A bit more background into this:
While tuning a model using LSTM and convolutions we find that using
hybridize with static_alloc and static_shape is 15% slower in the
latest revision vs in version 1.4.1 in which using hybridize with
static_alloc and static_shape is 10% faster than without.
Thanks for the good discussion.
I actually wasn't referring particularly to our conversations in
github with respect of the refactors, but it's nice from you to bring
them up. And it's ok to disagree in small things, hopefully we can
align in the big ones.
I understand that for TVM you might
To put a recent specific example and focus the discussion (there are
many as there are attributes), the shapes in the graph are a vector of
Shape set as an attribute using dmlc::any so this makes it very
difficult to debug the shapes when you have a graph object. I would
have it as a typed
On 2019/06/11 17:36:09, Pedro Larroy wrote:
> A bit more background into this:
>
> While tuning a model using LSTM and convolutions we find that using
> hybridize with static_alloc and static_shape is 15% slower in the
> latest revision vs in version 1.4.1 in which using hybridize with
>
The stack trace doesn't seem to come from MXNet, do you have more info?
On Tue, Jun 11, 2019 at 11:46 AM Zhi Zhang wrote:
>
>
>
> On 2019/06/11 17:36:09, Pedro Larroy wrote:
> > A bit more background into this:
> >
> > While tuning a model using LSTM and convolutions we find that using
> >
-1. Built from source, import mxnet in python cause Segfault.
back trace:
Thread 1 "python3" received signal SIGSEGV, Segmentation fault.
0x7fff3e8a9f20 in ?? ()
(gdb) bt
#0 0x7fff3e8a9f20 in ?? ()
#1 0x7fffebbf440c in ReadConfigFile(Configuration&,
std::__cxx11::basic_string,
> Re that particular case.
>
> The shape of vector will be typed after being fetched and won’t affect the
> general effort for programming. Getting the shape vector out contains
> around one line of code.
>
> The str to any map is defined to enable future compatibility of the
> general set of
Another data point. While working with a contributor, he/she is asking
to get access to the graph and values of the NDArray (me too) to be
able to reason more effectively about an enhacements to the operators:
https://github.com/apache/incubator-mxnet/pull/15120
I think gathering in the wiki
Hi guys,
Thanks for the updates. Currently, we are able to confirm Lin's issue with
Horovod, and there is a fix pending. [1]
Will update later today to see if we need to cancel this vote for the fix.
As for the hybridize with static alloc performance regression. IMO it does
not need to be a
-1
There's an autogenerated file that doesn't get cleaned up in the
scala-package folder when you run make clean. This causes the scaladoc
step to fail. I'm putting in workaround messaging in the error message
and that'll go into master, but if anyone wants to specifically run
the scaladocs for
Correction, I wanted to say:
1.5 is 33% faster than 1.4.1 when using hybridize without static_alloc
and static_shape.
We are claiming that static_alloc should improve speed and in this
case it makes it worse. Is that a blocker for the release?
Pedro.
On Tue, Jun 11, 2019 at 10:36 AM Pedro
On 2019/06/11 18:53:56, Pedro Larroy wrote:
> The stack trace doesn't seem to come from MXNet, do you have more info?
>
> On Tue, Jun 11, 2019 at 11:46 AM Zhi Zhang wrote:
> >
> >
> >
> > On 2019/06/11 17:36:09, Pedro Larroy wrote:
> > > A bit more background into this:
> > >
> > > While
We have thought very carefully when introducing type-erasures, including
considering the concerns you raised, and never-the-less have made the
decision
that resulted in the current design, which strikes the balance of
type-erasure and typing.
The original intention of the current design is to
Tested with CPU, 2.6x slower. comparing master vs 1.4.1.
Looks like a general regression.
On Tue, Jun 11, 2019 at 2:31 PM Lai Wei wrote:
>
> Hi guys,
>
> Thanks for the updates. Currently, we are able to confirm Lin's issue with
> Horovod, and there is a fix pending. [1]
> Will update later
14 matches
Mail list logo
