A simple test for zola even theme

2022-09-11T00:00:00+00:00
Introduction</h2>
This is my first post, which is a post to test zola.</p>

Markdown code block</h2>
fn </span>main</span>() {
</span>    </span>println!</span>(</span>"Hello world"</span>);
</span>}
</span></code></pre>
Math</h2>
Inline math: \(e = mc^2\)</p>
Display math:
$$e = mc^2$$</p>
Images</h2>
</p>
Table</h2>
a</th> table</th> in</th> markdown</th> !!</th></tr></thead>

1</td> 2</td> 3</td> 4</td> 5</td></tr>
1</td> we</td> ew</td> we</td> with a longish column inside it</td></tr>
</tbody></table>
List</h2>

First

A</li>
</ul>
</li>
Second

B</li>
</ul>
</li>
</ul>
Numbered List</h2>

First

A</li>
B</li>
C</li>
</ol>
</li>
Second</li>
Third</li>
</ol>
Iframe</h2>

    </iframe>
</div>


在matplotlib中使用latex渲染
2019-09-12T11:35:20+00:00
代码</h2>
在代码中加入如下的代码就可以在matplotlib中使用latex渲染文字</p>
import </span>matplotlib
</span>matplotlib.use(</span>"pgf"</span>)
</span>pgf_with_custom_preamble </span>= </span>{
</span>    </span># "font.size": 40,
</span>    </span>"pgf.rcfonts"</span>: </span>False</span>,
</span>    </span>"text.usetex"</span>: </span>True</span>,
</span>    </span>"pgf.preamble"</span>: [
</span>        </span># math setup:
</span>        </span>r</span>"\usepackage{unicode-math}"</span>,
</span>        </span>r</span>"\usepackage</span>[utf8]</span>{inputenc}"</span>,
</span>        </span>r</span>"\usepackage{xeCJK}"</span>,
</span>        </span>r</span>"</span>\s</span>etmainfont{Times New Roman} % Set main font for ASCII"</span>,
</span>        </span>r</span>"</span>\s</span>etsansfont{Times New Roman}"</span>,
</span>        </span>r</span>"</span>\s</span>etmonofont{Times New Roman}"
</span>        </span>r</span>"</span>\s</span>etCJKmainfont{SimSun}"</span>,
</span>        </span>r</span>"</span>\s</span>etCJKsansfont{SimSun}"</span>,
</span>        </span>r</span>"</span>\s</span>etCJKmonofont{SimSun}"
</span>    ],
</span>    </span>"pdf.fonttype"</span>: </span>42</span>,
</span>}
</span>matplotlib.rcParams.update(pgf_with_custom_preamble)
</span></code></pre>
完整的代码如下：</p>
import </span>matplotlib
</span>matplotlib.use(</span>"pgf"</span>)
</span>pgf_with_custom_preamble </span>= </span>{
</span>    </span># "font.size": 40,
</span>    </span>"pgf.rcfonts"</span>: </span>False</span>,
</span>    </span>"text.usetex"</span>: </span>True</span>,
</span>    </span>"pgf.preamble"</span>: [
</span>        </span># math setup:
</span>        </span>r</span>"\usepackage{unicode-math}"</span>,
</span>        </span>r</span>"\usepackage</span>[utf8]</span>{inputenc}"</span>,
</span>        </span>r</span>"\usepackage{xeCJK}"</span>,
</span>        </span>r</span>"</span>\s</span>etmainfont{Times New Roman} % Set main font for ASCII"</span>,
</span>        </span>r</span>"</span>\s</span>etsansfont{Times New Roman}"</span>,
</span>        </span>r</span>"</span>\s</span>etmonofont{Times New Roman}"
</span>        </span>r</span>"</span>\s</span>etCJKmainfont{SimSun}"</span>,
</span>        </span>r</span>"</span>\s</span>etCJKsansfont{SimSun}"</span>,
</span>        </span>r</span>"</span>\s</span>etCJKmonofont{SimSun}"
</span>    ],
</span>    </span>"pdf.fonttype"</span>: </span>42</span>,
</span>}
</span>matplotlib.rcParams.update(pgf_with_custom_preamble)
</span>
</span>import </span>matplotlib.pyplot </span>as </span>plt
</span>import </span>pandas </span>as </span>pd
</span>import </span>seaborn </span>as </span>sns
</span>
</span>sns.set()
</span>df </span>= </span>pd.read_hdf(</span>'./label.h5'</span>)
</span>
</span>## 性别分布
</span>gender </span>= </span>df[</span>'gender'</span>]
</span>
</span>gender </span>= </span>gender.fillna(</span>'unkonwn'</span>)
</span>counts </span>= </span>pd.value_counts(gender)
</span>plt.figure(figsize</span>=</span>(</span>10</span>, </span>8</span>))
</span>counts.plot.pie(legend</span>=</span>True</span>, autopct</span>=</span>lambda </span>p: </span>'</span>{:.2f}</span>%(</span>{:.0f}</span>)'</span>.format(
</span>    p, (p</span>/</span>100</span>)</span>*</span>counts.sum()), title</span>=</span>"Gender distribution"</span>)
</span>plt.savefig(</span>'./figures/gender.png'</span>)
</span>plt.savefig(</span>'./figures/gender'</span>)
</span>
</span>## 年龄分布
</span>
</span>age </span>= </span>df[</span>'age'</span>]
</span>plt.clf()
</span>age </span>= </span>age.dropna()
</span>sns.distplot(age)
</span>plt.title(</span>'kde plot'</span>)
</span>plt.savefig(</span>'./figures/age.png'</span>)
</span>plt.savefig(</span>'./figures/age'</span>)
</span>
</span>## 诊断分布
</span>
</span>tmp </span>= </span>df.iloc[:, </span>2</span>:]
</span>plt.clf()
</span>counts </span>= </span>tmp.sum() </span>/ </span>tmp.shape[</span>0</span>]
</span>counts.plot(kind</span>=</span>'bar'</span>, title</span>=</span>'Percentage'</span>)
</span>plt.tight_layout()
</span>plt.savefig(</span>'./figures/diagnose.png'</span>)
</span>
</span>plt.savefig(</span>'./figures/diagnose'</span>)
</span>
</span></code></pre>
结果如图：</p>
</p>
</p>
</p>


