cjolivier01 commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367813362
Rules are nice, but IMHO, customer adoption right now trumps everything.
Well, not EVERYHING, but it's important.
cjolivier01 commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367813362
Rules are nice, but IMHO, customer adoption right now trumps everything
This
marcoabreu commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367810206
Considering the fact that we are not making any releases for v0.11, it
doesn't really make sense to work on that branch. Or does the customer
Lorna-Liu commented on issue #9517: Check failed: err == CUDNN_STATUS_SUCCESS
(6 vs. 0) CUDNN_STATUS_ARCH_MISMATCH
URL:
https://github.com/apache/incubator-mxnet/issues/9517#issuecomment-367805855
@liangyong928
I was wondering if you have solve the problem. I just encountered exactly
Lorna-Liu commented on issue #9517: Check failed: err == CUDNN_STATUS_SUCCESS
(6 vs. 0) CUDNN_STATUS_ARCH_MISMATCH
URL:
https://github.com/apache/incubator-mxnet/issues/9517#issuecomment-367805855
@liangyong928
I was wondering if you have solve the problem. I just encountered exactly
wellner commented on issue #9656: Sparse Tensor support in scala api
URL:
https://github.com/apache/incubator-mxnet/issues/9656#issuecomment-367802177
I'd second this request for support of sparse tensors/ndarrays in Scala.
Thanks!
dabraude commented on issue #9860: [WIP] CMake NNPack support
URL: https://github.com/apache/incubator-mxnet/pull/9860#issuecomment-367794784
@piiswrong NNPack's CmakeLists.txt downloads the dependencies, I can change
it so that they are not submodules.
marcoabreu commented on issue #9820: Flaky
test_gluon_model_zoo_gpu.test_training @ Python3: MKLDNN-GPU
URL:
https://github.com/apache/incubator-mxnet/issues/9820#issuecomment-367792349
Great news, thanks for diving deep on this problem!
zheng-da commented on issue #9820: Flaky test_gluon_model_zoo_gpu.test_training
@ Python3: MKLDNN-GPU
URL:
https://github.com/apache/incubator-mxnet/issues/9820#issuecomment-367791248
After one day of testing, I think I know where the problem is now.
Basically, the current implementation
nswamy commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367788415
@piiswrong @szha
Some background:
One of the customer is using MXNet-Scala Interface of 0.11 version. They are
soon to go live with their
This is an automated email from the ASF dual-hosted git repository.
zhasheng pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git
The following commit(s) were added to refs/heads/master by this push:
new e1d8c66 Enable the reporting of
szha closed pull request #9805: Enable the reporting of cross-entropy or nll
loss value when training CNN network using the models defined by
example/image-classification
URL: https://github.com/apache/incubator-mxnet/pull/9805
This is a PR merged from a forked repository.
As GitHub
piiswrong commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367773072
We retrofit critical bugfixes to previous releases. This is a new feature.
If you do it for 0.11, then you'll need to do it for all the releases
piiswrong commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367773072
We retrofit critical bugfixes to previous releases. This is a new feature.
If you do it for 0.11, then you'll need to do it for all the releases
piiswrong commented on issue #9860: [WIP] CMake NNPack support
URL: https://github.com/apache/incubator-mxnet/pull/9860#issuecomment-367772409
This is too many submodules.
@Maratyszcza Do you have a bundled release or something like that? A single
submodule that has a snap shot to all
Maratyszcza commented on a change in pull request #9860: [WIP] CMake NNPack
support
URL: https://github.com/apache/incubator-mxnet/pull/9860#discussion_r170047405
##
File path: CMakeLists.txt
##
@@ -551,6 +552,66 @@ if(NOT EXISTS
Maratyszcza commented on issue #9860: [WIP] CMake NNPack support
URL: https://github.com/apache/incubator-mxnet/pull/9860#issuecomment-367771148
PeachPy has `pre-generated` branch, which is a mirror of `master` branch
after running `python setup.py generate`. I suggest you use the HEAD of
szha commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367746320
Yeah, that would be the right approach. Semver doesn?t have to apply for
0.x, so cherry pick should be fine.
cjolivier01 commented on a change in pull request #9770: eye operator, for
default storage type
URL: https://github.com/apache/incubator-mxnet/pull/9770#discussion_r170006130
##
File path: src/operator/tensor/init_op.h
##
@@ -428,6 +405,39 @@ void FillComputeZerosEx(const
cjolivier01 commented on issue #9744: Performance regression when
OMP_NUM_THREADS environment variable is not set
URL:
https://github.com/apache/incubator-mxnet/issues/9744#issuecomment-367727571
Can you also explain what these settings do? I know it's in a manual
somewhere, but from
cjolivier01 commented on issue #9744: Performance regression when
OMP_NUM_THREADS environment variable is not set
URL:
https://github.com/apache/incubator-mxnet/issues/9744#issuecomment-367727199
Are you saying I should set the environment variable in the code:
cjolivier01 commented on issue #9744: Performance regression when
OMP_NUM_THREADS environment variable is not set
URL:
https://github.com/apache/incubator-mxnet/issues/9744#issuecomment-367727199
Are you sauing I should set the environment variable in the code:
cjolivier01 commented on issue #9744: Performance regression when
OMP_NUM_THREADS environment variable is not set
URL:
https://github.com/apache/incubator-mxnet/issues/9744#issuecomment-367726497
I am only vaguely familiar with the KMP stuff. Can you please elaborate?
