jan 2, 2018 at 9:15 PM Hen wrote:
> Few items that flagged for me in the report:
>
> 1/ "We are work
://cwiki.apache.org/confluence/display/MXNET/Development+Process
Wishing all of you a very happy new year 2018!
Regards,
Bhavin Thaker.
On Fri, Dec 22, 2017 at 12:36 AM, kellen sunderland <
kellen.sunderl...@gmail.com> wrote:
> Sheng’s comments are correct. GitHub will save review com
how-to link here:
https://mxnet.incubator.apache.org/community/contribute.html
Please vote and provide your feedback.
Regards,
Bhavin Thaker.
Marco: we were thinking on the lines of following Semantic versioning but
have not made the proposal to dev@ yet — plan to do that in a week or so.
Bhavin Thaker.
On Tue, Dec 12, 2017 at 4:04 PM Marco de Abreu
wrote:
> Are we following semantic versioning https://semver.org/ for MX
Hi Marco,
Thanks for your work on the CI.
Is it possible/ok to share the docs via googledocs link until you get write
permissions for the apache wiki page?
Bhavin Thaker.
On Tue, Dec 5, 2017 at 4:06 PM, Marco de Abreu wrote:
> Hello MXNet community,
>
> as discussed i
Hi Sergio,
Thanks for guidance.
The binary file has been removed now.
See: https://github.com/dmlc/tvm/tree/master/apps/android_rpc/gradle/wrapper
We are testing the changes currently and expect to submit a new RC by end
of today.
Regards,
Bhavin Thaker.
On Thu, Nov 30, 2017 at 9:40 AM Sergio
T use GPU
context on a CPU-only instance.
Bhavin Thaker.
On Sun, Nov 26, 2017 at 7:01 PM, Steffen Rochel
wrote:
> +1 (non-binding)
> verified building from source and tested on macOS Sierra
>
> On Sun, Nov 26, 2017 at 10:32 AM Chris Olivier
> wrote:
>
> > +1 (bindi
.
https://stackoverflow.com/questions/6364314/why-is-a-pthread-mutex-considered-slower-than-a-futex
Bhavin Thaker.
On Thu, Nov 23, 2017 at 10:00 PM kellen sunderland <
kellen.sunderl...@gmail.com> wrote:
> I think Haitao is right given some stacks we've recently looked at.
Do we have performance/profiling measurement evidence to prove/disprove the
claims of the impact of mutex bottleneck in MXNet?
Bhavin Thaker.
On Thu, Nov 23, 2017 at 11:03 AM Tianqi Chen
wrote:
> If they are not bottleneck. Then resorting to standard library solution is
> always pre
+1 for better [1] PR Titles. As suggested by Madan and use by Spark, the
current PR template seems to be ignored by folks and so we may want to
simplify it to:
Q1. What changes were proposed in this pull request?
Q2. How was this patch tested?
+1 to either [2] Jira OR [3] PR labels.
Bhavin
+1 for [1] Jenkins (A setup separated from Apache Jenkins) - with various
plugins.
Bhavin Thaker.
On Fri, Nov 10, 2017 at 11:39 AM, Madan Jampani
wrote:
> +1 for (1)
>
> On Thu, Nov 9, 2017 at 4:41 PM, Meghna Baijal
> wrote:
>
> > Hi All,
> > A need has been identif
I agree about your point on correctness -- do you know of any known
correctness issues with Ninja?
These build times seem to be NOT with GPU builds and distributed kvstore
enabled -- could you please confirm? nvcc builds take a significant time.
Bhavin Thaker.
On Thu, Nov 2, 2017 at 8:45 AM
-build.org/manual.html
Thanks,
Bhavin Thaker.
On Thu, Nov 2, 2017 at 7:57 AM, Pedro Larroy
wrote:
> Hi
>
> I would like to update the MXNet build instructions.
>
> In particular I was thinking that it would be a good idea to update
> the instructions to use CMake + N
about, what is the expected SLA for a response from them? When users deploy
MXNet in production environments, they will expect some form of SLA for
support and a patch release.
