[GitHub] yajiedesign commented on issue #10613: Add Windows MKLDNN Building Instruction
yajiedesign commented on issue #10613: Add Windows MKLDNN Building Instruction URL: https://github.com/apache/incubator-mxnet/pull/10613#issuecomment-383278835 i am fixed it https://github.com/apache/incubator-mxnet/pull/10629 This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] zhaoxy2018 opened a new issue #10634: How to get internal output with C++ API?
zhaoxy2018 opened a new issue #10634: How to get internal output with C++ API? URL: https://github.com/apache/incubator-mxnet/issues/10634 I want to get some intermediate results in the network with the C++ API. I found an old issue but the link suggested in the solution (https://github.com/dmlc/mxnet/tree/master/example/cpp/image-classification) is no longer valid now. So what is the proper way to inspect network intermediate results with the new C++ API? This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] xinyu-intel commented on issue #10613: Add Windows MKLDNN Building Instruction
xinyu-intel commented on issue #10613: Add Windows MKLDNN Building Instruction URL: https://github.com/apache/incubator-mxnet/pull/10613#issuecomment-383282908 @yajiedesign Good J:) I think it is very helpful to simplify my workflow! This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183207896 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: not sure if the omp and mklml are also in 3rdparty\\mkldnn\\ This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183208559 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: no omp mklml is in the root path.i am copy it in cmake, the problem is that I can't figure out the location of mkldnn This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] yajiedesign commented on issue #10626: Could you help build windows pypi package mxnet-cu90 for version 1.0.0 and version 1.1.0?
yajiedesign commented on issue #10626: Could you help build windows pypi package mxnet-cu90 for version 1.0.0 and version 1.1.0? URL: https://github.com/apache/incubator-mxnet/issues/10626#issuecomment-383291805 all done This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] chinakook commented on issue #10633: [MXNET-346] Hard Sigmoid Operator
chinakook commented on issue #10633: [MXNET-346] Hard Sigmoid Operator URL: https://github.com/apache/incubator-mxnet/pull/10633#issuecomment-383292186 It’s widely used in LSTM in Keras. Could add this op into LSTM or GRU, etc.. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183209056 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: may be 3rdparty\\mkldnn\\Release\\mkldnn.dll This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183209056 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: maybe 3rdparty\\mkldnn\\src\\Release\\mkldnn.dll This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183209056 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: maybe 3rdparty\\mkldnn\\Release\\mkldnn.dll This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183209056 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: maybe 3rdparty\\mkldnn\\src\\mkldnn.dll This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
xinyu-intel commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183209561 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: seems 3rdparty\mkldnn\src\mkldnn.dll is right but still some deps are missing. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183209684 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: yes,I don't know what is missing.Maybe msvcrt120.dll. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183209713 ## File path: Jenkinsfile ## @@ -358,7 +358,7 @@ try { mkdir pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libmxnet.lib pkg_%BUILD_NAME%\\lib copy build_%BUILD_NAME%\\libmxnet.dll pkg_%BUILD_NAME%\\build - copy build_%BUILD_NAME%\\mkldnn.dll pkg_%BUILD_NAME%\\build + copy build_%BUILD_NAME%\\3rdparty\\mkldnn\\mkldnn.dll pkg_%BUILD_NAME%\\build copy build_%BUILD_NAME%\\libiomp5md.dll pkg_%BUILD_NAME%\\build Review comment: need some one login and look at it with depends. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] mrkn opened a new pull request #10635: Stop ignoring the given name of CompositeEvalMetric
mrkn opened a new pull request #10635: Stop ignoring the given name of CompositeEvalMetric URL: https://github.com/apache/incubator-mxnet/pull/10635 mxnet.metric.CompositeEvalMetric ignores the given name in its constructor. It shoudn't be ignored. Before: ``` >>> acc = mx.metric.CompositeEvalMetric(name='composite_xyz') >>> acc.name 'composite' ``` After: ``` >>> acc = mx.metric.CompositeEvalMetric(name='composite_xyz') >>> acc.name 'composite_xyz' ``` This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] barkntuncer opened a new issue #10636: How to install mxnet for CUDA8.0 properly? mxnet-cu80 error