Are you saying
chsin commented on a change in pull request #9809: fix optimizer bug in
CPP-Package
URL: https://github.com/apache/incubator-mxnet/pull/9809#discussion_r169996703
##
File path: cpp-package/include/mxnet-cpp/optimizer.hpp
##
@@ -125,13 +125,16 @@ inline float
ZiyueHuang commented on issue #9770: eye operator, for default storage type
URL: https://github.com/apache/incubator-mxnet/pull/9770#issuecomment-367719581
Thanks for your comments. Sorry for the late response.
@piiswrong I increased the size and the number of trials and added M=0
chsin closed issue #9855: Why no Codeowner for /cpp-package/?
URL: https://github.com/apache/incubator-mxnet/issues/9855
This is an automated message from the Apache Git Service.
To respond to the message, please log on
pengzhao-intel commented on issue #9744: Performance regression when
OMP_NUM_THREADS environment variable is not set
URL:
https://github.com/apache/incubator-mxnet/issues/9744#issuecomment-364357479
We found the similar issue that the performance will be worse if we don't
set
pengzhao-intel commented on issue #9744: Performance regression when
OMP_NUM_THREADS environment variable is not set
URL:
https://github.com/apache/incubator-mxnet/issues/9744#issuecomment-367711172
@cjolivier01 Thanks for the fix. I have tested in my local with your code.
The
KellenSunderland opened a new pull request #9861: Add CuDNN (v6) support to
Jetson builds
URL: https://github.com/apache/incubator-mxnet/pull/9861
## Description ##
Add CuDNN (v6) support to Jetson builds. This should make the builds for
Jetson devices 20-35% faster for many tasks.
KellenSunderland commented on issue #9861: Add CuDNN (v6) support to Jetson
builds
URL: https://github.com/apache/incubator-mxnet/pull/9861#issuecomment-367703404
Tested offline. Everything looks good, cudnn is working properly with
Jetpack 3.1
dabraude commented on issue #9860: [WIP] CMake NNPack support
URL: https://github.com/apache/incubator-mxnet/pull/9860#issuecomment-367696110
The dependencies other than NNPack can be downloaded at configure time
instead of being sub-modules if that is preferable.
marcoabreu commented on issue #9859: How to run model trained with mxnet 1.0 on
android
URL:
https://github.com/apache/incubator-mxnet/issues/9859#issuecomment-367694828
@larroy
This is an automated message from the Apache
KellenSunderland closed pull request #9861: Add CuDNN (v6) support to Jetson
builds
URL: https://github.com/apache/incubator-mxnet/pull/9861
This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:
As
This is an automated email from the ASF dual-hosted git repository.
marcoabreu pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git
The following commit(s) were added to refs/heads/master by this push:
new 6d2b6e3 Improve
KellenSunderland commented on issue #9861: Add CuDNN (v6) support to Jetson
builds
URL: https://github.com/apache/incubator-mxnet/pull/9861#issuecomment-367693725
Checking a few things offline.
This is an automated message
marcoabreu closed pull request #9849: License tool
URL: https://github.com/apache/incubator-mxnet/pull/9849
This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:
As this is a foreign pull request (from
marcoabreu commented on issue #9798: fix cmake
URL: https://github.com/apache/incubator-mxnet/pull/9798#issuecomment-367693035
Thanks you. I just checked and it's not part of the PR test chain. Could you
please make sure that it's also being used in
marcoabreu commented on issue #9857: Core dump DROPOUT_PERF.TimingGPU
URL:
https://github.com/apache/incubator-mxnet/issues/9857#issuecomment-367692383
No, this would waste too many resources. Generally, multiple failures can be
fixed by one change. If we wouldn't follow fail-fast, a lot
marcoabreu commented on a change in pull request #9861: Add CuDNN (v6) support
to Jetson builds
URL: https://github.com/apache/incubator-mxnet/pull/9861#discussion_r169963987
##
File path: docker_multiarch/Dockerfile.build.jetson
##
@@ -21,13 +21,7 @@ ADD
KellenSunderland commented on a change in pull request #9861: Add CuDNN (v6)
support to Jetson builds
URL: https://github.com/apache/incubator-mxnet/pull/9861#discussion_r169964125
##
File path: docker_multiarch/Dockerfile.build.jetson
##
@@ -21,13 +21,7 @@ ADD
ZiyueHuang commented on issue #9848: add infer_type for regression ops, fix
issues#9847
URL: https://github.com/apache/incubator-mxnet/pull/9848#issuecomment-367690061
By the way, `L2NormalizationOp` and `InstanceNormOp` don't have an
infer_type function.