Regards,
Bhavin Thaker.
On Wed, Nov 1, 2017 at 8:20 AM, Pedro Larroy
wrote:
> +1 That would be great.
>
>
failures for PRs.
There are other places (nightly) to do extensive tests.
PR-unit-tests are sanity tests and must be quick, reliable and consistent
for every PR.
Bhavin Thaker.
On Mon, Oct 16, 2017 at 8:51 AM Pedro Larroy
wrote:
> That's not true. random() and similar functions are based o
+1 to disable old slack, but keep old messages viewable.
Bhavin Thaker.
On Mon, Oct 16, 2017 at 8:23 AM Steffen Rochel
wrote:
> As we have now migrated the majority of people from apache-mxnet.slack.com
> to the-asf.slack.com I would like to suggest to shut down
> apache-mxnet.
MXNet (Incubating).
Regards,
Bhavin Thaker.
On Mon, Oct 16, 2017 at 5:56 AM Pedro Larroy
wrote:
> Hi
>
> Some of the unit tests are extremely costly in terms of memory and compute.
>
> As an example in the gluon tests we are loading all the datasets.
>
> test_gluon_data.tes
Hi John,
I went through the mentioned section and scanned the entire page. It is
very helpful. Thanks!
Regards,
Bhavin Thaker.
On Fri, Oct 6, 2017 at 4:32 PM, John D. Ament wrote:
> Bhavin,
>
> You may want to review http://www.apache.org/dev/infra-contact#request-
> checklist .
learn from the mistakes of other Apache
projects and also contribute to that guidelines list.
Later, we can propose our guidelines page to contribute back to the *main
Apache guidelines page for incubating projects*.
Bhavin Thaker.
On Thu, Oct 5, 2017 at 10:24 PM, Sebastian wrote:
> I'm
Hi Seb: please use a different email thread for new topics of discussion.
Hi Jun: I think Seb may be referring to Volta V100 support in MXNet and NOT
P4/P40 inference accelerators.
Corrections/clarifications welcome.
Bhavin Thaker.
On Mon, Oct 2, 2017 at 8:22 PM Jun Wu wrote:
> Thanks
+1
On Fri, Sep 15, 2017 at 4:45 PM Naveen Swamy wrote:
> + 1
>
> On Fri, Sep 15, 2017 at 4:32 PM, Vikram Madan
> wrote:
>
> > +1
> >
> > On Fri, Sep 15, 2017 at 3:40 PM, Lupesko, Hagay
> wrote:
> >
> > > +1
> > >
> > > I think that a dedicated forum would be nice. PyTorch have a nice one:
> >
-review and only
after sufficient amount of sanity build-tests have passed.
Let the machines work harder for humans and not the other way around.
Bhavin Thaker.
On Tue, Sep 12, 2017 at 12:20 PM Chris Olivier
wrote:
> The majority of these iterations is to trigger a build on the broken CI
, Sep 11, 2017 at 5:30 PM, Bhavin Thaker
> wrote:
>
> > +1 : Chris has an important question.
> >
> > How do we ensure that a feature/task for a particular Apache project is
> not
> > worked on by multiple folks at the same time?
> >
> > Is t
working on
what.
Bhavin Thaker.
On Mon, Sep 11, 2017 at 5:06 AM Joern Kottmann wrote:
> A good way to coordinate is to use the dev list and/or state on the
> issues that you are interested in it and are working on it.
>
> We use Jira for years at Apache OpenNLP for all the issues we dea
high bar. This is difficult to
quantify but is a quality that is expected to be demonstrated in various
conversations of the person being nominated to be the committer.
Bhavin Thaker.
On Wed, Aug 9, 2017 at 11:03 PM, Sebastian wrote:
>
> Another thing might be that having too many committers
Comments:
[ ](mxnet) Markus Weimer
Comments:
[ ](mxnet) Henri Yandell
Comments:
--snip—
Thanks,
Bhavin Thaker.
49 matches
Mail list logo