barkntuncer opened a new issue #10636: How to install mxnet for CUDA8.0 properly? mxnet-cu80 error URL: https://github.com/apache/incubator-mxnet/issues/10636 ## Description Trying to install mxnet for CUDA-8.0 with mxnet-cu80 while following [here](https://mxnet.incubator.apache.org/install/index.html). When I try the python code below on the page, it gives error. ## Environment info (Required) Ubuntu 16.4 Python 3.5.2 CUDA-8.0 nvidia 950m ``` What to do: 1. Download the diagnosis script from https://raw.githubusercontent.com/apache/incubator-mxnet/master/tools/diagnose.py 2. Run the script using `python diagnose.py` and paste its output here. --Python Info-- Version : 3.5.2 Compiler : GCC 5.4.0 20160609 Build: ('default', 'Nov 23 2017 16:37:01') Arch : ('64bit', 'ELF') Pip Info--- Version : 8.1.1 Directory: /usr/lib/python3/dist-packages/pip --MXNet Info--- Version : 0.11.0 Directory: /usr/local/lib/python3.5/dist-packages/mxnet Commit Hash : 53274b4a2b0d73f3fbdb10cfb5f9ed0c8263fda7 --System Info-- Platform : Linux-4.13.0-37-generic-x86_64-with-Ubuntu-16.04-xenial system : Linux node : barkntuncer release : 4.13.0-37-generic version : #42~16.04.1-Ubuntu SMP Wed Mar 7 16:03:28 UTC 2018 --Hardware Info-- machine : x86_64 processor: x86_64 Architecture: x86_64 CPU op-mode(s):32-bit, 64-bit Byte Order:Little Endian CPU(s):8 On-line CPU(s) list: 0-7 Thread(s) per core:2 Core(s) per socket:4 Socket(s): 1 NUMA node(s): 1 Vendor ID: GenuineIntel CPU family:6 Model: 94 Model name:Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz Stepping: 3 CPU MHz: 2600.000 CPU max MHz: 3500, CPU min MHz: 800, BogoMIPS: 5184.00 Virtualization:VT-x L1d cache: 32K L1i cache: 32K L2 cache: 256K L3 cache: 6144K NUMA node0 CPU(s): 0-7 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb invpcid_single pti retpoline intel_pt rsb_ctxsw tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx rdseed adx smap clflushopt xsaveopt xsavec xgetbv1 xsaves dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp --Network Test-- Setting timeout: 10 Timing for Gluon Tutorial(cn): https://zh.gluon.ai, DNS: 0.0031 sec, LOAD: 0.2751 sec. Timing for PYPI: https://pypi.python.org/pypi/pip, DNS: 0.1011 sec, LOAD: 1.1429 sec. Timing for Gluon Tutorial(en): http://gluon.mxnet.io, DNS: 0.0016 sec, LOAD: 0.1321 sec. Timing for FashionMNIST: https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/dataset/fashion-mnist/train-labels-idx1-ubyte.gz, DNS: 0.0009 sec, LOAD: 0.5194 sec. Timing for Conda: https://repo.continuum.io/pkgs/free/, DNS: 0.0014 sec, LOAD: 0.2902 sec. Timing for MXNet: https://github.com/apache/incubator-mxnet, DNS: 0.0010 sec, LOAD: 1.0151 sec. ``` Package used (Python/R/Scala/Julia): (I'm using PYTHON) For Scala user, please provide: 1. Java version: (`java -version`) 2. Maven version: (`mvn -version`) 3. Scala runtime if applicable: (`scala -version`) For R user, please provide R `sessionInfo()`: ## Build info (Required if built from source) Compiler (gcc/clang/mingw/visual studio): MXNet commit hash: (Paste the output of `git rev-parse HEAD` here.) Build config: (Paste the content of config.mk, or the build command.) ## Error Message: (Paste the complete error message, including stack trace.) terminate called after throwing an instance of 'dmlc::Errorterminate called after throwing an instance of 'dmlc::Error' ' what(): [18:32:00] /home/travis/build/dmlc/mxnet-distro/mxnet-build/mshadow/mshadow/./tensor_gpu-inl.h:35: Check failed: e == cudaSuccess CUDA: unknown error Stack trace returned 9 entries: [bt] (0) /usr/local/lib/python3.5/dist-packages/mxnet/libmxnet.so(+0x2ab998) [0x7fb7a6b54998] [bt] (1) /usr/local/lib/python3.5/dist-packages/mxnet/libmxnet.so(+0x2abda8) [0x7fb7a6b54da8] [bt] (2) /usr/local/lib/python3
[GitHub] marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183215160 ## File path: CMakeLists.txt ## @@ -18,8 +18,8 @@ mxnet_option(USE_CUDNN"Build with cudnn support" ON) # one could se mxnet_option(USE_SSE "Build with x86 SSE instruction support" ON) mxnet_option(USE_LAPACK "Build with lapack support" ON IF NOT MSVC) mxnet_option(USE_MKL_IF_AVAILABLE "Use MKL if found" ON) -mxnet_option(USE_MKLML_MKL"Use MKLDNN variant of MKL (if MKL found)" ON IF USE_MKL_IF_AVAILABLE AND UNIX AND (NOT APPLE)) -mxnet_option(USE_MKLDNN "Use MKLDNN variant of MKL (if MKL found)" ON IF USE_MKL_IF_AVAILABLE AND UNIX AND (NOT APPLE)) +mxnet_option(USE_MKLML_MKL"Use MKLDNN variant of MKL (if MKL found)" OFF IF (NOT APPLE)) Review comment: Please keep the USE_MKL_IF_AVAILABLE argument - you're changing the default behaviour here This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183215160 ## File path: CMakeLists.txt ## @@ -18,8 +18,8 @@ mxnet_option(USE_CUDNN"Build with cudnn support" ON) # one could se mxnet_option(USE_SSE "Build with x86 SSE instruction support" ON) mxnet_option(USE_LAPACK "Build with lapack support" ON IF NOT MSVC) mxnet_option(USE_MKL_IF_AVAILABLE "Use MKL if found" ON) -mxnet_option(USE_MKLML_MKL"Use MKLDNN variant of MKL (if MKL found)" ON IF USE_MKL_IF_AVAILABLE AND UNIX AND (NOT APPLE)) -mxnet_option(USE_MKLDNN "Use MKLDNN variant of MKL (if MKL found)" ON IF USE_MKL_IF_AVAILABLE AND UNIX AND (NOT APPLE)) +mxnet_option(USE_MKLML_MKL"Use MKLDNN variant of MKL (if MKL found)" OFF IF (NOT APPLE)) Review comment: Please keep the USE_MKL_IF_AVAILABLE argument This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183215179 ## File path: Jenkinsfile ## @@ -311,7 +311,7 @@ try { bat """mkdir build_vc14_gpu call "C:\\Program Files (x86)\\Microsoft Visual Studio 14.0\\VC\\bin\\x86_amd64\\vcvarsx86_amd64.bat" cd build_vc14_gpu - cmake -G \"NMake Makefiles JOM\" -DUSE_CUDA=1 -DUSE_CUDNN=1 -DUSE_NVRTC=1 -DUSE_OPENCV=1 -DUSE_OPENMP=1 -DUSE_PROFILER=1 -DUSE_BLAS=open -DUSE_LAPACK=1 -DUSE_DIST_KVSTORE=0 -DCUDA_ARCH_NAME=All -DCMAKE_CXX_FLAGS_RELEASE="/FS /MD /O2 /Ob2 /DNDEBUG" -DCMAKE_BUILD_TYPE=Release ${env.WORKSPACE}""" + cmake -G \"NMake Makefiles JOM\" -DUSE_CUDA=1 -DUSE_CUDNN=1 -DUSE_NVRTC=1 -DUSE_OPENCV=1 -DUSE_OPENMP=1 -DUSE_PROFILER=1 -DUSE_BLAS=open -DUSE_LAPACK=1 -DUSE_DIST_KVSTORE=0 -DCUDA_ARCH_NAME=Maxwell -DCMAKE_CXX_FLAGS_RELEASE="/FS /MD /O2 /Ob2 /DNDEBUG" -DCMAKE_BUILD_TYPE=Release ${env.WORKSPACE}""" Review comment: Please elaborate this change? We're trying to cover all supported archs This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183215222 ## File path: Jenkinsfile ## @@ -334,6 +334,47 @@ try { } } }, +'Build GPU MKLDNN windows':{ + node('mxnetwindows-cpu') { +timeout(time: max_time, unit: 'MINUTES') { + ws('workspace/build-gpu') { +withEnv(['OpenBLAS_HOME=C:\\mxnet\\openblas', 'OpenCV_DIR=C:\\mxnet\\opencv_vc14', 'CUDA_PATH=C:\\CUDA\\v8.0','BUILD_NAME=vc14_gpu_mkldnn']) { +init_git_win() +bat """mkdir build_%BUILD_NAME% + call "C:\\Program Files (x86)\\Microsoft Visual Studio 14.0\\VC\\bin\\x86_amd64\\vcvarsx86_amd64.bat" + cd build_%BUILD_NAME% + copy ${env.WORKSPACE}\\3rdparty\\mkldnn\\config_template.vcxproj.user ${env.WORKSPACE}\\config_template.vcxproj.user /y + cmake -G \"NMake Makefiles JOM\" -DUSE_CUDA=1 -DUSE_CUDNN=1 -DUSE_NVRTC=1 -DUSE_OPENCV=1 -DUSE_OPENMP=1 -DUSE_PROFILER=1 -DUSE_BLAS=open -DUSE_LAPACK=1 -DUSE_DIST_KVSTORE=0 -DCUDA_ARCH_NAME=Maxwell -DUSE_MKLDNN=1 -DCMAKE_CXX_FLAGS_RELEASE="/FS /MD /O2 /Ob2 /DNDEBUG" -DCMAKE_BUILD_TYPE=Release ${env.WORKSPACE}""" +bat ''' +call "C:\\Program Files (x86)\\Microsoft Visual Studio 14.0\\VC\\bin\\x86_amd64\\vcvarsx86_amd64.bat" +cd build_%BUILD_NAME% +set /a cores=36 * 2 +jom -j 72 Review comment: Please use the proper env variables and don't hardcode these values This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183215251 ## File path: cmake/FirstClassLangCuda.cmake ## @@ -125,8 +125,8 @@ endif () # mshadow_select_nvcc_arch_flags(out_variable) function(mshadow_select_nvcc_arch_flags out_variable) - set(CUDA_ARCH_LIST "Auto" CACHE STRING "Select target NVIDIA GPU achitecture.") - set_property( CACHE CUDA_ARCH_LIST PROPERTY STRINGS "" "All" "Common" ${CUDA_KNOWN_GPU_ARCHITECTURES} ) + set(CUDA_ARCH_LIST "" CACHE STRING "Select target NVIDIA GPU achitecture.") Review comment: default behaviour change? This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183215285 ## File path: python/mxnet/libinfo.py ## @@ -37,6 +38,8 @@ def find_lib_path(): logging.warning("MXNET_LIBRARY_PATH should be an absolute path, instead of: %s", lib_from_env) else: +if os.name == 'nt': +os.environ['PATH'] = os.path.dirname(lib_from_env) + ';' + os.environ['PATH'] Review comment: Please append to the end of path instead of the beginning to prevent path masking This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
marcoabreu commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183215290 ## File path: python/mxnet/libinfo.py ## @@ -69,6 +72,8 @@ def find_lib_path(): if len(lib_path) == 0: raise RuntimeError('Cannot find the MXNet library.\n' + 'List of candidates:\n' + str('\n'.join(dll_path))) +if os.name == 'nt': +os.environ['PATH'] = os.path.dirname(lib_path[0]) + ';' + os.environ['PATH'] Review comment: same here This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] barkntuncer closed issue #10636: How to install mxnet for CUDA8.0 properly? mxnet-cu80 error
barkntuncer closed issue #10636: How to install mxnet for CUDA8.0 properly? mxnet-cu80 error URL: https://github.com/apache/incubator-mxnet/issues/10636 This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] barkntuncer commented on issue #10636: How to install mxnet for CUDA8.0 properly? mxnet-cu80 error
barkntuncer commented on issue #10636: How to install mxnet for CUDA8.0 properly? mxnet-cu80 error URL: https://github.com/apache/incubator-mxnet/issues/10636#issuecomment-383318396 When I reinstall NVIDIA driver, it fixed This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] haojin2 commented on issue #10633: [MXNET-346] Hard Sigmoid Operator
