rahul003 opened a new issue #11549: Pip package can be faster (OpenCV version?) URL: https://github.com/apache/incubator-mxnet/issues/11549 ## Description When I build MXNet from source using same build flags as pip package uses, I get much faster speeds. Not sure which versions of dependencies could be different in our case. I used all standard dependencies versions that come with AWS Deep Learning AMI. This could be because of the OpenCV version used to build. Happy to provide any information needed on this front. Here is a comparison of running Resnet50v1 on the Imagenet dataset using 8 V100s on a p3.16xl machine, all else being equal. **Built from source: ~5200 samples/sec Nightly pip package: ~3500 samples/sec** ## Environment info ``` ----------Python Info---------- Version : 3.6.4 Compiler : GCC 7.2.0 Build : ('default', 'Jan 16 2018 18:10:19') Arch : ('64bit', '') ------------Pip Info----------- Version : 9.0.1 Directory : /home/ubuntu/anaconda3/lib/python3.6/site-packages/pip ----------MXNet Info----------- No MXNet installed. ----------System Info---------- Platform : Linux-4.4.0-1061-aws-x86_64-with-debian-stretch-sid system : Linux node : ip-172-31-15-196 release : 4.4.0-1061-aws version : #70-Ubuntu SMP Fri May 25 21:47:34 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): 64 On-line CPU(s) list: 0-63 Thread(s) per core: 2 Core(s) per socket: 16 Socket(s): 2 NUMA node(s): 2 Vendor ID: GenuineIntel CPU family: 6 Model: 79 Model name: Intel(R) Xeon(R) CPU E5-2686 v4 @ 2.30GHz Stepping: 1 CPU MHz: 1200.402 CPU max MHz: 3000.0000 CPU min MHz: 1200.0000 BogoMIPS: 4600.10 Hypervisor vendor: Xen Virtualization type: full L1d cache: 32K L1i cache: 32K L2 cache: 256K L3 cache: 46080K NUMA node0 CPU(s): 0-15,32-47 NUMA node1 CPU(s): 16-31,48-63 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq monitor est ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single kaiser fsgsbase bmi1 hle avx2 smep bmi2 erms invpcid rtm rdseed adx xsaveopt ida ----------Network Test---------- Setting timeout: 10 Timing for MXNet: https://github.com/apache/incubator-mxnet, DNS: 0.0057 sec, LOAD: 0.4343 sec. Timing for Gluon Tutorial(en): http://gluon.mxnet.io, DNS: 0.1129 sec, LOAD: 0.1909 sec. Timing for Gluon Tutorial(cn): https://zh.gluon.ai, DNS: 0.2744 sec, LOAD: 0.1096 sec. Timing for FashionMNIST: https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/dataset/fashion-mnist/train-labels-idx1-ubyte.gz, DNS: 0.0097 sec, LOAD: 0.0958 sec. Timing for PYPI: https://pypi.python.org/pypi/pip, DNS: 0.0022 sec, LOAD: 0.3638 sec. Timing for Conda: https://repo.continuum.io/pkgs/free/, DNS: 0.0136 sec, LOAD: 0.0810 sec. ``` Package used (Python/R/Scala/Julia): Python ## Build info Compiler: gcc 5.4 MXNet commit hash: 99ac2f571b3c4431e8a7bd64ad13035e22259cf9 Build config: USE_OPENMP=1 USE_DIST_KVSTORE=1 USE_BLAS=openblas USE_LAPACK=1 USE_NVRTC=1 USE_CUDA=1 USE_CUDA_PATH=/usr/local/cuda USE_CUDNN=1 USE_SIGNAL_HANDLER=1 USE_OPENCV=1 Here's the output of ldd on my binary ``` ldd lib/libmxnet.so linux-vdso.so.1 => (0x00007ffd9bdb0000) /usr/lib/libmpi_cxx.so (0x00007fab7d839000) libcudart.so.9.0 => /usr/local/cuda/lib64/libcudart.so.9.0 (0x00007fab7d5cc000) libcublas.so.9.0 => /usr/local/cuda/lib64/libcublas.so.9.0 (0x00007fab7a196000) libcurand.so.9.0 => /usr/local/cuda/lib64/libcurand.so.9.0 (0x00007fab76232000) libcusolver.so.9.0 => /usr/local/cuda/lib64/libcusolver.so.9.0 (0x00007fab71637000) libopenblas.so.0 => /usr/local/lib/libopenblas.so.0 (0x00007fab706a2000) librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fab7049a000) libopencv_imgcodecs.so.3.4 => /usr/local/lib/libopencv_imgcodecs.so.3.4 (0x00007fab6feb2000) libopencv_imgproc.so.3.4 => /usr/local/lib/libopencv_imgproc.so.3.4 (0x00007fab6d5c0000) libopencv_core.so.3.4 => /usr/local/lib/libopencv_core.so.3.4 (0x00007fab6c68b000) libcudnn.so.7 => /usr/local/cuda/lib64/libcudnn.so.7 (0x00007fab5b1f4000) libcufft.so.9.0 => /usr/local/cuda/lib64/libcufft.so.9.0 (0x00007fab53153000) libcuda.so.1 => /usr/lib/x86_64-linux-gnu/libcuda.so.1 (0x00007fab522d5000) libnvrtc.so.9.0 => /usr/local/cuda/lib64/libnvrtc.so.9.0 (0x00007fab50a8a000) libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fab50708000) libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fab503ff000) libgomp.so.1 => /usr/lib/x86_64-linux-gnu/libgomp.so.1 (0x00007fab501dd000) libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fab4ffc7000) libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fab4fdaa000) libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fab4f9e0000) /lib64/ld-linux-x86-64.so.2 (0x00007fab8df73000) libmpi.so.12 => /usr/lib/libmpi.so.12 (0x00007fab4f70a000) libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fab4f506000) libgfortran.so.3 => /usr/lib/x86_64-linux-gnu/libgfortran.so.3 (0x00007fab4f1db000) libjpeg.so.8 => /usr/lib/x86_64-linux-gnu/libjpeg.so.8 (0x00007fab4ef82000) libpng12.so.0 => /lib/x86_64-linux-gnu/libpng12.so.0 (0x00007fab4ed5d000) libtiff.so.5 => /usr/lib/x86_64-linux-gnu/libtiff.so.5 (0x00007fab4eae9000) libjasper.so.1 => /usr/lib/x86_64-linux-gnu/libjasper.so.1 (0x00007fab4e894000) libIlmImf-2_2.so.22 => /usr/lib/x86_64-linux-gnu/libIlmImf-2_2.so.22 (0x00007fab4e3c6000) libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007fab4e1ac000) libnvidia-fatbinaryloader.so.384.111 => /usr/lib/x86_64-linux-gnu/libnvidia-fatbinaryloader.so.384.111 (0x00007fab4df5a000) libibverbs.so.1 => /usr/lib/libibverbs.so.1 (0x00007fab4dd4b000) libopen-rte.so.12 => /usr/lib/libopen-rte.so.12 (0x00007fab4dad1000) libopen-pal.so.13 => /usr/lib/libopen-pal.so.13 (0x00007fab4d834000) libquadmath.so.0 => /usr/lib/x86_64-linux-gnu/libquadmath.so.0 (0x00007fab4d5f5000) liblzma.so.5 => /lib/x86_64-linux-gnu/liblzma.so.5 (0x00007fab4d3d3000) libjbig.so.0 => /usr/lib/x86_64-linux-gnu/libjbig.so.0 (0x00007fab4d1c5000) libHalf.so.12 => /usr/lib/x86_64-linux-gnu/libHalf.so.12 (0x00007fab4cf82000) libIex-2_2.so.12 => /usr/lib/x86_64-linux-gnu/libIex-2_2.so.12 (0x00007fab4cd64000) libIlmThread-2_2.so.12 => /usr/lib/x86_64-linux-gnu/libIlmThread-2_2.so.12 (0x00007fab4cb5d000) libhwloc.so.5 => /usr/lib/x86_64-linux-gnu/libhwloc.so.5 (0x00007fab4c923000) libutil.so.1 => /lib/x86_64-linux-gnu/libutil.so.1 (0x00007fab4c720000) libnuma.so.1 => /usr/lib/x86_64-linux-gnu/libnuma.so.1 (0x00007fab4c515000) libltdl.so.7 => /usr/lib/x86_64-linux-gnu/libltdl.so.7 (0x00007fab4c30b000) ```
---------------------------------------------------------------- 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: [email protected] With regards, Apache Git Services