使用docker搭建深度学习环境
2018-06-10T10:18:22+00:00
Docker</h2>
Docker是一种虚拟化容器，可能把运行环境与操作系统隔离起来，使开发环境与操作系统分离，方便多个开发环境的管理。与虚拟机相比，它更加轻量，方便。</p>
系统环境</h2>

Ubuntu 16.04 LTS</li>
Nvidia GTX Titan Xp(其它Nvidia的显卡应该也可以)</li>
</ul>
安装Nvidia驱动（已经安装过的可以跳过）</h2>
可以参照这篇文章</a></p>
sudo add-apt-repository ppa:graphics-drivers/ppa
</span>sudo apt update
</span>sudo apt install nvidia-387
</span></code></pre>
安装Docker-ce和Nvidia-Docker</h2>
安装Docker-ce</h3>
docker-ce的安装可以参照这里的官方文档</a></p>
## 卸载Ubuntu官方源中的docker.io（如果之前安装了的话）
</span>$ sudo apt-get remove docker docker-engine docker.io
</span>## 更新apt package index.
</span>$ sudo apt-get update
</span>## 安装一些要用到了一些包
</span>$ sudo apt-get install \
</span>    apt-transport-https \
</span>    ca-certificates \
</span>    curl \
</span>    software-properties-common
</span>## 添加docker官方的GPG KEY
</span>$ curl -fsSL https://download.docker.com/linux/ubuntu/gpg </span>| </span>sudo apt-key add -
</span>## 添加docker官方的源
</span>$ sudo add-apt-repository \
</span>   </span>"deb [arch=amd64] https://download.docker.com/linux/ubuntu \
</span>   $(</span>lsb_release -cs</span>) \
</span>   stable"
</span>## 更新apt package index.
</span>$ sudo apt-get update
</span>## 安装docker-ce
</span>$ sudo apt-get install docker-ce
</span></code></pre>
安装Nvidia-docker</h3>
在docker-ce中并不能使用GPU,要安装nvidia-docker才可以，nvidia-docker的安装可以参照其官方文档</a></p>
## 配置软件源
</span>$ curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey </span>| </span>\
</span>  </span>sudo apt-key add -
</span>$ distribution=$(</span>.</span> /etc/os-release</span>;</span>echo </span>$ID$VERSION_ID)
</span>$ curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list </span>| </span>\
</span>  </span>sudo tee /etc/apt/sources.list.d/nvidia-docker.list
</span>sudo apt-get update
</span>## 安装Nvidia-docker
</span>$ sudo apt-get install nvidia-docker2
</span></code></pre>
配置Docker</h3>
## 启动Docker服务
</span>sudo service docker start
</span>## 将当前用户加入docker group
</span>sudo gpasswd -a $USER docker
</span></code></pre>
优化Dockerhub连接(可选)</h3>
Dockerhub</a>是一个托管docker镜像的地方，就像Github</a>是一个托管代码的地方一样。上面有大量别人已经配置好的docker环境可以直接使用，十分方便。但是在国内访问十分缓慢。</p>
可以参照这个博客</a>使用阿里云容器hub对Dockerhub进行加速，阿里云容器hub的加速是免费的，实测在学校能达到大概1M/s，而直接连接Dockerhub的速度则惨不忍睹。</p>
</p>
复制上面的专属加速地层地址后编辑/etc/docker//etc/docker/daemon.json</code>
在后面加入如下一行(注意json的格式，要写在{}</code>里面，上一行结尾要有,</code>)：</p>
"registry-mirrosrs"</span>:[</span>"刚才复制的地址"</span>]
</span></code></pre>
</p>
至此Docker和Nvidia环境安装配置完成</p>
使用docker</h2>
简单使用</h3>
一些常用的镜像可以直接在dockerhub上搜索，然后通过dcoker运行。比如下面的这个Tensorflow</a>, 通过下面的这个命令即可开箱及用。</p>
nvidia-docker run -it -p 8888:8888 tensorflow/tensorflow:latest-gpu
</span></code></pre>
使用Nvidia提供的cuda&cudnn环境镜像</h3>
如果在电脑上安装多个版本的cuda</code>、cudnn</code>等库并且管理其版本是一件很麻烦的事情。有了docker之后可以使用Nvidia官方提供的docker镜像，在这些镜像中cuda和cudnn都是已经安装配置好了的。如果dockerhub上面没有我们想要的一些环境的话，可以在这个环境的基础上继续构建我们的环境。</p>
以8.0-cudnn5-devel-ubuntu16.04</code>为例：</p>
nvidia-docker run --name test -it nvidia/cuda:8.0-cudnn5-devel-ubuntu16.04 bash
</span></code></pre>
如下图片所示可以进入对应的环境，然后可以像正常使用Linux一样继续构建环境。</p>
</p>
要退出环境的话，输入exit</code>或者使用快捷键ctrl=d</code>就可以了。</p>
配置好环境退出后后使用可以使用docker commit</code>将container</code>保存成image</code>从而使环境持久化，方便生成同样的环境。</p>