haojin2 commented on issue #10633: [MXNET-346] Hard Sigmoid Operator URL: https://github.com/apache/incubator-mxnet/pull/10633#issuecomment-383328533 @chinakook Thanks for your advice! I'll definitely talk to related persons about this. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183222432 ## File path: cmake/FirstClassLangCuda.cmake ## @@ -125,8 +125,8 @@ endif () # mshadow_select_nvcc_arch_flags(out_variable) function(mshadow_select_nvcc_arch_flags out_variable) - set(CUDA_ARCH_LIST "Auto" CACHE STRING "Select target NVIDIA GPU achitecture.") - set_property( CACHE CUDA_ARCH_LIST PROPERTY STRINGS "" "All" "Common" ${CUDA_KNOWN_GPU_ARCHITECTURES} ) + set(CUDA_ARCH_LIST "" CACHE STRING "Select target NVIDIA GPU achitecture.") Review comment: no.still is "Empty". This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183222490 ## File path: Jenkinsfile ## @@ -311,7 +311,7 @@ try { bat """mkdir build_vc14_gpu call "C:\\Program Files (x86)\\Microsoft Visual Studio 14.0\\VC\\bin\\x86_amd64\\vcvarsx86_amd64.bat" cd build_vc14_gpu - cmake -G \"NMake Makefiles JOM\" -DUSE_CUDA=1 -DUSE_CUDNN=1 -DUSE_NVRTC=1 -DUSE_OPENCV=1 -DUSE_OPENMP=1 -DUSE_PROFILER=1 -DUSE_BLAS=open -DUSE_LAPACK=1 -DUSE_DIST_KVSTORE=0 -DCUDA_ARCH_NAME=All -DCMAKE_CXX_FLAGS_RELEASE="/FS /MD /O2 /Ob2 /DNDEBUG" -DCMAKE_BUILD_TYPE=Release ${env.WORKSPACE}""" + cmake -G \"NMake Makefiles JOM\" -DUSE_CUDA=1 -DUSE_CUDNN=1 -DUSE_NVRTC=1 -DUSE_OPENCV=1 -DUSE_OPENMP=1 -DUSE_PROFILER=1 -DUSE_BLAS=open -DUSE_LAPACK=1 -DUSE_DIST_KVSTORE=0 -DCUDA_ARCH_NAME=Maxwell -DCMAKE_CXX_FLAGS_RELEASE="/FS /MD /O2 /Ob2 /DNDEBUG" -DCMAKE_BUILD_TYPE=Release ${env.WORKSPACE}""" Review comment: This only speeds up compilation speed. Theoretically, it will not affect the correctness of programs. In fact, even if you use "ALL", it is not a test of "ALL" at the time of testing.the NV driver will only choose a suitable one. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc
yajiedesign commented on a change in pull request #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#discussion_r183222432 ## File path: cmake/FirstClassLangCuda.cmake ## @@ -125,8 +125,8 @@ endif () # mshadow_select_nvcc_arch_flags(out_variable) function(mshadow_select_nvcc_arch_flags out_variable) - set(CUDA_ARCH_LIST "Auto" CACHE STRING "Select target NVIDIA GPU achitecture.") - set_property( CACHE CUDA_ARCH_LIST PROPERTY STRINGS "" "All" "Common" ${CUDA_KNOWN_GPU_ARCHITECTURES} ) + set(CUDA_ARCH_LIST "" CACHE STRING "Select target NVIDIA GPU achitecture.") Review comment: no.still is "Empty". This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] xinyu-intel commented on issue #10629: [MXNET-343]fix Mkldnn with msvc
xinyu-intel commented on issue #10629: [MXNET-343]fix Mkldnn with msvc URL: https://github.com/apache/incubator-mxnet/pull/10629#issuecomment-383351919 @yajiedesign @marcoabreu It seems all check passed, I'll modify readme in #10613 when this pr merged. Thank you:) This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil commented on issue #1356: override global initialization method in layer configuration
ThomasDelteil commented on issue #1356: override global initialization method in layer configuration URL: https://github.com/apache/incubator-mxnet/issues/1356#issuecomment-383355423 I am trying to load a np array into a Dense layer as weight like this: ```python weights = np.zeros((512,512)) np.fill_diagonal(weights,1) last_layer = gluon.nn.Dense( 512, use_bias=False, weight_initializer=mx.initializer.Load({'last_layer_weight':nd.array(weights, ctx)}), prefix='last_layer_', in_units=512 ) last_layer.initialize() ``` I get: ``` --- AssertionErrorTraceback (most recent call last) in () 8 in_units=512 9 ) ---> 10 last_layer.initialize() ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/block.py in initialize(self, init, ctx, verbose, force_reinit) 380 Whether to force re-initialization if parameter is already initialized. 381 """ --> 382 self.collect_params().initialize(init, ctx, verbose, force_reinit) 383 384 def hybridize(self, active=True, **kwargs): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in initialize(self, init, ctx, verbose, force_reinit) 685 init.set_verbosity(verbose=verbose) 686 for _, v in self.items(): --> 687 v.initialize(None, ctx, init, force_reinit=force_reinit) 688 689 def zero_grad(self): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in initialize(self, init, ctx, default_init, force_reinit) 338 339 self._deferred_init = (init, ctx, default_init, None) --> 340 self._finish_deferred_init() 341 342 def reset_ctx(self, ctx): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in _finish_deferred_init(self) 240 ctx=context.cpu()) 241 initializer.create(default_init)( --> 242 initializer.InitDesc(self.name, {'__init__': init}), data) 243 244 self._init_impl(data, ctx) ~/anaconda3/lib/python3.6/site-packages/mxnet/initializer.py in __call__(self, desc, arr) 136 if init: 137 # when calling Variable initializer --> 138 create(init)._init_weight(desc, arr) 139 self._verbose_print(desc, init, arr) 140 else: ~/anaconda3/lib/python3.6/site-packages/mxnet/registry.py in create(*args, **kwargs) 147 return create(**name) 148 --> 149 assert isinstance(name, string_types), "%s must be of string type"%nickname 150 151 if name.startswith('['): AssertionError: initializer must be of string type ``` This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil commented on issue #1356: override global initialization method in layer configuration
