Install and Configure Caffe on windows 10


Series

Guide

requirements:

  • windows: 10
  • caffe: caffe-windows
  • nvidia driver: gtx 1060 382.05 (gtx 970m)
  • GPU arch(s): sm_61 (sm_52)
  • cuda: 8.0
  • cudnn: 5.0.5
  • opencv: 3.1.0 WITH_CUDA (compiled from source)
  • other libs: libraries_v140_x64_py27_1.1.0.tar.bz2

cuda+cudnn

  1. download and install driver by standalone for GTX 970 or GTX 1060 from here.
  2. download and install cuda_8.0.61_win10.exe, skip install nvidia driver and install toolkit only.
  3. download and install cudnn-8.0-windows10-x64-v5.0-ga.zip.

nvidia driver

driver can be installed by standalone or from cuda_xxx_win10.exe.
we choose to install by standalone

download proper driver for GTX 970 or GTX 1060 eg: 398.36-notebook-win10-64bit-international-whql.exe from here

download driver

cuda toolkit

ref: cuda install guides for windows

download cuda_8.0.61_win10.exe from here

The CUDA Toolkit installs the CUDA driver and tools needed to create, build and run a CUDA application as well as libraries, header files, CUDA samples source code, and other resources

cuda_8.0.61_win10.exe includes: Nvidia driver + toolkit.

install to

  • driver install to C:/Program Files/NVIDIA Corporation and C:/ProgramData/NVIDIA Corporation
  • tookit install to C:/Program Files/NVIDIA GPU Computing Toolkit,which contains headers,libs,tools for compiling CUDA applications. C:/ProgramData/NVIDIA GPU Computing Toolkit contains cuda plugins for Visual Studio.

cuda driver

cuda toolkit

cuda driver data

cuda toolkit data

verify

cd C:\ProgramData\NVIDIA Corporation\CUDA Samples\v9.2\bin\win64\Release
./deviceQuery.exe

cudnn

extract cudnn-8.0-windows10-x64-v5.0-ga.zip and copy include,liband bin to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0

cudnn

check cuda

nvidia driver and cuda software installation

compile

download

  1. place caffe-windows at C:/compile/caffe-windows
  2. extract libraries_v140_x64_py27_1.1.0.tar.bz2 to C:\Users\zunli\.caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries

config

edit C:\Users\zunli\.caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries\caffe-builder-config.cmake

# BOOST config
set(BOOST_ROOT "C:/Boost/")
set(BOOST_INCLUDEDIR ${BOOST_ROOT}/include/boost-1_64 CACHE PATH "")
set(BOOST_LIBRARYDIR ${BOOST_ROOT}/lib CACHE PATH "")
set(Boost_USE_MULTITHREADED ON CACHE BOOL "")
set(Boost_USE_STATIC_LIBS ON CACHE BOOL "")
set(Boost_USE_STATIC_RUNTIME OFF CACHE BOOL "")

edit caffe-windows/cmake/Dependencies.cmake

set(Boost_USE_STATIC_LIBS ON)
find_package(Boost 1.64 REQUIRED COMPONENTS system thread filesystem)

Tips:
(1) we use C:\Boost\ 1.64 to replace caffe dependencies C:\Users\zunli\.caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries\ 1.61, because we have compile PCL 1.8.1 with Boost 1.64 static.
(2) we use caffe C:\Users\zunli\.caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries\x64\vc14\lib to replace C:/Program Files/opencv. (opencv3.1 <====opencv3.4)

configure caffe with

cd caffe
mkdir build && cd build && cmake-gui ..

with options

BLAS                 Open # Atlas, Open, MKL
BUILD_SHARED_LIBS        OFF # build static library
CMAKE_CONFIGURATION_TYPES   Release
CMAKE_CXX_RELEASE_FLAGS    /MD /O2 /Ob2 /DNDEBUG /MP