$ docker commit test test:test
</span>sha256:e9338a3b4aede076ab566c44b1127f5c1b18fd3dcb7cb3e02e19856f21db0c5b
</span></code></pre>
可以通过下面的命令新建一个同样的环境</p>
nvidia-docker run -it --name test01 test:test bash
</span></code></pre>
如果只是想运行刚才的container</code>的话，可以用docker container start</code>命令。</p>
更加详细的使用方法可以参照Docker的官方文档</a>和一些中文教程。</p>
使用Dockerfile</h3>
Dockerfile可以用来自动化构建Docker Image.</p>
构建caffe-sal</h4>
参考caffe官方的Dockerfile</a>编写自己的Docerfile</p>
目录结构</p>
</p>
Dockerfile的内容</p>
FROM</span> nvidia/cuda
</span>
</span>RUN </span>apt-get update && apt-get install -y --no-install-recommends \
</span> cpio \
</span>        build-essential \
</span>        git \
</span>        wget \
</span>        numactl \
</span>        vim \
</span>        libopenblas-dev\
</span>        screen \
</span>        libmlx4-1 libmlx5-1 ibutils  rdmacm-utils libibverbs1 ibverbs-utils perftest infiniband-diags \
</span>        openmpi-bin libopenmpi-dev \
</span>        libboost-all-dev \
</span>        libgflags-dev \
</span>        libgoogle-glog-dev \
</span>        libhdf5-serial-dev \
</span>        libleveldb-dev \
</span>        liblmdb-dev \
</span>        libopencv-dev \
</span>        libprotobuf-dev \
</span>        libsnappy-dev \
</span>        protobuf-compiler
</span>
</span>ENV </span>CAFFE_ROOT=/opt/caffe
</span>WORKDIR </span>$CAFFE_ROOT
</span>
</span>ADD </span>caffe-sal caffe-sal
</span>WORKDIR </span>$CAFFE_ROOT/caffe-sal
</span>ADD </span>Makefile.config Makefile.config
</span>
</span>RUN </span>make -j40
</span></code></pre>
Makefile.config的内容</p>
### Refer to http://caffe.berkeleyvision.org/installation.html
</span>## Contributions simplifying and improving our build system are welcome!
</span>
</span>## cuDNN acceleration switch (uncomment to build with cuDNN).
</span>## USE_CUDNN := 1
</span>
</span>## CPU-only switch (uncomment to build without GPU support).
</span>## CPU_ONLY := 1
</span>
</span>## To customize your choice of compiler, uncomment and set the following.
</span>## N.B. the default for Linux is g++ and the default for OSX is clang++
</span>## CUSTOM_CXX := g++
</span>
</span>## CUDA directory contains bin/ and lib/ directories that we need.
</span>CUDA_DIR </span>:= </span>/usr/local/cuda
</span>## On Ubuntu 14.04, if cuda tools are installed via
</span>## "sudo apt-get install nvidia-cuda-toolkit" then use this instead:
</span>## CUDA_DIR := /usr
</span>
</span>## CUDA architecture setting: going with all of them.
</span>## For CUDA < 6.0, comment the *_50 lines for compatibility.
</span>CUDA_ARCH </span>:= </span>-gencode arch=compute_30,code=sm_30 \
</span>  -gencode arch=compute_35,code=sm_35 \
</span>  -gencode arch=compute_50,code=sm_50 \
</span>  -gencode arch=compute_50,code=compute_50
</span>
</span>## BLAS choice:
</span>## atlas for ATLAS (default)
</span>## mkl for MKL
</span>## open for OpenBlas
</span>BLAS </span>:= </span>open
</span>## Custom (MKL/ATLAS/OpenBLAS) include and lib directories.
</span>## Leave commented to accept the defaults for your choice of BLAS
</span>## (which should work)!
</span>## BLAS_INCLUDE := /path/to/your/blas
</span>## BLAS_LIB := /path/to/your/blas