KellenSunderland opened a new pull request #9861: Add CuDNN (v6) support to
Jetson builds
URL: https://github.com/apache/incubator-mxnet/pull/9861
## Description ##
Add CuDNN (v6) support to Jetson builds. This should make the builds for
Jetson devices 20-35% faster for many tasks.
ZiyueHuang commented on issue #9848: add infer_type for regression ops, fix
issues#9847
URL: https://github.com/apache/incubator-mxnet/pull/9848#issuecomment-367688993
Unittest is written in symbolic mode while the issue is in imperative mode.
In symbolic mode, infer_type pass has a
larroy commented on issue #9849: License tool
URL: https://github.com/apache/incubator-mxnet/pull/9849#issuecomment-367686078
@marcoabreu please merge?
This is an automated message from the Apache Git Service.
To respond to
yajiedesign commented on issue #9798: fix cmake
URL: https://github.com/apache/incubator-mxnet/pull/9798#issuecomment-367683536
@marcoabreu yes.the windows night build use it.
This is an automated message from the Apache Git
marcoabreu commented on issue #9798: fix cmake
URL: https://github.com/apache/incubator-mxnet/pull/9798#issuecomment-367673339
I'm not on my computer right now. could you check whether cuda toolset is
actually applied during compilation in CI?
marcoabreu commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367671272
In my opinion, the right approach here would be that this change is getting
done on the master and people who are interested in this operator
marcoabreu commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367670741
We are not running any CI for v0.11 and as far as I remember, was the plan
to only support the last two versions (in this case v1.0 and v1.1)
yajiedesign commented on issue #9798: fix cmake
URL: https://github.com/apache/incubator-mxnet/pull/9798#issuecomment-367669459
sorry for later.
I originally set up CUDA_TOOLSET for the purpose of selecting the CUDA
version in CI.
I will try to add a default auto option.
marcoabreu commented on issue #9860: [WIP] CMake NNPack support
URL: https://github.com/apache/incubator-mxnet/pull/9860#issuecomment-367669600
That's a LOT of third party dependencies and they're coming from a private
repository.
yajiedesign commented on a change in pull request #9798: fix cmake
URL: https://github.com/apache/incubator-mxnet/pull/9798#discussion_r169942811
##
File path: CMakeLists.txt
##
@@ -46,6 +46,11 @@ if(USE_CUDA AND NOT USE_OLDCMAKECUDA)
(${CMAKE_VERSION}
dabraude commented on issue #9860: CMake NNPack support
URL: https://github.com/apache/incubator-mxnet/pull/9860#issuecomment-367623208
The test that failed is flaky #9856
This is an automated message from the Apache Git
juliusshufan commented on issue #9805: Enable the reporting of cross-entropy or
nll loss value when training CNN network using the models defined by
example/image-classification
URL: https://github.com/apache/incubator-mxnet/pull/9805#issuecomment-366863185
# I rerun the failed GPU
juliusshufan commented on issue #9805: Enable the reporting of cross-entropy or
nll loss value when training CNN network using the models defined by
example/image-classification
URL: https://github.com/apache/incubator-mxnet/pull/9805#issuecomment-367618399
@szha Hi, I noticed the preci
juliusshufan commented on issue #9805: Enable the reporting of cross-entropy or
nll loss value when training CNN network using the models defined by
example/image-classification
URL: https://github.com/apache/incubator-mxnet/pull/9805#issuecomment-366863185
### I rerun the failed GPU case
dabraude commented on issue #9573: Fix bug with symbol loading
URL: https://github.com/apache/incubator-mxnet/pull/9573#issuecomment-367613732
failed tests relate to #9856
This is an automated message from the Apache Git
dabraude commented on issue #9573: Fix bug with symbol loading
URL: https://github.com/apache/incubator-mxnet/pull/9573#issuecomment-367613732
failed tests relate to #9856
This is an automated message from the Apache Git
lupesko commented on issue #9858: softsign activation function
URL: https://github.com/apache/incubator-mxnet/pull/9858#issuecomment-367608109
@szha This will help MXNet users who can't migrate from 0.11 to 1.0 just
yet, but need the operator.
dabraude opened a new pull request #9860: Nnpack
URL: https://github.com/apache/incubator-mxnet/pull/9860
## Description ##
This is the first part of #9719
## Checklist ##
### Essentials ###
- [X] Passed code style checking (`make lint`)
- [X] Changes are complete
101 - 159 of 159 matches
Mail list logo