atom.xml

<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>一只羊的碎碎念</title>
  
  
  <link href="https://www.holmeyoung.com/atom.xml" rel="self"/>
  
  <link href="https://www.holmeyoung.com/"/>
  <updated>2022-10-06T07:36:56.304Z</updated>
  <id>https://www.holmeyoung.com/</id>
  
  <author>
    <name>福尔摩小阳</name>
    
  </author>
  
  <generator uri="https://hexo.io/">Hexo</generator>
  
  <entry>
    <title>python 多线程与多进程</title>
    <link href="https://www.holmeyoung.com/blog/python-threading-multiprocessing/"/>
    <id>https://www.holmeyoung.com/blog/python-threading-multiprocessing/</id>
    <published>2019-08-27T15:24:15.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h2 id="多进程"><a href="#多进程" class="headerlink" title="多进程"></a>多进程</h2><h3 id="进程池阻塞与非阻塞"><a href="#进程池阻塞与非阻塞" class="headerlink" title="进程池阻塞与非阻塞"></a>进程池阻塞与非阻塞</h3><p><span id="jump"></span></p><h4 id="异步进程池（非阻塞）"><a href="#异步进程池（非阻塞）" class="headerlink" title="异步进程池（非阻塞）"></a>异步进程池（非阻塞）</h4><figure class="highlight python"><table><tr><td class="code"><pre><span class="line"><span class="keyword">from</span> multiprocessing <span class="keyword">import</span> Pool</span><br><span class="line"><span class="function"><span class="keyword">def</span> <span class="title">test</span><span class="params">(i)</span>:</span></span><br><span class="line">    <span class="keyword">print</span> (i)</span><br><span class="line"><span class="keyword">if</span> __name__==<span class="string">"__main__"</span>:</span><br><span class="line">pool = Pool(processes=<span class="number">10</span>)</span><br><span class="line"><span class="keyword">for</span> i  <span class="keyword">in</span> xrange(<span class="number">500</span>):</span><br><span class="line"><span class="string">'''</span></span><br><span class="line"><span class="string">For循环中执行步骤：</span></span><br><span class="line"><span class="string">（1）循环遍历，将500个子进程添加到进程池（相对父进程会阻塞）</span></span><br><span class="line"><span class="string">（2）每次执行10个子进程，等一个子进程执行完后，立马启动新的子进程。（相对父进程不阻塞）</span></span><br><span class="line"><span class="string"></span></span><br><span class="line"><span class="string">apply_async为异步进程池写法。</span></span><br><span class="line"><span class="string">异步指的是启动子进程的过程，与父进程本身的执行（print）是异步的，而For循环中往进程池添加子进程的过程，与父进程本身的执行却是同步的。</span></span><br><span class="line"><span class="string">'''</span></span><br><span class="line">    pool.apply_async(test, args=(i,)) <span class="comment">#维持执行的进程总数为10，当一个进程执行完后启动一个新进程.       </span></span><br><span class="line"><span class="keyword">print</span> (<span class="string">'test'</span>)</span><br><span class="line">pool.close()</span><br><span class="line">pool.join()</span><br></pre></td></tr></table></figure><p>执行顺序：For循环内执行了2个步骤，第一步：将500个对象放入进程池（阻塞）。第二步：同时执行10个子进程（非阻塞），有结束的就立即添加，维持10个子进程运行。（apply_async方法的会在执行完for循环的添加步骤后，直接执行后面的print语句，而apply方法会等所有进程池中的子进程运行完以后再执行后面的print语句）</p><p>注意：调用join之前，先调用close或者terminate方法，否则会出错。执行完close后不会有新的进程加入到pool,join函数等待所有子进程结束。</p><h4 id="同步进程池（阻塞）"><a href="#同步进程池（阻塞）" class="headerlink" title="同步进程池（阻塞）"></a>同步进程池（阻塞）</h4><figure class="highlight python"><table><tr><td class="code"><pre><span class="line"><span class="keyword">from</span> multiprocessing <span class="keyword">import</span> Pool</span><br><span class="line"><span class="function"><span class="keyword">def</span> <span class="title">test</span><span class="params">(p)</span>:</span></span><br><span class="line">       <span class="keyword">print</span> (p)</span><br><span class="line">       time.sleep(<span class="number">3</span>)</span><br><span class="line"><span class="keyword">if</span> __name__==<span class="string">"__main__"</span>:</span><br><span class="line">pool = Pool(processes=<span class="number">10</span>)</span><br><span class="line"><span class="keyword">for</span> i  <span class="keyword">in</span> xrange(<span class="number">500</span>):</span><br><span class="line"><span class="string">'''</span></span><br><span class="line"><span class="string">实际测试发现，for循环内部执行步骤：</span></span><br><span class="line"><span class="string">（1）遍历500个可迭代对象，往进程池放一个子进程</span></span><br><span class="line"><span class="string">（2）执行这个子进程，等子进程执行完毕，再往进程池放一个子进程，再执行。