ThomasDelteil commented on issue #1356: override global initialization method in layer configuration URL: https://github.com/apache/incubator-mxnet/issues/1356#issuecomment-383355423 I am trying to load a np array into a Dense layer as weight like this: ```python weights = np.zeros((512,512)) np.fill_diagonal(weights,1) last_layer = gluon.nn.Dense( 512, use_bias=False, weight_initializer=mx.initializer.Load({'last_layer_weight':nd.array(weights, ctx)}), prefix='last_layer_', in_units=512 ) last_layer.initialize() ``` I get: ``` --- AssertionErrorTraceback (most recent call last) in () 8 in_units=512 9 ) ---> 10 last_layer.initialize() ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/block.py in initialize(self, init, ctx, verbose, force_reinit) 380 Whether to force re-initialization if parameter is already initialized. 381 """ --> 382 self.collect_params().initialize(init, ctx, verbose, force_reinit) 383 384 def hybridize(self, active=True, **kwargs): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in initialize(self, init, ctx, verbose, force_reinit) 685 init.set_verbosity(verbose=verbose) 686 for _, v in self.items(): --> 687 v.initialize(None, ctx, init, force_reinit=force_reinit) 688 689 def zero_grad(self): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in initialize(self, init, ctx, default_init, force_reinit) 338 339 self._deferred_init = (init, ctx, default_init, None) --> 340 self._finish_deferred_init() 341 342 def reset_ctx(self, ctx): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in _finish_deferred_init(self) 240 ctx=context.cpu()) 241 initializer.create(default_init)( --> 242 initializer.InitDesc(self.name, {'__init__': init}), data) 243 244 self._init_impl(data, ctx) ~/anaconda3/lib/python3.6/site-packages/mxnet/initializer.py in __call__(self, desc, arr) 136 if init: 137 # when calling Variable initializer --> 138 create(init)._init_weight(desc, arr) 139 self._verbose_print(desc, init, arr) 140 else: ~/anaconda3/lib/python3.6/site-packages/mxnet/registry.py in create(*args, **kwargs) 147 return create(**name) 148 --> 149 assert isinstance(name, string_types), "%s must be of string type"%nickname 150 151 if name.startswith('['): AssertionError: initializer must be of string type ``` update: To accomplish the following you can use: ``` weights = np.zeros((512,512)) np.fill_diagonal(weights,1) last_layer = gluon.nn.Dense(512, in_units=512, use_bias=False, weight_initializer=mx.init.Constant(nd.array(weights))) last_layer.collect_params().initialize(ctx=ctx) last_layer.params['dense0_weight'].data() ``` This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil commented on issue #1356: override global initialization method in layer configuration
ThomasDelteil commented on issue #1356: override global initialization method in layer configuration URL: https://github.com/apache/incubator-mxnet/issues/1356#issuecomment-383355423 I am trying to load a np array into a Dense layer as weight like this: ```python weights = np.zeros((512,512)) np.fill_diagonal(weights,1) last_layer = gluon.nn.Dense( 512, use_bias=False, weight_initializer=mx.initializer.Load({'last_layer_weight':nd.array(weights, ctx)}), prefix='last_layer_', in_units=512 ) last_layer.initialize() ``` I get: ``` --- AssertionErrorTraceback (most recent call last) in () 8 in_units=512 9 ) ---> 10 last_layer.initialize() ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/block.py in initialize(self, init, ctx, verbose, force_reinit) 380 Whether to force re-initialization if parameter is already initialized. 