</span>
</span>## Homebrew puts openblas in a directory that is not on the standard search path
</span>## BLAS_INCLUDE := $(shell brew --prefix openblas)/include
</span>## BLAS_LIB := $(shell brew --prefix openblas)/lib
</span>
</span>## This is required only if you will compile the matlab interface.
</span>## MATLAB directory should contain the mex binary in /bin.
</span>## MATLAB_DIR := /usr/local
</span>## MATLAB_DIR := /Applications/MATLAB_R2012b.app
</span>
</span>## NOTE: this is required only if you will compile the python interface.
</span>## We need to be able to find Python.h and numpy/arrayobject.h.
</span>PYTHON_INCLUDE </span>:= </span>/usr/include/python2.7 \
</span>  /usr/lib/python2.7/dist-packages/numpy/core/include
</span>## Anaconda Python distribution is quite popular. Include path:
</span>## Verify anaconda location, sometimes it's in root.
</span>## ANACONDA_HOME := $(HOME)/anaconda
</span>## PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
</span>  # $(ANACONDA_HOME)/include/python2.7 \
</span>  # $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include \
</span>
</span>## We need to be able to find libpythonX.X.so or .dylib.
</span>PYTHON_LIB </span>:= </span>/usr/lib
</span>## PYTHON_LIB := $(ANACONDA_HOME)/lib
</span>
</span>## Homebrew installs numpy in a non standard path (keg only)
</span>## PYTHON_INCLUDE += $(dir $(shell python -c 'import numpy.core; print(numpy.core.__file__)'))/include
</span>## PYTHON_LIB += $(shell brew --prefix numpy)/lib
</span>
</span>## Uncomment to support layers written in Python (will link against Python libs)
</span>## WITH_PYTHON_LAYER := 1
</span>
</span>## Whatever else you find you need goes here.
</span>INCLUDE_DIRS </span>:= </span>$(</span>PYTHON_INCLUDE</span>) ${</span>HOME</span>}</span>/.local/include /usr/local/include /usr/include/hdf5/serial/
</span>LIBRARY_DIRS </span>:= </span>$(</span>PYTHON_LIB</span>) ${</span>HOME</span>}</span>/.local/lib /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu/hdf5/serial/
</span>
</span>## If Homebrew is installed at a non standard location (for example your home directory) and you use it for general dependencies
</span>## INCLUDE_DIRS += $(shell brew --prefix)/include
</span>## LIBRARY_DIRS += $(shell brew --prefix)/lib
</span>
</span>## Uncomment to use `pkg-config` to specify OpenCV library paths.
</span>## (Usually not necessary -- OpenCV libraries are normally installed in one of the above $LIBRARY_DIRS.)
</span>## USE_PKG_CONFIG := 1
</span>
</span>BUILD_DIR </span>:= </span>build
</span>DISTRIBUTE_DIR </span>:= </span>distribute
</span>
</span>## Uncomment for debugging. Does not work on OSX due to https://github.com/BVLC/caffe/issues/171
</span>## DEBUG := 1
</span>
</span>## The ID of the GPU that 'make runtest' will use to run unit tests.
</span>TEST_GPUID </span>:= </span>0
</span>
</span>## enable pretty build (comment to see full commands)
</span>Q </span>?= </span>@
</span>
</span></code></pre>
## 运行构建命令
</span>$ docker build --rm -t caffe-sal:latest .
</span></code></pre>
构建结果：</p>
</p>
运行构建好的Image</code>:</p>
nvidia-docker run it caffe-sal:latest bash
</span></code></pre>
运行结果</p>
</p>