（同时只执行一个子进程）</span></span><br><span class="line"><span class="string">for循环执行完毕，再执行print函数。</span></span><br><span class="line"><span class="string">'''</span></span><br><span class="line">    pool.apply(test, args=(i,))   <span class="comment">#维持执行的进程总数为10，当一个进程执行完后启动一个新进程.</span></span><br><span class="line"><span class="keyword">print</span> (<span class="string">'test'</span>)</span><br><span class="line">pool.close()</span><br><span class="line">pool.join()</span><br></pre></td></tr></table></figure><p>说明：for循环内执行的步骤顺序，往进程池中添加一个子进程，执行子进程，等待执行完毕再添加一个子进程…..等500个子进程都执行完了，再执行print (‘test’)。（从结果来看，并没有多进程并发）</p><h4 id="REF"><a href="#REF" class="headerlink" title="REF"></a>REF</h4><blockquote><p><a href="https://thief.one/2016/11/24/Multiprocessing-Pool/" target="_blank" rel="noopener">https://thief.one/2016/11/24/Multiprocessing-Pool/</a></p></blockquote><h3 id="多进程pool"><a href="#多进程pool" class="headerlink" title="多进程pool"></a>多进程pool</h3><p>见 <a href="#jump">多进程.进程池阻塞与非阻塞</a></p><h3 id="多进程pool-queue"><a href="#多进程pool-queue" class="headerlink" title="多进程pool queue"></a>多进程pool queue</h3><figure class="highlight python"><table><tr><td class="code"><pre><span class="line"><span class="keyword">import</span> os</span><br><span class="line"><span class="keyword">from</span> multiprocessing <span class="keyword">import</span> Manager, Pool</span><br><span class="line"><span class="keyword">import</span> sys</span><br><span class="line"></span><br><span class="line"><span class="function"><span class="keyword">def</span> <span class="title">get_hash_queue</span><span class="params">(i, n, image_path, q_put)</span>:</span></span><br><span class="line">    <span class="comment"># print the doing process</span></span><br><span class="line">    sys.stdout.write(<span class="string">"\rLoading process: &#123;&#125;/&#123;&#125;"</span>.format(str(n-i), str(n)))</span><br><span class="line">    sys.stdout.flush()</span><br><span class="line">    <span class="string">'''Dealwith image_path here.'''</span></span><br><span class="line">    q_put.put_nowait([image_path])</span><br><span class="line">    <span class="keyword">return</span></span><br><span class="line"></span><br><span class="line"><span class="keyword">if</span> __name__ == <span class="string">"__main__"</span>:</span><br><span class="line">    <span class="comment"># init pool and queue</span></span><br><span class="line">    q_get = Manager().Queue()</span><br><span class="line">    q_put = Manager().Queue()</span><br><span class="line">    pool = Pool(processes = <span class="number">4</span>)</span><br><span class="line"></span><br><span class="line">    <span class="comment"># put pic path into get queue: q_get</span></span><br><span class="line">    pics = os.listdir(folder)</span><br><span class="line">    <span class="keyword">for</span> pic <span class="keyword">in</span> pics:</span><br><span class="line">        path = os.path.join(folder, pic)</span><br><span class="line">        q_get.put(path)</span><br><span class="line">    </span><br><span class="line">    <span class="comment"># use multi process to get hash result and put the result into q_put</span></span><br><span class="line">    n = q_get.qsize()</span><br><span class="line">    <span class="keyword">while</span> <span class="literal">True</span>:</span><br><span class="line">        <span class="keyword">try</span>:</span><br><span class="line">            image_path = q_get.get_nowait()</span><br><span class="line">            i = q_get.qsize()</span><br><span class="line">        <span class="keyword">except</span>:</span><br><span class="line">            <span class="keyword">break</span></span><br><span class="line">        <span class="keyword">else</span>:</span><br><span class="line">            pool.apply_async(get_hash_queue, (i, n, image_path, q_put, ))</span><br><span class="line">    pool.close()</span><br><span class="line">    pool.join()</span><br><span class="line">    <span class="string">'''Dealwith q_put here.'''