381 """ --> 382 self.collect_params().initialize(init, ctx, verbose, force_reinit) 383 384 def hybridize(self, active=True, **kwargs): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in initialize(self, init, ctx, verbose, force_reinit) 685 init.set_verbosity(verbose=verbose) 686 for _, v in self.items(): --> 687 v.initialize(None, ctx, init, force_reinit=force_reinit) 688 689 def zero_grad(self): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in initialize(self, init, ctx, default_init, force_reinit) 338 339 self._deferred_init = (init, ctx, default_init, None) --> 340 self._finish_deferred_init() 341 342 def reset_ctx(self, ctx): ~/anaconda3/lib/python3.6/site-packages/mxnet/gluon/parameter.py in _finish_deferred_init(self) 240 ctx=context.cpu()) 241 initializer.create(default_init)( --> 242 initializer.InitDesc(self.name, {'__init__': init}), data) 243 244 self._init_impl(data, ctx) ~/anaconda3/lib/python3.6/site-packages/mxnet/initializer.py in __call__(self, desc, arr) 136 if init: 137 # when calling Variable initializer --> 138 create(init)._init_weight(desc, arr) 139 self._verbose_print(desc, init, arr) 140 else: ~/anaconda3/lib/python3.6/site-packages/mxnet/registry.py in create(*args, **kwargs) 147 return create(**name) 148 --> 149 assert isinstance(name, string_types), "%s must be of string type"%nickname 150 151 if name.startswith('['): AssertionError: initializer must be of string type ``` update: To accomplish the following you can use: ```python last_layer = gluon.nn.Dense( 512, in_units=512, use_bias=False, weight_initializer=mx.init.Constant(nd.array(np.identity(512))) ) last_layer.collect_params().initialize(ctx=ctx) last_layer.params['dense0_weight'].data() ``` This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha closed issue #10626: Could you help build windows pypi package mxnet-cu90 for version 1.0.0 and version 1.1.0?
szha closed issue #10626: Could you help build windows pypi package mxnet-cu90 for version 1.0.0 and version 1.1.0? URL: https://github.com/apache/incubator-mxnet/issues/10626 This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil opened a new pull request #10637: [MXNET-352] Document behavior of mx.initializer.Constant
ThomasDelteil opened a new pull request #10637: [MXNET-352] Document behavior of mx.initializer.Constant URL: https://github.com/apache/incubator-mxnet/pull/10637 ## Description ## As I was looking for a way to initialize the value of my weights to a given specific value, I couldn't find any straightforward way to do that without hacking into the parameters and setting the value directly. Turns out you can use the mx.initializer.Constant, even though the documentation states that it is only use for scalar. If you give a NDArray of the right shape, you can initialize your weights using the `Constant` initializer, which is much cleaner than setting the data directly since you don't have to play around the names of the layers and parameters. ## Checklist ## ### Essentials ### Please feel free to remove inapplicable items for your PR. - [ ] The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant [JIRA issue](https://issues.apache.org/jira/projects/MXNET/issues) created (except PRs with tiny changes) - [ ] Changes are complete (i.e. I finished coding on this PR) - [ ] All changes have test coverage: - Unit tests are added for small changes to verify correctness (e.g. adding a new operator) - Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore) - Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL) - [ ] Code is well-documented: - For user-facing API changes, API doc string has been updated. - For new C++ functions in header files, their functionalities and arguments are documented. - For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable - Check the API doc at http://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$PR_ID/$BUILD_ID/index.html - [ ] To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change ### Changes ### - [ ] Feature1, tests, (and when applicable, API doc) - [ ] Feature2, tests, (and when applicable, API doc) ## Comments ## - If this change is a backward incompatible change, why must this change be made. - Interesting edge cases to note here This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] piiswrong commented on issue #10628: [MXNET-342] Fix the multi worker Dataloader
piiswrong commented on issue #10628: [MXNET-342] Fix the multi worker Dataloader URL: https://github.com/apache/incubator-mxnet/pull/10628#issuecomment-383356835 this is a generic issue not specific to recordiodataset. Any dataset that opens a file could be affected. Does it behave the same way if the Dataset is written in python and opens file with `open`? This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil opened a new issue #10638: [Feature Request] Gluon model zoo allow fine-tuning
ThomasDelteil opened a new issue #10638: [Feature Request] Gluon model zoo allow fine-tuning URL: https://github.com/apache/incubator-mxnet/issues/10638 Currently it is a little bit of a faff to fine-tune a model from the model-zoo. The straight dope takes the option of downloading both the pre-trained and non-trained model and assigning the trained features to the non-trained model. http://gluon.mxnet.io/chapter08_computer-vision/fine-tuning.html As done in the naming tutorial, you can also re assign the `.output` layer of the model from the model zoo but that again requires knowledge of the code. I propose that when calling: ```python vision.resnet18_v1(pretrained=True, classes=10) ``` instead of getting an error with: ``` AssertionError: Failed loading Parameter 'resnetv10_dense0_weight' from saved params: shape incompatible expected (10, 512) vs saved (1000, 512) ``` We get the pre-trained model with the last layer replaced with a 10-unit dense layer, which makes more sense from a user perspective. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning
szha commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning URL: https://github.com/apache/incubator-mxnet/issues/10638#issuecomment-383357918 The semantics of such call might be ambiguous. Given that we already consistently name the network parts as "features" and "output", I think an easier approach is to just document this structure and make it a convention. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] szha commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning
szha commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning URL: https://github.com/apache/incubator-mxnet/issues/10638#issuecomment-383357918 The semantics of such call might be ambiguous. Given that we already consistently name the network parts as "features" and "output", I think an easier approach is to just document this structure and convention. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil commented on issue #10628: [MXNET-342] Fix the multi worker Dataloader
ThomasDelteil commented on issue #10628: [MXNET-342] Fix the multi worker Dataloader URL: https://github.com/apache/incubator-mxnet/pull/10628#issuecomment-383357995 It would behave the same way if the dataset relies on reading the file at run-time. To make it clearer, instead of checking the `RecordIODataset`, we could have `RecordIODataset` inheriting from a new abstract class `FileReadingDataset` for example, that documents the behavior and has an abstract method `reload_file` that we call on the worker loop similarly as proposed in this PR. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning
ThomasDelteil commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning URL: https://github.com/apache/incubator-mxnet/issues/10638#issuecomment-383358167 I agree that documenting the structure of the network would be very helpful to start with. And also documenting the fact that the `classes` argument is only valid with `pre_trained=False`, in the docs and by throwing an appropriate exception. However I don't really see the ambiguity in the semantics? Could you elaborate in which case there would be ambiguity? This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning
ThomasDelteil commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning URL: https://github.com/apache/incubator-mxnet/issues/10638#issuecomment-383358379 Also @szha , I think another common use case is to use a network as a featurizer. Currently the only way I found without going to the symbolic representation is to replace the output layer with a dense layer without bias and with the identity matrix as weight. It would be nice if we could have an option to set `Featurizer=True` and it would set the output layer to a pass-through `HybridLambda` for example. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] luoyetx commented on issue #10544: name_scope/prefix doesn't work
luoyetx commented on issue #10544: name_scope/prefix doesn't work URL: https://github.com/apache/incubator-mxnet/issues/10544#issuecomment-383358915 parameters of `SymbolBlock` is not saved. ```python import mxnet as mx from mxnet import gluon as gl, nd from mxnet.gluon import nn class Net(gl.HybridBlock): def __init__(self): super(Net, self).__init__() with self.name_scope(): backbone = gl.model_zoo.vision.resnet50_v1() data = mx.sym.var('data') featnames = ['stage1_activation2', 'stage2_activation3', 'stage3_activation5'] out_names = ['_'.join([backbone.name, featname, 'output']) for featname in featnames] internals = backbone(data).get_internals() outs = [internals[out_name] for out_name in out_names] self.backbone = gl.SymbolBlock(outs, data, params=backbone.collect_params()) self.body = nn.Conv2D(3, 1) def hybrid_forward(self, F, x): x = self.body(x) return self.backbone(x) ctx = mx.cpu() net = Net() net.initialize(mx.init.Normal(), ctx=ctx) net.hybridize() net(nd.random.normal(shape=(1, 3, 224, 224))) net.save_params('./test.params') for k, v in nd.load('./test.params').items(): print(k) for k, v in net.collect_params().items(): print(k) ``` gets ``` body.bias body.weight ``` and ``` net5_resnetv10_conv0_weight net5_resnetv10_batchnorm0_gamma net5_resnetv10_batchnorm0_beta net5_resnetv10_stage1_conv0_weight net5_resnetv10_stage1_conv0_bias net5_resnetv10_stage1_batchnorm0_gamma net5_resnetv10_stage1_batchnorm0_beta net5_resnetv10_stage1_conv1_weight net5_resnetv10_stage1_batchnorm1_gamma net5_resnetv10_stage1_batchnorm1_beta net5_resnetv10_stage1_conv2_weight net5_resnetv10_stage1_conv2_bias net5_resnetv10_stage1_batchnorm2_gamma net5_resnetv10_stage1_batchnorm2_beta