電腦配置
系統(tǒng)：Ubuntu16.04
GPU：NVIDIA GTX1080
安裝依賴
sudo apt-get update
sudo apt-get upgrade
sudo apt-get install -y build-essential cmake git pkg-config
sudo apt-get install -y libprotobuf-dev libleveldb-dev libsnappy-dev libhdf5-serial-dev protobuf-compiler
sudo apt-get install -y libatlas-base-dev libgflags-dev libgoogle-glog-dev liblmdb-dev
sudo apt-get install -y --no-install-recommends libboost-all-dev
# (Python general)
sudo apt-get install -y python-pip
# (Python 2.7 development files)
sudo apt-get install -y python-dev
sudo apt-get install -y python-numpy python-scipy
安裝Caffe
(1) 從github上獲取caffe (get git: sudo apt-get install git)：
git clone https://github.com/BVLC/caffe.git
在caffe-master/python文件夾內，使用root執(zhí)行依賴項的檢查與安裝：
sudo su
cd caffe-master/python
for req in $(cat requirements.txt); do pip install $req; done
以上過程報錯寞缝，需要解決后再進行丈钙，可通過升級pip解決部分問題：
pip --version
pip install -u pip
(2) cp Makefile.config：
sudo cp Makefile.config.example Makefile.config
(3) 打開并修改配置文件：
sudo gedit Makefile.config #打開Makefile.config文件
3.1 使用CUDNN：
將
#USE_CUDNN := 1
修改成：
USE_CUDNN := 1
3.2 使用OpenCV
USE_OPENCV := 1 （可選）
OPENCV_VERSION := 3 （如果打開了USE_OPENCV，而且opencv版本為3時）
3.3 Python來編寫layer
將
#WITH_PYTHON_LAYER := 1
修改為
WITH_PYTHON_LAYER := 1
3.4 路徑變化
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib
修改為：
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu /usr/lib/x86_64-linux-gnu/hdf5/serial
修改makefile文件
打開makefile文件蛉腌，做如下修改：
將：
NVCCFLAGS +=-ccbin=$(CXX) -Xcompiler-fPIC $(COMMON_FLAGS)
替換為：
NVCCFLAGS += -D_FORCE_INLINES -ccbin=$(CXX) -Xcompiler -fPIC $(COMMON_FLAGS)
編譯
make all -j8 -數(shù)字6為并行編譯進程數(shù)目榜掌，一般設置為CPU物理核心數(shù)溶锭，可以使用 $(nproc) 來替換或echo $(nproc) 來查看CPU物理核心數(shù)
make test -j8
make runtest -j8
make pycaffe -此時應該已經編譯完成宝恶，再次執(zhí)行以確認
忽視以下錯誤即可：
nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
測試
MNIST數(shù)據集測試
配置caffe完成后，我們可以利用MNIST數(shù)據集對caffe進行測試，過程如下：
1.將終端定位到Caffe根目錄
cd ~/caffe
2.下載MNIST數(shù)據庫并解壓縮
./data/mnist/get_mnist.sh
3.將其轉換成Lmdb數(shù)據庫格式
./examples/mnist/create_mnist.sh
4.訓練網絡
./examples/mnist/train_lenet.sh
訓練開始：
I0922 13:14:47.332710 1418 caffe.cpp:218] Using GPUs 0
I0922 13:14:47.390920 1418 caffe.cpp:223] GPU 0: TITAN X (Pascal)
I0922 13:14:47.593374 1418 solver.cpp:44] Initializing solver from parameters:
test_iter: 100
test_interval: 500
base_lr: 0.01
display: 100
max_iter: 10000
lr_policy: "inv"
gamma: 0.0001
power: 0.75
momentum: 0.9
weight_decay: 0.0005
snapshot: 5000
snapshot_prefix: "examples/mnist/lenet"
solver_mode: GPU
device_id: 0
...
I0922 13:15:00.678990 1418 solver.cpp:447] Snapshotting to binary proto file examples/mnist/lenet_iter_10000.caffemodel
I0922 13:15:00.682633 1418 sgd_solver.cpp:273] Snapshotting solver state to binary proto file examples/mnist/lenet_iter_10000.solverstate
I0922 13:15:00.684578 1418 solver.cpp:310] Iteration 10000, loss = 0.00334544
I0922 13:15:00.684592 1418 solver.cpp:330] Iteration 10000, Testing net (#0)
I0922 13:15:00.732586 1429 data_layer.cpp:73] Restarting data prefetching from start.
I0922 13:15:00.733906 1418 solver.cpp:397] Test net output #0: accuracy = 0.9908
I0922 13:15:00.733917 1418 solver.cpp:397] Test net output #1: loss = 0.0276493 (* 1 = 0.0276493 loss)
I0922 13:15:00.733922 1418 solver.cpp:315] Optimization Done.
I0922 13:15:00.733924 1418 caffe.cpp:259] Optimization Done.
參考：
http://blog.csdn.net/aifei7320/article/details/73252635（此文記錄了部分錯誤記錄）
http://www.reibang.com/p/7df78120803a