编译caffe时使用本地的protobuf
2018-06-08T14:37:46+00:00
下载并编译protobuf</h2>
Protobuf的项目主页可以在github</a>上找到。在Release页面</a>可以下载对应的代码。下载protobuf-cpp-3.5.1.tar.gz</a>，别的版本应该也可以。</p>
wget https://github.com/google/protobuf/releases/download/v3.5.1/protobuf-cpp-3.5.1.tar.gz
</span>tar -xvf protobuf-cpp-3.5.1.tar.gz
</span>cd</span> protobuf-3.5.1
</span>./configure --prefix</span>=</span>"$</span>HOME</span>/.local"
</span>make -j20
</span>make install
</span></code></pre>
编译caffe</h2>
unrar x caffe-sal.rar
</span>cd</span> caffe-sal
</span>cp Makefile.config.example Makefile.config
</span></code></pre>
编辑Makefile.config</code></p>
修改</p>
INCLUDE_DIRS </span>:= </span>$(</span>PYTHON_INCLUDE</span>)</span> /usr/local/include
</span>LIBRARY_DIRS </span>:= </span>$(</span>PYTHON_LIB</span>)</span> /usr/local/lib /usr/lib
</span></code></pre>
为</p>
INCLUDE_DIRS </span>:= </span>$(</span>PYTHON_INCLUDE</span>) ${</span>HOME</span>}</span>/.local/include /usr/local/include /usr/include/hdf5/serial/
</span>LIBRARY_DIRS </span>:= </span>$(</span>PYTHON_LIB</span>) ${</span>HOME</span>}</span>/.local/lib /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu/hdf5/serial/
</span></code></pre>
设置PATH</code>环境变量，因为要使用刚才编译好的protoc</code>,然后进行编译,j后面的数字是要使用的线程数，一般为cpu的线程数+1,只影响编译的速度</p>
$ export PATH=${HOME}/.local/bin:$PATH
</span>$ which protoc
</span>/home/ncepu/.local/bin/protoc
</span>$ make -j41
</span></code></pre>
每次运行程序前设置环境变量LD_LIBRARY_PATH</code>,或者把它加到.bashrc里面去。</p>
export </span>LD_LIBRARY_PATH</span>=</span>${</span>HOME</span>}/.local/lib:$</span>LD_LIBRARY_PATH
</span></code></pre>
查看链接库的版本</h2>
用下面的命令查看caffe链接的库的版本，可以看到是刚才编译好的版本。</p>
ldd build/tools/caffe </span>| </span>grep protobuf
</span>libprotobuf.so.15 =</span>></span> /home/ncepu/.local/lib/libprotobuf.so.15 (0x00007f7795fc1000</span>)
</span></code></pre>