net5_resnetv10_stage1_conv3_weight net5_resnetv10_stage1_batchnorm3_gamma net5_resnetv10_stage1_batchnorm3_beta net5_resnetv10_stage1_conv4_weight net5_resnetv10_stage1_conv4_bias net5_resnetv10_stage1_batchnorm4_gamma net5_resnetv10_stage1_batchnorm4_beta net5_resnetv10_stage1_conv5_weight net5_resnetv10_stage1_batchnorm5_gamma net5_resnetv10_stage1_batchnorm5_beta net5_resnetv10_stage1_conv6_weight net5_resnetv10_stage1_conv6_bias net5_resnetv10_stage1_batchnorm6_gamma net5_resnetv10_stage1_batchnorm6_beta net5_resnetv10_stage1_conv7_weight net5_resnetv10_stage1_conv7_bias net5_resnetv10_stage1_batchnorm7_gamma net5_resnetv10_stage1_batchnorm7_beta net5_resnetv10_stage1_conv8_weight net5_resnetv10_stage1_batchnorm8_gamma net5_resnetv10_stage1_batchnorm8_beta net5_resnetv10_stage1_conv9_weight net5_resnetv10_stage1_conv9_bias net5_resnetv10_stage1_batchnorm9_gamma net5_resnetv10_stage1_batchnorm9_beta net5_resnetv10_stage2_conv0_weight net5_resnetv10_stage2_conv0_bias net5_resnetv10_stage2_batchnorm0_gamma net5_resnetv10_stage2_batchnorm0_beta net5_resnetv10_stage2_conv1_weight net5_resnetv10_stage2_batchnorm1_gamma net5_resnetv10_stage2_batchnorm1_beta net5_resnetv10_stage2_conv2_weight net5_resnetv10_stage2_conv2_bias net5_resnetv10_stage2_batchnorm2_gamma net5_resnetv10_stage2_batchnorm2_beta net5_resnetv10_stage2_conv3_weight net5_resnetv10_stage2_batchnorm3_gamma net5_resnetv10_stage2_batchnorm3_beta net5_resnetv10_stage2_conv4_weight net5_resnetv10_stage2_conv4_bias net5_resnetv10_stage2_batchnorm4_gamma net5_resnetv10_stage2_batchnorm4_beta net5_resnetv10_stage2_conv5_weight net5_resnetv10_stage2_batchnorm5_gamma net5_resnetv10_stage2_batchnorm5_beta net5_resnetv10_stage2_conv6_weight net5_resnetv10_stage2_conv6_bias net5_resnetv10_stage2_batchnorm6_gamma net5_resnetv10_stage2_batchnorm6_beta net5_resnetv10_stage2_conv7_weight net5_resnetv10_stage2_conv7_bias net5_resnetv10_stage2_batchnorm7_gamma net5_resnetv10_stage2_batchnorm7_beta net5_resnetv10_stage2_conv8_weight net5_resnetv10_stage2_batchnorm8_gamma net5_resnetv10_stage2_batchnorm8_beta net5_resnetv10_stage2_conv9_weight net5_resnetv10_stage2_conv9_bias net5_resnetv10_stage2_batchnorm9_gamma net5_resnetv10_stage2_batchnorm9_beta net5_resnetv10_stage2_conv10_weight net5_resnetv10_stage2_conv10_bias net5_resnetv10_stage2_batchnorm10_gamma net5_resnetv10_stage2_batchnorm10_beta net5_resnetv10_stage2_conv11_weight net5_resnetv10_stage2_batchnorm11_gamma net5_resnetv10_stage2_batchnorm11_beta net5_resnetv10_stage2_conv12_weight net5_resnetv10_stage2_conv12_bias net5_resnetv10_stage2_batchnorm12_gamma net5_resnetv10_stage2_batchnorm12_beta net5_resnetv10_stage3_conv0_weight net5_resnetv10_stage3_con
[GitHub] szha commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning
szha commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning URL: https://github.com/apache/incubator-mxnet/issues/10638#issuecomment-383359131 In the proposed design, without the knowledge of actual implementation, the call may either be interpreted as "get me a pre-trained model and replace the output layer with a size of 10", or "get me a pre-trained model that was trained on 10 classes". For using the network as feature extractor, simply use the net.features block. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning
ThomasDelteil commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning URL: https://github.com/apache/incubator-mxnet/issues/10638#issuecomment-383358167 I agree that documenting the structure of the network would be very helpful to start with. And also documenting the fact that the `classes` argument is only valid with `pre_trained=False`, in the docs and by throwing an appropriate exception. However I don't really see the ambiguity in the semantics? Could you elaborate in which case there would be ambiguity? edit: for mobilenet and squeezenet, it is not completely straightforward as the last layers are not dense layers, so the method used should be the one featured in the straight dope. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] ThomasDelteil commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning
ThomasDelteil commented on issue #10638: [Feature Request] Gluon model zoo allow fine-tuning URL: https://github.com/apache/incubator-mxnet/issues/10638#issuecomment-383359513 :+1: that's much simpler indeed, thanks. Maybe having a `PreTrained` class that documents the `.features` and `.output`, and various usage could be a good idea. I see the ambiguity now, one way to circumvent this ambiguity could be that setting specifically the number of classes means getting an untrained last layer. But the issue would be that it is not backward compatible with people having already set pre_trained=True and num_class=1000 :/ Though we could make an exception for these and returning the pre-trained layer. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services