[GitHub] eric-haibin-lin commented on issue #8294: NCCL integration

eric-haibin-lin commented on issue #8294: NCCL integration URL: https://github.com/apache/incubator-mxnet/pull/8294#issuecomment-343719635 @mbaijal We'll need your help setting up a new CI test with NCCL build after this is merged.

[GitHub] cjolivier01 closed pull request #8622: Fix changed OMP call in LaunchEx, add broadcast perf test

cjolivier01 closed pull request #8622: Fix changed OMP call in LaunchEx, add broadcast perf test URL: https://github.com/apache/incubator-mxnet/pull/8622 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

[GitHub] eric-haibin-lin commented on a change in pull request #8180: Add wide and deep model into sparse example

eric-haibin-lin commented on a change in pull request #8180: Add wide and deep model into sparse example URL: https://github.com/apache/incubator-mxnet/pull/8180#discussion_r150389300 ## File path: example/sparse/wide_deep_classification.py ## @@ -0,0 +1,125 @@ +#

[GitHub] eric-haibin-lin commented on issue #8601: Why is the accuracy overestimated?

eric-haibin-lin commented on issue #8601: Why is the accuracy overestimated? URL: https://github.com/apache/incubator-mxnet/issues/8601#issuecomment-343715907 Are you using k-fold validation? This is an automated message

[GitHub] shuokay commented on issue #8623: How to set different learning rate to different layers with pretained model which only have symbols json file and params file?

shuokay commented on issue #8623: How to set different learning rate to different layers with pretained model which only have symbols json file and params file? URL: https://github.com/apache/incubator-mxnet/issues/8623#issuecomment-343715866 After digging into source code, I think it is

[GitHub] eric-haibin-lin commented on a change in pull request #8620: Support sparse for custom python operators

eric-haibin-lin commented on a change in pull request #8620: Support sparse for custom python operators URL: https://github.com/apache/incubator-mxnet/pull/8620#discussion_r150401097 ## File path: python/mxnet/operator.py ## @@ -513,6 +515,51 @@ def infer_type(self,

[GitHub] eric-haibin-lin commented on a change in pull request #8620: Support sparse for custom python operators

eric-haibin-lin commented on a change in pull request #8620: Support sparse for custom python operators URL: https://github.com/apache/incubator-mxnet/pull/8620#discussion_r150401177 ## File path: src/operator/custom/custom.cc ## @@ -342,19 +351,99 @@ void Backward(const

[GitHub] eric-haibin-lin commented on a change in pull request #8620: Support sparse for custom python operators

eric-haibin-lin commented on a change in pull request #8620: Support sparse for custom python operators URL: https://github.com/apache/incubator-mxnet/pull/8620#discussion_r150401068 ## File path: python/mxnet/operator.py ## @@ -513,6 +515,51 @@ def infer_type(self,

[GitHub] shuokay opened a new issue #8623: How to set different learning rate to different layers with pretained model which only have symbols json file and params file?

shuokay opened a new issue #8623: How to set different learning rate to different layers with pretained model which only have symbols json file and params file? URL: https://github.com/apache/incubator-mxnet/issues/8623 Model downloaded from model zoo only have symbol json file and model

[GitHub] jiarenyf commented on issue #8606: How to quick share weight like the tensorflow using reused_params !!!

jiarenyf commented on issue #8606: How to quick share weight like the tensorflow using reused_params !!! URL: https://github.com/apache/incubator-mxnet/issues/8606#issuecomment-343714396 I also want to share params with two different inputs, but I find it's not easy to quickly share

[GitHub] cjolivier01 commented on issue #8572: use first class cuda with cmake 3.9 and cuda9.0 support

cjolivier01 commented on issue #8572: use first class cuda with cmake 3.9 and cuda9.0 support URL: https://github.com/apache/incubator-mxnet/pull/8572#issuecomment-343713031 Actually, this doesn;t build at all on my machine :( Can't find cuda.h

[GitHub] cjolivier01 commented on issue #8572: use first class cuda with cmake 3.9 and cuda9.0 support

cjolivier01 commented on issue #8572: use first class cuda with cmake 3.9 and cuda9.0 support URL: https://github.com/apache/incubator-mxnet/pull/8572#issuecomment-343712994 Does this build all CUDA architectures? The old one used to detect. Building all architectures is super-slow during