CUDA_ARCH_BIN  3.0 3.5 5.0 5.2 6.0 6.1 # very time-consuming
CUDA_ARCH_NAME Manual
CUDA_ARCH_PTX 3.0

Use Boost 1.64

caffe cuda arch

cudnn

opencv with cuda

configure and output

Selecting Windows SDK version 10.0.14393.0 to target Windows 10.0.15063.
Boost version: 1.64.0
Found the following Boost libraries:
  system
  thread
  filesystem
  chrono
  date_time
  atomic
Found gflags  (include: C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/include, library: gflags_shared)
Found glog    (include: C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/include, library: glog)
Found PROTOBUF Compiler: C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/bin/protoc.exe
Found lmdb    (include: C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/include, library: lmdb)
Found LevelDB (include: C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/include, library: leveldb)
Found Snappy  (include: C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/include, library: snappy_static;optimized;C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/lib/caffezlib.lib;debug;C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/lib/caffezlibd.lib)
CUDA detected: 8.0
Found cuDNN: ver. 5.0.5 found (include: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0/include, library: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0/lib/x64/cudnn.lib)
Added CUDA NVCC flags for: sm_61
OpenCV found (C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries)
Found OpenBLAS libraries: C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/lib/libopenblas.dll.a
Found OpenBLAS include: C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/include
NumPy ver. 1.11.3 found (include: C:/Python27/lib/site-packages/numpy/core/include)
Boost version: 1.64.0
Found the following Boost libraries:
  python

******************* Caffe Configuration Summary *******************
General:
  Version           :   1.0.0
  Git               :   unknown
  System            :   Windows
  C++ compiler      :   C:/Program Files (x86)/Microsoft Visual Studio 14.0/VC/bin/x86_amd64/cl.exe
  Release CXX flags :   /MD /O2 /Ob2 /DNDEBUG /MP /DWIN32 /D_WINDOWS /W3 /GR /EHsc
  Debug CXX flags   :   /MDd /Zi /Ob0 /Od /RTC1 /DWIN32 /D_WINDOWS /W3 /GR /EHsc
  Build type        :   Release

  BUILD_SHARED_LIBS :   OFF
  BUILD_python      :   ON
  BUILD_matlab      :   OFF
  BUILD_docs        :   
  CPU_ONLY          :   OFF
  USE_OPENCV        :   ON
  USE_LEVELDB       :   ON
  USE_LMDB          :   ON
  USE_NCCL          :   OFF
  ALLOW_LMDB_NOLOCK :   OFF

Dependencies:
  BLAS              :   Yes (Open)
  Boost             :   Yes (ver. 1.64)
  glog              :   Yes
  gflags            :   Yes
  protobuf          :   Yes (ver. 3.1.0)
  lmdb              :   Yes (ver. 0.9.70)
  LevelDB           :   Yes (ver. 1.18)
  Snappy            :   Yes (ver. 1.1.1)
  OpenCV            :   Yes (ver. 3.1.0)
  CUDA              :   Yes (ver. 8.0)

NVIDIA CUDA:
  Target GPU(s)     :   Auto
  GPU arch(s)       :   sm_61
  cuDNN             :   Yes (ver. 5.0.5)

Python:
  Interpreter       :   C:/Python27/python.exe (ver. 2.7.13)
  Libraries         :   C:/Python27/libs/python27.lib (ver 2.7.13)
  NumPy             :   C:/Python27/lib/site-packages/numpy/core/include (ver 1.11.3)

Install:
  Install path      :   C:/car_libs/caffe

Configuring done

build and install

tips: Visual Studio 2015 can not generate shared library. So we build static caffe library.

CMake Error at CMakeLists.txt:66 (message):
  The Visual Studio generator cannot build a shared library.  Use the Ninja
  generator instead.