</span></span><br></pre></td></tr></table></figure><h3 id="多进程pool-map"><a href="#多进程pool-map" class="headerlink" title="多进程pool.map"></a>多进程pool.map</h3><figure class="highlight python"><table><tr><td class="code"><pre><span class="line"><span class="keyword">import</span> os</span><br><span class="line"><span class="keyword">import</span> cv2</span><br><span class="line"><span class="keyword">import</span> shutil</span><br><span class="line"><span class="keyword">from</span> multiprocessing <span class="keyword">import</span> Pool</span><br><span class="line"><span class="keyword">import</span> argparse</span><br><span class="line"></span><br><span class="line">pic_suffix = set()</span><br><span class="line">pic_suffix.update([<span class="string">'.BMP'</span>, <span class="string">'.DIB'</span>, <span class="string">'.JPEG'</span>, <span class="string">'.JPG'</span>, <span class="string">'.JPE'</span>, <span class="string">'.PNG'</span>, <span class="string">'.PBM'</span>, <span class="string">'.PGM'</span>, <span class="string">'.PPM'</span>, <span class="string">'.SR'</span>, <span class="string">'.RAS'</span>, <span class="string">'.TIFF'</span>, <span class="string">'.TIF'</span>, <span class="string">'.EXR'</span>, <span class="string">'.JP2'</span>])</span><br><span class="line"></span><br><span class="line"><span class="function"><span class="keyword">def</span> <span class="title">image_judge</span><span class="params">(image_path)</span>:</span></span><br><span class="line">    <span class="keyword">try</span>:</span><br><span class="line">        new_path = os.path.join(<span class="string">'./trash'</span>, os.path.split(image_path)[<span class="number">-1</span>])</span><br><span class="line">        <span class="comment"># not image</span></span><br><span class="line">        <span class="keyword">if</span> os.path.splitext(image_path)[<span class="number">-1</span>].upper() <span class="keyword">not</span> <span class="keyword">in</span> pic_suffix:</span><br><span class="line">            <span class="keyword">print</span> (<span class="string">'Not image %s'</span> % (new_path))</span><br><span class="line">            shutil.move(image_path, <span class="string">'./trash'</span>)</span><br><span class="line">            <span class="keyword">return</span></span><br><span class="line">        </span><br><span class="line">        <span class="comment"># image is none</span></span><br><span class="line">        image = cv2.imread(image_path)</span><br><span class="line">        <span class="keyword">if</span> image <span class="keyword">is</span> <span class="literal">None</span>:</span><br><span class="line">            <span class="keyword">print</span> (<span class="string">'Image none: %s'</span> % (new_path))</span><br><span class="line">            shutil.move(image_path, <span class="string">'./trash'</span>)</span><br><span class="line">            <span class="keyword">return</span></span><br><span class="line">    <span class="keyword">except</span>:</span><br><span class="line">        <span class="comment"># catch except                                            </span></span><br><span class="line">        <span class="keyword">print</span> (<span class="string">'Image error: %s'</span> % (new_path))</span><br><span class="line">        shutil.move(image_path, <span class="string">'./trash'</span>)</span><br><span class="line">        <span class="keyword">return</span></span><br><span class="line"></span><br><span class="line"><span class="function"><span class="keyword">def</span> <span class="title">entry_function</span><span class="params">(folder)</span>:</span></span><br><span class="line">    os.makedirs(<span class="string">'./trash'</span>, exist_ok=<span class="literal">True</span>)</span><br><span class="line">    pic_list = []</span><br><span class="line">    pics = os.listdir(folder)</span><br><span class="line">    <span class="keyword">for</span> pic <span class="keyword">in</span> pics:</span><br><span class="line">        pic_list.append(os.path.join(folder, pic))</span><br><span class="line">    pool = Pool()</span><br><span class="line">    pool.map(image_judge, pic_list)</span><br><span class="line">    pool.close()</span><br><span class="line">    pool.join()</span><br><span class="line"></span><br><span class="line"><span class="keyword">if</span> __name__ == <span class="string">"__main__"</span>:</span><br><span class="line">    parser = argparse.ArgumentParser()</span><br><span class="line">    parser.add_argument(<span class="string">'--folder'</span>, type = str, required = <span class="literal">True</span>, help = <span class="string">'path to folder which contains the images'</span>)</span><br><span class="line">    args = parser.parse_args()</span><br><span class="line"></span><br><span class="line">    entry_function(args.folder)</span><br></pre></td></tr></table></figure><h2 id="多线程"><a href="#多线程" class="headerlink" title="多线程"></a>多线程</h2><h3 id="多线程queue"><a href="#多线程queue" class="headerlink" title="多线程queue"></a>多线程queue</h3><figure class="highlight python"><table><tr><td class="code"><pre><span class="line"><span class="keyword">import</span> os</span><br><span class="line"><span class="keyword">import</span> threading</span><br><span class="line"><span class="keyword">import</span> queue</span><br><span class="line"></span><br><span class="line"><span class="function"><span class="keyword">def</span> <span class="title">fetch_img_func</span><span class="params">(q)</span>:</span></span><br><span class="line">    <span class="keyword">while</span> <span class="literal">True</span>:</span><br><span class="line">        <span class="keyword">try</span>:</span><br><span class="line">            <span class="comment"># unblock read from queue</span></span><br><span class="line">            url = q.get_nowait() <span class="comment"># https://example.com/example.jpg</span></span><br><span class="line">        <span class="keyword">except</span>:</span><br><span class="line">            <span class="keyword">break</span></span><br><span class="line">        <span class="keyword">else</span>:</span><br><span class="line">            <span class="keyword">try</span>:</span><br><span class="line">                <span class="string">'''do something with url here'''</span></span><br><span class="line"></span><br><span class="line"><span class="keyword">if</span> __name__ == <span class="string">"__main__"</span>:</span><br><span class="line">    q = queue.Queue()</span><br><span class="line"></span><br><span class="line">    <span class="string">'''</span></span><br><span class="line"><span class="string">    Put something into queue here.</span></span><br><span class="line"><span class="string">    eg:</span></span><br><span class="line"><span class="string">        url = https://example.com/example.jpg</span></span><br><span class="line"><span class="string">        q.put(url)</span></span><br><span class="line"><span class="string">    '''</span></span><br><span class="line"></span><br><span class="line">    threads = []</span><br><span class="line">    <span class="keyword">for</span> i <span class="keyword">in</span> range(args.thread):</span><br><span class="line">        thread = threading.Thread(target=fetch_img_func, args=(q, ))</span><br><span class="line">        threads.append(thread)</span><br><span class="line">        thread.start()</span><br><span class="line">    <span class="keyword">for</span> i <span class="keyword">in</span> threads:</span><br><span class="line">        i.join()</span><br></pre></td></tr></table></figure><h3 id="多线程下载"><a href="#多线程下载" class="headerlink" title="多线程下载"></a>多线程下载</h3><p><a href="https://github.com/Holmeyoung/image-process/blob/master/image_download.py" target="_blank" rel="noopener">https://github.com/Holmeyoung/image-process/blob/master/image_download.py</a></p>]]></content>
    