[GitHub] cjolivier01 opened a new pull request #8622: Fix changed OMP call in LaunchEx, add broadcast perf test

cjolivier01 opened a new pull request #8622: Fix changed OMP call in LaunchEx, add broadcast perf test URL: https://github.com/apache/incubator-mxnet/pull/8622 @piiswrong ## Description ## Build problem in master fixed Performance test for broadcast op ## Checklist ##

[GitHub] burness opened a new pull request #8621: fix row_sparse demo tutorials doc

burness opened a new pull request #8621: fix row_sparse demo tutorials doc URL: https://github.com/apache/incubator-mxnet/pull/8621 ## Description ## [row sparse demo tutorials](https://mxnet.incubator.apache.org/versions/master/tutorials/sparse/row_sparse.html), it has a mistake that

[GitHub] yajiedesign commented on issue #8619: Revising steps for building MXNet on Windows platform

yajiedesign commented on issue #8619: Revising steps for building MXNet on Windows platform URL: https://github.com/apache/incubator-mxnet/pull/8619#issuecomment-343712231 There are problems with vs2017 +cuda9.0 now.some cu can't build.

[GitHub] anirudh2290 opened a new pull request #8620: Support sparse for custom python operators

anirudh2290 opened a new pull request #8620: Support sparse for custom python operators URL: https://github.com/apache/incubator-mxnet/pull/8620 ## Description ## This PR is to support sparse for custom python operators. Specifically, one should be able to use sparse storage types with

[GitHub] piiswrong commented on issue #8578: Add development instructions and update installation instructions

piiswrong commented on issue #8578: Add development instructions and update installation instructions URL: https://github.com/apache/incubator-mxnet/pull/8578#issuecomment-343709642 yes This is an automated message from the

[GitHub] piiswrong closed pull request #8616: Fix python 3 docstring

piiswrong closed pull request #8616: Fix python 3 docstring URL: https://github.com/apache/incubator-mxnet/pull/8616 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

[incubator-mxnet] branch master updated: Fix python 3 docstring (#8616)

This is an automated email from the ASF dual-hosted git repository. jxie 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 79b00cd Fix python 3 docstring (#8616)

[GitHub] szha commented on issue #2111: Is it normal that MXNet consumes much more system memory than Caffe during training in GPU mode?

szha commented on issue #2111: Is it normal that MXNet consumes much more system memory than Caffe during training in GPU mode? URL: https://github.com/apache/incubator-mxnet/issues/2111#issuecomment-343704547 This issue is closed due to lack of activity in the last 90 days. Feel free to

[GitHub] szha closed issue #7154: mx.contrib.sym.ctc_loss slow down extremely on gpu with large alphabet_size,but faster on cpu

szha closed issue #7154: mx.contrib.sym.ctc_loss slow down extremely on gpu with large alphabet_size,but faster on cpu URL: https://github.com/apache/incubator-mxnet/issues/7154 This is an automated message from the Apache

[GitHub] szha closed issue #2111: Is it normal that MXNet consumes much more system memory than Caffe during training in GPU mode?

szha closed issue #2111: Is it normal that MXNet consumes much more system memory than Caffe during training in GPU mode? URL: https://github.com/apache/incubator-mxnet/issues/2111 This is an automated message from the

[GitHub] szha commented on issue #7154: mx.contrib.sym.ctc_loss slow down extremely on gpu with large alphabet_size,but faster on cpu

szha commented on issue #7154: mx.contrib.sym.ctc_loss slow down extremely on gpu with large alphabet_size,but faster on cpu URL: https://github.com/apache/incubator-mxnet/issues/7154#issuecomment-343704545 This issue is closed due to lack of activity in the last 90 days. Feel free to

[GitHub] larroy commented on issue #8578: Add development instructions and update installation instructions

larroy commented on issue #8578: Add development instructions and update installation instructions URL: https://github.com/apache/incubator-mxnet/pull/8578#issuecomment-343701620 @piiswrong so in ubuntu we should leave the instructions about Make?

[GitHub] baijin93 commented on issue #2925: FCN32s: crop offset is different from the berkeley released network.

baijin93 commented on issue #2925: FCN32s: crop offset is different from the berkeley released network. URL: https://github.com/apache/incubator-mxnet/issues/2925#issuecomment-343700653 @tornadomeet Is it correct to use `center_crop=True` here instead of manually computing the `offset`s?