Build with Release x64 with Visual Studio 2015 and 38 modules will be generated and We Install to C:/car_libs/caffe/.
build with vs

build result.
build result

install to C:/car_libs/caffe.

caffe usage

CMakeLists.txt

# Boost
if(MSVC)
    # use static boost on windows
    set(Boost_USE_STATIC_LIBS ON) # 
else()
    # use release boost on linux
    set(Boost_USE_STATIC_LIBS OFF)
endif(MSVC)

set(Boost_USE_MULTITHREAD ON)
# Find Boost package 1.64 (caffe also use Boost 1.64)
find_package(Boost 1.64 REQUIRED COMPONENTS serialization date_time system filesystem thread timer math_tr1)

# opencv 
SET(OpenCV_DIR "C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/")
find_package(OpenCV REQUIRED COMPONENTS core highgui imgproc features2d calib3d) # nofree for 2.4

# caffe
set(Caffe_DIR "C:/car_libs/caffe/share/Caffe/")
find_package(Caffe)

when we use caffe lib in our program, errors will occur. And we need to fix CaffeTargets-release.cmake file。

usage error fix

(1) error with shared.lib

LNK1181    unable to open“gflags_shared.lib” 

solution:

vim C:/car_libs/caffe/share/Caffe/CaffeTargets-release.cmake

# remove _shared -shared
:1,$s/_shared//g
:1,$s/-shared//g

(2) error with hdf5

hdf5.lib===>libcaffehdf5.lib
hdf5_hl.lib===>libcaffehdf5_hl.lib

 :1,$s/hdf5/libcaffehdf5/g

(3) error with libopenblas

LNK1181    unable to open“libopenblas.dll.a.lib”

solution:

cd C:\Users\zunli\.caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries\lib and

  • copy libopenblas.a ===> libopenblas.a.lib
  • copy libopenblas.dll.a ===> libopenblas.dll.a.lib

(4) error NtClose

error LNK2019: 无法解析的外部符号 NtClose,该符号在函数 mdb_env_map 中被引用

solution:

copy `C:/Program Files (x86)/Windows Kits/10/Lib/10.0.14393.0/um/x64/ntdll.lib` to `C:\Users\zunli\.caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries\lib`
copy `C:\Windows\SysWOW64\ntdll.dll` to `C:\Users\zunli\.caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries\bin`

CaffeTargets-release.cmake

edit C:\car_libs\caffe\share\Caffe\CaffeTargets-release.cmake

#----------------------------------------------------------------
# Generated CMake target import file for configuration "Release".
#----------------------------------------------------------------

# Commands may need to know the format version.
set(CMAKE_IMPORT_FILE_VERSION 1)

# Import target "caffe" for configuration "Release"
set_property(TARGET caffe APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
set_target_properties(caffe PROPERTIES
  IMPORTED_LINK_INTERFACE_LANGUAGES_RELEASE "CXX"
  IMPORTED_LINK_INTERFACE_LIBRARIES_RELEASE 
"caffeproto;C:/Boost/lib/libboost_system-vc140-mt-1_64.lib;C:/Boost/lib/libboost_thread-vc140-mt-1_64.lib;C:/Boost/lib/libboost_filesystem-vc140-mt-1_64.lib;C:/Boost/lib/libboost_chrono-vc140-mt-1_64.lib;C:/Boost/lib/libboost_date_time-vc140-mt-1_64.lib;C:/Boost/lib/libboost_atomic-vc140-mt-1_64.lib;C:/Boost/lib/libboost_python-vc140-mt-1_64.lib;caffehdf5.lib;caffehdf5_cpp.lib;caffehdf5_hl.lib;caffehdf5_hl_cpp.lib;caffezlib.lib;caffezlibstatic.lib;gflags;glog;leveldb.lib;libcaffehdf5.lib;libcaffehdf5_cpp.lib;libcaffehdf5_hl.lib;libcaffehdf5_hl_cpp.lib;libprotobuf.lib;libprotoc.lib;lmdb.lib;snappy.lib;snappy_static.lib;libopenblas.dll.a.lib;ntdll.lib;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0/lib/x64/cudart.lib;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0/lib/x64/curand.lib;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0/lib/x64/cublas.lib;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0/lib/x64/cublas_device.lib;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0/lib/x64/cudnn.lib;opencv_core;opencv_highgui;opencv_imgproc;opencv_imgcodecs;C:/Python27/libs/python27.lib;"
  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/lib/caffe.lib"
  )

list(APPEND _IMPORT_CHECK_TARGETS caffe )
list(APPEND _IMPORT_CHECK_FILES_FOR_caffe "${_IMPORT_PREFIX}/lib/caffe.lib" )

