dlib GPU版本安装（待解决）

dlib GPU版本安装（待解决）

1、CUDA安装

参考

• 神经网络学习小记录42——windows下的tensorflow-gpu=1.13.2环境配置

• cuda安装教程+cudnn安装教程

2、dlib GPU版本安装参考_1

• dlib gpu版本

git clone https://github.com/davisking/dlib.git
cd dlib
mkdir build
cd build
cmake .. -DDLIB_USE_CUDA=1 -DUSE_AVX_INSTRUCTIONS=1 -D"CUDA_HOST_COMPILER=C:/Program Files (x86)/Microsoft Visual Studio 14.0/VC/bin/amd64/cl.exe"
cmake --build .
cd ..
python3.6 setup.py install --yes USE_AVX_INSTRUCTIONS --yes DLIB_USE_CUDA

cmake .. -DDLIB_USE_CUDA=1 -DUSE_AVX_INSTRUCTIONS=1命令报错：

CMake Warning at dlib/CMakeLists.txt:614 (message):
  You have CUDA installed, but we can't use it unless you put visual studio
  in 64bit mode.

Found CUDA, but CMake was unable to find the cuBLAS libraries that should be part of every basic CUDA install. Your CUDA install is somehow broken or incomplete. Since cuBLAS is required for dlib to use CUDA we won't use CUDA.
-- DID NOT FIND CUDA
-- Disabling CUDA support for dlib.  DLIB WILL NOT USE CUDA
-- C++11 activated.
-- Configuring done
-- Generating done
-- Build files have been written to: E:/dlib/dlib/build

主要原因是CUDA 10.1和CUDA 9.0目录结构不同，libcublas.so在不同文件夹下。

可以看Github上的讨论：
https://github.com/davisking/dlib/issues/1035

3、dlib GPU版本安装参考_2（只能这么整了）

参考

• Windows安装Python的CMake+dlib+Face_Recognition. 该版本和pip直接安装dlib不同，需要cmake编译，但是编译成功后，print(dlib.DLIB_USE_CUDA)显示为False

dlib安装包 & 训练文件下载地址： http://dlib.net/files/