    
      
      
    <summary type="html">&lt;h2 id=&quot;多进程&quot;&gt;&lt;a href=&quot;#多进程&quot; class=&quot;headerlink&quot; title=&quot;多进程&quot;&gt;&lt;/a&gt;多进程&lt;/h2&gt;&lt;h3 id=&quot;进程池阻塞与非阻塞&quot;&gt;&lt;a href=&quot;#进程池阻塞与非阻塞&quot; class=&quot;headerlink&quot; title=&quot;进程池</summary>
      
    
    
    
    <category term="python" scheme="https://www.holmeyoung.com/categories/python/"/>
    
    
    <category term="python" scheme="https://www.holmeyoung.com/tags/python/"/>
    
  </entry>
  
  <entry>
    <title>hexo 通过 gitlab-ci 部署到 github</title>
    <link href="https://www.holmeyoung.com/blog/hexo-gitlab-ci-github/"/>
    <id>https://www.holmeyoung.com/blog/hexo-gitlab-ci-github/</id>
    <published>2019-05-10T11:19:17.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h2 id="目的"><a href="#目的" class="headerlink" title="目的"></a>目的</h2><p>将 <code>Markdown</code> 文件提交到 <code>GitLab</code> ，通过 <code>GitLab</code> 的 <code>ci</code> 工具，自动采用 <code>hexo</code> 将其渲染为 <code>html</code> 文件并部署到 <code>GitHub</code> 。可以真正做到</p><blockquote><p>只维护 <code>Markdown</code> 的博客源文件即可</p><p>提交 <code>Markdown</code> 就能看博客</p><p>源文件、配置文件长久备份</p><p>跨平台、跨终端</p></blockquote><h2 id="GitHub-操作"><a href="#GitHub-操作" class="headerlink" title="GitHub 操作"></a>GitHub 操作</h2><h3 id="新建-Pages-仓库"><a href="#新建-Pages-仓库" class="headerlink" title="新建 Pages 仓库"></a>新建 Pages 仓库</h3><p>假如你的 <code>GitHub</code> 用户名为 <code>xxx</code> ，新建 <strong>共有仓库</strong> <code>xxx.github.io</code></p><h3 id="获取-tokens"><a href="#获取-tokens" class="headerlink" title="获取 tokens"></a>获取 tokens</h3><p>在 <code>GitHub</code> 上获取一个 <strong>只用于</strong> 操作 <code>repo</code> 的 <code>Personal access tokens</code></p><p>具体路径为</p><p><code>Settings</code> – <code>Developer settings</code> – <code>Personal access tokens</code> – <code>Generate new token</code></p><blockquote><p>注意权限勾选 <code>repo</code> 即可</p></blockquote><h2 id="GitLab-操作"><a href="#GitLab-操作" class="headerlink" title="GitLab 操作"></a>GitLab 操作</h2><h3 id="新建自己的-hexo-仓库"><a href="#新建自己的-hexo-仓库" class="headerlink" title="新建自己的 hexo 仓库"></a>新建自己的 hexo 仓库</h3><p>在 <code>gitlab</code> 新建一个自己的 <strong>私有仓库</strong> 。在这里我们假设命名为 <code>blog</code> 。</p><h3 id="clone一份官方-hexo-仓库到本地"><a href="#clone一份官方-hexo-仓库到本地" class="headerlink" title="clone一份官方 hexo 仓库到本地"></a>clone一份官方 hexo 仓库到本地</h3><p>地址：<a href="https://gitlab.com/pages/hexo" target="_blank" rel="noopener">https://gitlab.com/pages/hexo</a></p><p>将其 <code>clone</code> 到本地，删除 <code>.git</code> 文件夹，并将该文件夹初始化为自己在 <code>gitlab新建的仓库</code> 即 <code>blog</code> 。</p><h2 id="文件配置"><a href="#文件配置" class="headerlink" title="文件配置"></a>文件配置</h2><ul><li><p>修改 <code>hexo</code> 的 <code>config.yml</code> 文件</p><figure class="highlight yaml"><table><tr><td class="code"><pre><span class="line"><span class="attr">url:</span> <span class="string">your-blog-url</span></span><br><span class="line"><span class="attr">root:</span> <span class="string">/</span></span><br><span class="line"></span><br><span class="line"><span class="attr">deploy:</span></span><br><span class="line">  <span class="attr">type:</span> <span class="string">git</span></span><br><span class="line">  <span class="attr">repository:</span></span><br><span class="line">    <span class="attr">github:</span> <span class="string">https://personal-access-token:x-oauth-basic@github.com/xxx/xxx.github.io.git</span></span><br><span class="line">  <span class="attr">branch:</span> <span class="string">master</span></span><br></pre></td></tr></table></figure><blockquote><p>将 <code>your-blog-url</code> 替换为自己的博客地址</p><p>把上面从 <code>GitHub</code> 获取的 <code>personal-access-token</code> 放到这里来</p><p>将 <code>xxx</code> 替换为自己的用户名</p></blockquote></li></ul><ul><li><p>修改 <code>.gitlab-ci.yml</code> 文件</p><figure class="highlight yaml"><table><tr><td class="code"><pre><span class="line"><span class="attr">image:</span> <span class="string">node:10.15.3</span></span><br><span