Windows下载关联python
2018-05-01T18:30:37+00:00
安装anaconda</h2>
下载anaconda</h3>
可以从清华大学开源镜像站</a>下载, 比从官方下载要快上很多。</p>
wget https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/Anaconda3-5.1.0-Windows-x86_64.exe
</span></code></pre>
运行下载的文件完成安装。</p>
设置环境变量</h2>

右键此电脑</code>，选择属性</code>, 高级系统设置</code>, 环境变量</code>。</li>
新建立如下环境变量CONDA_HOME</code>为anaconda的安装目录。</li>
编辑环境变量path</code>, 加入%CONDA_HOME%\</code>和%CONDA_HOME%\Scripts</code>。</li>
</ul>
</p>
</p>
下载winpython</h2>

从github上下载Winpython</a></li>
修改winpython/associate.py</li>
注释unregister(sys.prefix)</code>这一行</li>
运行python -m winpython.associate</code></li>
</ul>
git clone https://github.com/winpython/winpython.git
</span>cd</span> winpython
</span>sed -i </span>'s/unregister(sys.prefix)/#register(sys.prefix)/g'</span> winpython/associate.py
</span>python -m winpython.associate
</span></code></pre>
</p>
修改注册表</h2>
打开注册表编辑器，查找python.File</code>，修改</p>
计算机\HKEY_CURRENT_USER\Software\Classes\Python.File\shell\Edit with Spyder\command</code></p>
的值：</p>
"D:\ProgramData\Miniconda3\pythonw.exe" "D:\ProgramData\Miniconda3\Scripts\spyder" "%1"</code></p>
为：</p>
"D:\ProgramData\Miniconda3\pythonw.exe" "D:\ProgramData\Miniconda3\Scripts\spyder-script.py" "%1"</code></p>
</p>
A simple test for zola even theme

Introduction</h2> This is my first post, which is a post to test zola.</p>

Math</h2> Inline math: \(e = mc^2\)</p> Display math: $$e = mc^2$$</p>

Images</h2> </p>

在matplotlib中使用latex渲染

使用docker搭建深度学习环境

Docker</h2> Docker是一种虚拟化容器，可能把运行环境与操作系统隔离起来，使开发环境与操作系统分离，方便多个开发环境的管理。与虚拟机相比，它更加轻量，方便。</p>

安装Docker-ce和Nvidia-Docker</h2>

使用docker</h2>

编译caffe时使用本地的protobuf

Windows下载关联python

安装anaconda</h2>

设置环境变量</h2> 右键此电脑</code>，选择属性</code>, 高级系统设置</code>, 环境变量</code>。</li> 新建立如下环境变量CONDA_HOME</code>为anaconda的安装目录。</li>

Introduction</h2>
This is my first post, which is a post to test zola.</p>

Math</h2>
Inline math: \(e = mc^2\)</p>
Display math: $$e = mc^2$$</p>

Images</h2>
</p>

Docker</h2>
Docker是一种虚拟化容器，可能把运行环境与操作系统隔离起来，使开发环境与操作系统分离，方便多个开发环境的管理。与虚拟机相比，它更加轻量，方便。</p>