[GitHub] analog-cbarber commented on issue #8312: Gradient function not returning enough gradients

analog-cbarber commented on issue #8312: Gradient function not returning enough gradients URL: https://github.com/apache/incubator-mxnet/issues/8312#issuecomment-343698329 What happened with this fix? Is it going to get merged into master or is the fix incorrect? I built piiswrong:aux

[GitHub] eftiquar opened a new pull request #8619: Revising steps for building MXNet on Windows platform

eftiquar opened a new pull request #8619: Revising steps for building MXNet on Windows platform URL: https://github.com/apache/incubator-mxnet/pull/8619 ## Description ## The steps for building MXNet on Windows have been enhanced by adding more details to make it more customer

[GitHub] piiswrong commented on issue #8618: Segfault for custom op without inputs

piiswrong commented on issue #8618: Segfault for custom op without inputs URL: https://github.com/apache/incubator-mxnet/issues/8618#issuecomment-343691265 Thanks for reporting. I'm working on a fix This is an automated

[GitHub] piiswrong commented on issue #8423: Re-implement segnet in MXnet

piiswrong commented on issue #8423: Re-implement segnet in MXnet URL: https://github.com/apache/incubator-mxnet/pull/8423#issuecomment-343691025 Result on the official site is: Model | Global Accuracy | Class Accuracy | Mean I/U -- | -- | -- | -- Segnet-Basic | 82.8% | 62.3% |

[incubator-mxnet] branch master updated: Pull more optimize and simplification changes from tuner branch (#8599)

This is an automated email from the ASF dual-hosted git repository. jxie 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 8ba5de8 Pull more optimize and

[GitHub] piiswrong commented on issue #8578: Add development instructions and update installation instructions

piiswrong commented on issue #8578: Add development instructions and update installation instructions URL: https://github.com/apache/incubator-mxnet/pull/8578#issuecomment-343690666 cmake is an extra dependency. This is an

[GitHub] piiswrong closed pull request #8599: Pull more optimize and simplification changes from tuner branch

piiswrong closed pull request #8599: Pull more optimize and simplification changes from tuner branch URL: https://github.com/apache/incubator-mxnet/pull/8599 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

[GitHub] piiswrong closed pull request #8566: optimize broadcast

piiswrong closed pull request #8566: optimize broadcast URL: https://github.com/apache/incubator-mxnet/pull/8566 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

[incubator-mxnet] branch master updated: optimize broadcast (#8566)

This is an automated email from the ASF dual-hosted git repository. jxie 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 23a9294 optimize broadcast (#8566)

[incubator-mxnet] branch master updated: generalize array dataset (#8612)

This is an automated email from the ASF dual-hosted git repository. jxie 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 9572340 generalize array dataset

[GitHub] piiswrong closed pull request #8612: generalize array dataset

piiswrong closed pull request #8612: generalize array dataset URL: https://github.com/apache/incubator-mxnet/pull/8612 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

[GitHub] larroy commented on issue #8578: Add development instructions and update installation instructions

larroy commented on issue #8578: Add development instructions and update installation instructions URL: https://github.com/apache/incubator-mxnet/pull/8578#issuecomment-343688330 @rahul003 we did a quick test, and seems 10-20% faster.

[GitHub] larroy commented on issue #8578: Add development instructions and update installation instructions

larroy commented on issue #8578: Add development instructions and update installation instructions URL: https://github.com/apache/incubator-mxnet/pull/8578#issuecomment-343688235 @piiswrong isn't it better to use CMake in the official build instructions in all platforms also for

[GitHub] eric-haibin-lin commented on issue #8500: program crash when run sparse model predict

eric-haibin-lin commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-343685478 - io: Are you using `LibSVMIter` for io? - metrics: it might affect speed depending on your implementation. Are you

[GitHub] liumilan commented on issue #8500: program crash when run sparse model predict

liumilan commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-343683671 Linear_classfication.py is also changed ,call model.fit directly # model # The positive class weight, says how much

[GitHub] liumilan commented on issue #8500: program crash when run sparse model predict

liumilan commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-343683671 Linear_classfication.py is also changed # model # The positive class weight, says how much more we should upweight

[GitHub] liumilan commented on issue #8500: program crash when run sparse model predict

liumilan commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-343683550 I just set 2 environment values in mxnet. export MXNET_GPU_COPY_NTHREADS=4 export MXNET_CPU_WORKER_NTHREADS=20

[GitHub] liumilan commented on issue #8500: program crash when run sparse model predict

liumilan commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-343683356 The legacy code is implented by mpi for lr,used in my product,not mxnet. The trainning code is modified a little by your