class="line"></span><br><span class="line"><span class="attr">cache:</span></span><br><span class="line">  <span class="attr">paths:</span></span><br><span class="line">    <span class="bullet">-</span> <span class="string">node_modules/</span></span><br><span class="line"></span><br><span class="line"><span class="attr">before_script:</span></span><br><span class="line">  <span class="bullet">-</span> <span class="string">git</span> <span class="string">config</span> <span class="string">--global</span> <span class="string">user.name</span> <span class="string">"github user name"</span></span><br><span class="line">  <span class="bullet">-</span> <span class="string">git</span> <span class="string">config</span> <span class="string">--global</span> <span class="string">user.email</span> <span class="string">"gitub email"</span></span><br><span class="line">  <span class="bullet">-</span> <span class="string">npm</span> <span class="string">install</span> <span class="string">-g</span> <span class="string">hexo</span></span><br><span class="line">  <span class="bullet">-</span> <span class="string">npm</span> <span class="string">install</span> <span class="string">-g</span> <span class="string">hexo-cli</span></span><br><span class="line">  <span class="bullet">-</span> <span class="string">npm</span> <span class="string">install</span></span><br><span class="line">  <span class="bullet">-</span> <span class="string">npm</span> <span class="string">install</span> <span class="string">hexo-deployer-git</span> <span class="string">--save</span></span><br><span class="line"></span><br><span class="line"><span class="attr">pages:</span></span><br><span class="line">  <span class="attr">stage:</span> <span class="string">deploy</span></span><br><span class="line">  <span class="attr">script:</span></span><br><span class="line">    <span class="bullet">-</span> <span class="string">hexo</span> <span class="string">clean</span></span><br><span class="line">    <span class="bullet">-</span> <span class="string">hexo</span> <span class="string">generate</span></span><br><span class="line">    <span class="bullet">-</span> <span class="string">hexo</span> <span class="string">deploy</span></span><br><span class="line">  <span class="attr">artifacts:</span></span><br><span class="line">    <span class="attr">paths:</span></span><br><span class="line">      <span class="bullet">-</span> <span class="string">public</span></span><br><span class="line">  <span class="attr">only:</span></span><br><span class="line">    <span class="bullet">-</span> <span class="string">master</span></span><br></pre></td></tr></table></figure></li></ul><h2 id="GitHub域名"><a href="#GitHub域名" class="headerlink" title="GitHub域名"></a>GitHub域名</h2><p>在 <code>source</code> 文件夹新建文件 <code>CNAME</code> 并写入你的域名，例如</p><figure class="highlight plain"><table><tr><td class="code"><pre><span class="line">www.holmeyoung.com</span><br></pre></td></tr></table></figure><h2 id="总览"><a href="#总览" class="headerlink" title="总览"></a>总览</h2><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">$ tree -L 2       </span><br><span class="line">.</span><br><span class="line">├── _config.yml</span><br><span class="line">├── package.json</span><br><span class="line">├── scaffolds</span><br><span class="line">│   ├── draft.md</span><br><span class="line">│   ├── page.md</span><br><span class="line">│   └── post.md</span><br><span class="line">├── <span class="built_in">source</span></span><br><span class="line">│   ├── about</span><br><span class="line">│   ├── categories</span><br><span class="line">│   ├── CNAME</span><br><span class="line">│   ├── _posts</span><br><span class="line">│   └── tags</span><br><span class="line">└── themes</span><br><span class="line">    └── next</span><br></pre></td></tr></table></figure><p>现在只需要写好 <code>markdown</code> 文件，放到 <code>_posts</code> 文件夹，提交到 <code>GitLab</code> ，稍等编译结束，访问你的博客域名就能看见啦</p>]]></content>
    