# Import target "caffeproto" for configuration "Release"
set_property(TARGET caffeproto APPEND PROPERTY IMPORTED_CONFIGURATIONS RELEASE)
set_target_properties(caffeproto PROPERTIES
  IMPORTED_LINK_INTERFACE_LANGUAGES_RELEASE "CXX"
  IMPORTED_LINK_INTERFACE_LIBRARIES_RELEASE "C:/Users/zunli/.caffe/dependencies/libraries_v140_x64_py27_1.1.0/libraries/lib/libprotobuf.lib"
  IMPORTED_LOCATION_RELEASE "${_IMPORT_PREFIX}/lib/caffeproto.lib"
  )

list(APPEND _IMPORT_CHECK_TARGETS caffeproto )
list(APPEND _IMPORT_CHECK_FILES_FOR_caffeproto "${_IMPORT_PREFIX}/lib/caffeproto.lib" )

# Commands beyond this point should not need to know the version.
set(CMAKE_IMPORT_FILE_VERSION)

comiple errors with caffe.pb.h

tips: sometimes we not need to do this.

CMakeLists.txt

add_definitions( -DGLOG_NO_ABBREVIATED_SEVERITIES ) 
add_definitions( -DNOMINMAX )  # for pcl min,max
add_definitions( -DWIN32_LEAN_AND_MEAN ) 
#add_definitions( -DNO_STRICT ) # no use for caffe.pb.h

vim C:\car_libs\caffe\include\caffe\proto\caffe.pb.h

typedef ParamSpec_DimCheckMode DimCheckMode;
static const DimCheckMode STRICT = ParamSpec_DimCheckMode_STRICT;
static const DimCheckMode PERMISSIVE = ParamSpec_DimCheckMode_PERMISSIVE;

typedef V1LayerParameter_DimCheckMode DimCheckMode;
static const DimCheckMode STRICT = V1LayerParameter_DimCheckMode_STRICT;
static const DimCheckMode PERMISSIVE = V1LayerParameter_DimCheckMode_PERMISSIVE;

replace STRICT and PERMISSIVE to _STRICT and _PERMISSIVE.

typedef ParamSpec_DimCheckMode DimCheckMode;
static const DimCheckMode _STRICT = ParamSpec_DimCheckMode_STRICT;
static const DimCheckMode _PERMISSIVE = ParamSpec_DimCheckMode_PERMISSIVE;

typedef V1LayerParameter_DimCheckMode DimCheckMode;
static const DimCheckMode _STRICT = V1LayerParameter_DimCheckMode_STRICT;
static const DimCheckMode _PERMISSIVE = V1LayerParameter_DimCheckMode_PERMISSIVE;

caffe.pb.h compile errors

run exe

  • copy C:/car_libs/caffe/bin/*.dll dlls to bin/release folder.
  • copy Opencv dlls to bin/release folder.

Errors and Solutions

nvidia driver not compatible with windows 10

problem: install nvidia driver failed on windows 10
nvidia driver not compatible with windows 10

solution

  1. download Windows10Upgrade
  2. run Windows10Upgrade.exe to upgrade windows 10 to latest.
  3. install nvidia driver again.
  4. OK.

Reference

History

  • 20180413 created.

Author: kezunlin
Reprint policy: All articles in this blog are used except for special statements CC BY 4.0 reprint polocy. If reproduced, please indicate source kezunlin !
评论
  TOC