[GitHub] eric-haibin-lin commented on a change in pull request #8180: Add wide and deep model into sparse example

eric-haibin-lin commented on a change in pull request #8180: Add wide and deep model into sparse example URL: https://github.com/apache/incubator-mxnet/pull/8180#discussion_r150389300 ## File path: example/sparse/wide_deep_classification.py ## @@ -0,0 +1,125 @@ +#

[GitHub] eric-haibin-lin commented on issue #8500: program crash when run sparse model predict

eric-haibin-lin commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-343681725 What was your legacy code based on? Is it MXNet dense NDArray implementation? For cpu training, did you train setting

[GitHub] expnn opened a new issue #8617: Failed to detect tcmalloc or jemalloc for Fedora 26

expnn opened a new issue #8617: Failed to detect tcmalloc or jemalloc for Fedora 26 URL: https://github.com/apache/incubator-mxnet/issues/8617 The wildcard given by the current Makefile failed to detect the gperftools/jemalloc malloc libraries. Currently, it uses the following snippet

[GitHub] MennoK commented on issue #8600: How many new data points does data augmentation with mx.io.ImageRecordIter produce?

MennoK commented on issue #8600: How many new data points does data augmentation with mx.io.ImageRecordIter produce? URL: https://github.com/apache/incubator-mxnet/issues/8600#issuecomment-343664960 Thanks for answering. Is it wrong to increase your training data explicitly by

[GitHub] kpengboy opened a new pull request #8616: Fix python 3 docstring

kpengboy opened a new pull request #8616: Fix python 3 docstring URL: https://github.com/apache/incubator-mxnet/pull/8616 ## Description ## Fixes a docstring for Python 3. ## Checklist ## ### Essentials ### - [x] Passed code style checking (`make lint`) - [x] Changes are

[GitHub] KaiyuYue commented on issue #8615: [solved] the link problem of libmxnet.so

KaiyuYue commented on issue #8615: [solved] the link problem of libmxnet.so URL: https://github.com/apache/incubator-mxnet/issues/8615#issuecomment-343658302 [solved] I will close this issue. The type of problem like `libmxnet.so` cannot link `mkl`, `cudnn`, or any other libs you

[GitHub] KaiyuYue closed issue #8615: [solved] the link problem of libmxnet.so

KaiyuYue closed issue #8615: [solved] the link problem of libmxnet.so URL: https://github.com/apache/incubator-mxnet/issues/8615 This is an automated message from the Apache Git Service. To respond to the message, please

[GitHub] KaiyuYue commented on issue #8615: the link problem of libmxnet.so

KaiyuYue commented on issue #8615: the link problem of libmxnet.so URL: https://github.com/apache/incubator-mxnet/issues/8615#issuecomment-343658302 I will close this issue. The type of problem like `libmxnet.so` cannot link `mkl`, `cudnn`, or any other libs you want to bind. Or

[GitHub] mseeger commented on issue #8338: master branch cannot build on centos 7 with cuda-8.0

mseeger commented on issue #8338: master branch cannot build on centos 7 with cuda-8.0 URL: https://github.com/apache/incubator-mxnet/issues/8338#issuecomment-343655159 Cool, this should be tried. Weird it does not raise issues for all the other setups. Well, c++ compilers are not all

[GitHub] liumilan commented on issue #8500: program crash when run sparse model predict

liumilan commented on issue #8500: program crash when run sparse model predict URL: https://github.com/apache/incubator-mxnet/issues/8500#issuecomment-343654806 It cost 1868.217 in every epoch in one gpu,seemed so slowly. There are only 2 classes ,0 and 1.Batchsizes are assigned to

[GitHub] HaoLiuHust commented on issue #8523: multiprocessing Gluon DataLoader

HaoLiuHust commented on issue #8523: multiprocessing Gluon DataLoader URL: https://github.com/apache/incubator-mxnet/pull/8523#issuecomment-343654227 is it can be used for now? This is an automated message from the Apache Git

[GitHub] liumilan closed issue #8608: how can i print all weights in monitor

liumilan closed issue #8608: how can i print all weights in monitor URL: https://github.com/apache/incubator-mxnet/issues/8608 This is an automated message from the Apache Git Service. To respond to the message, please log

[GitHub] kaiyuyue opened a new issue #8615: the link problem of libmxnet.so

kaiyuyue opened a new issue #8615: the link problem of libmxnet.so URL: https://github.com/apache/incubator-mxnet/issues/8615 Dear all, This may be wired. I put the cuda, cudnn and other dependencies in a folder together. At the same time, I change their specific CLFAGS and LDFLAGS