    
      
      
    <summary type="html">&lt;h2 id=&quot;目的&quot;&gt;&lt;a href=&quot;#目的&quot; class=&quot;headerlink&quot; title=&quot;目的&quot;&gt;&lt;/a&gt;目的&lt;/h2&gt;&lt;p&gt;将 &lt;code&gt;Markdown&lt;/code&gt; 文件提交到 &lt;code&gt;GitLab&lt;/code&gt; ，通过 &lt;code&gt;GitLab&lt;/co</summary>
      
    
    
    
    <category term="安装记录" scheme="https://www.holmeyoung.com/categories/安装记录/"/>
    
    
    <category term="hexo" scheme="https://www.holmeyoung.com/tags/hexo/"/>
    
  </entry>
  
  <entry>
    <title>centos7 配置 redis 主从集群 + 哨兵模式</title>
    <link href="https://www.holmeyoung.com/blog/centos7-redis-cluster-sentinel/"/>
    <id>https://www.holmeyoung.com/blog/centos7-redis-cluster-sentinel/</id>
    <published>2019-04-11T04:21:44.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h2 id="搭建环境"><a href="#搭建环境" class="headerlink" title="搭建环境"></a>搭建环境</h2><p>三台CentOS7服务器</p><h2 id="实现目标"><a href="#实现目标" class="headerlink" title="实现目标"></a>实现目标</h2><p>在3台CentOS7服务器上搭建一主两从的redis集群，且配置哨兵模式，当master宕机，另外两个slave会选举出一个新的master，实现高可用。</p><table><thead><tr><th>master</th><th>slave0</th><th>slave1</th></tr></thead><tbody><tr><td>ip0</td><td>ip1</td><td>ip2</td></tr></tbody></table><h2 id="准备工作"><a href="#准备工作" class="headerlink" title="准备工作"></a>准备工作</h2><h3 id="安装额外的软件包"><a href="#安装额外的软件包" class="headerlink" title="安装额外的软件包"></a>安装额外的软件包</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">$ yum install epel-release</span><br></pre></td></tr></table></figure><h3 id="安装-redis"><a href="#安装-redis" class="headerlink" title="安装 redis"></a>安装 redis</h3><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> yum install redis</span></span><br></pre></td></tr></table></figure><h2 id="配置一主两从"><a href="#配置一主两从" class="headerlink" title="配置一主两从"></a>配置一主两从</h2><h3 id="master（ip0）"><a href="#master（ip0）" class="headerlink" title="master（ip0）"></a>master（ip0）</h3><ul><li><p>修改 <code>/etc/redis.conf</code></p>  <figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">$ vim /etc/redis.conf</span><br></pre></td></tr></table></figure><p>  找到一下内容并修改</p>  <figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="comment"># 将 127.0.0.1 修改为本机的IP（ip0）</span></span><br><span class="line"><span class="built_in">bind</span> ip0</span><br><span class="line"></span><br><span class="line"><span class="comment"># 将 no 改为 yes，以守护进程的形式运行</span></span><br><span class="line">daemonize yes</span><br></pre></td></tr></table></figure><p>  修改完毕</p></li></ul><ul><li><p>测试</p><p>  登录 <code>redis</code>  </p>  <figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-cli -h ip0</span></span><br></pre></td></tr></table></figure><p>  列出所有 <code>kyes *</code></p>  <figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-cli -h ip0</span></span><br><span class="line">ip0:6379&gt; keys *</span><br></pre></td></tr></table></figure><p>  设置 <code>key</code> <code>value</code></p>  <figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-cli -h ip0</span></span><br><span class="line">ip0:6379&gt; set name holmeyoung</span><br></pre></td></tr></table></figure><p>  得到 <code>key</code> 的 <code>value</code></p>  <figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-cli -h ip0</span></span><br><span class="line">ip0:6379&gt; get name</span><br></pre></td></tr></table></figure><p>  删除 <code>key</code> </p>  <figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-cli -h ip0</span></span><br><span class="line">ip0:6379&gt; del name</span><br></pre></td></tr></table></figure></li></ul><h3 id="slave（ip1、ip2）"><a href="#slave（ip1、ip2）" class="headerlink" title="slave（ip1、ip2）"></a>slave（ip1、ip2）</h3><ul><li><p>修改 <code>/etc/redis.conf</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> vim /etc/redis.conf</span></span><br></pre></td></tr></table></figure><p>找到一下内容并修改</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="comment"># 将 127.0.0.1 修改为本机的IP（eg：ip1）</span></span><br><span class="line"><span class="built_in">bind</span> ip1</span><br><span class="line"></span><br><span class="line"><span class="comment"># 将 no 改为 yes，以守护进程的形式运行</span></span><br><span class="line">daemonize yes</span><br><span class="line"></span><br><span class="line"><span class="comment"># 增加该 slave节点 所要从的 master节点 的 ip/port</span></span><br><span class="line"><span class="comment"># 找到 # slaveof &lt;masterip&gt; &lt;masterport&gt;</span></span><br><span class="line"><span class="comment"># 在下面增加一行</span></span><br><span class="line">slaveof ip0 6379</span><br></pre></td></tr></table></figure><p>修改完毕</p></li></ul><ul><li><p>测试</p><p>同 <code>master</code> </p></li></ul><h3 id="启动"><a href="#启动" class="headerlink" title="启动"></a>启动</h3><p>分别运行三台机器的 <code>redis-server</code> </p><blockquote><p>先运行 <code>master</code> </p></blockquote><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-server /etc/redis.conf</span></span><br></pre></td></tr></table></figure><h3 id="测试"><a href="#测试" class="headerlink" title="测试"></a>测试</h3><ul><li><p>在 <code>master</code> 上</p><p>连接 <code>redis</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-cli -h ip0</span></span><br></pre></td></tr></table></figure><p>打印信息</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">ip0:6379&gt; info Replication</span><br><span class="line"><span class="meta">#</span><span class="bash"> Replication</span></span><br><span class="line">role:master</span><br><span class="line">connected_slaves:2</span><br><span class="line">slave0:ip=ip1,port=6379,state=online,offset=3242902,lag=1</span><br><span class="line">slave1:ip=ip2,port=6379,state=online,offset=3243317,lag=0</span><br><span class="line">master_repl_offset:3243317</span><br><span class="line">repl_backlog_active:1</span><br><span class="line">repl_backlog_size:1048576</span><br><span class="line">repl_backlog_first_byte_offset:2194742</span><br><span class="line">repl_backlog_histlen:1048576</span><br></pre></td></tr></table></figure><p>写入一组 <code>key</code> <code>value</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">ip0:6379&gt; set name holmeyoung</span><br></pre></td></tr></table></figure></li></ul><ul><li><p>在 <code>slave0</code> 上</p><p>连接 <code>redis</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-cli -h ip1</span></span><br></pre></td></tr></table></figure><p>读取数值</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">ip1:6379&gt; get name</span><br></pre></td></tr></table></figure><p>输出 <code>holmeyoung</code> 则正常</p><blockquote><p>因为 <code>slave</code> 只有读权限，所以是不可以写入数据的</p></blockquote></li></ul><ul><li><p>在 <code>slave1</code> 上</p><p>同 <code>slave0</code> </p></li></ul><h2 id="配置哨兵模式"><a href="#配置哨兵模式" class="headerlink" title="配置哨兵模式"></a>配置哨兵模式</h2><h3 id="配置"><a href="#配置" class="headerlink" title="配置"></a>配置</h3><blockquote><p>三台机器配置相同</p></blockquote><p>修改 <code>/etc/redis-sentinel.conf</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> vim /etc/redis-sentinel.conf</span></span><br></pre></td></tr></table></figure><p>增改以下内容</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">#</span><span class="bash"> 保护模式，这个必须选择no，不然当master宕机，slave无法自动转换为master</span></span><br><span class="line">protected-mode no</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 以守护进程运行</span></span><br><span class="line">daemonize yes</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 日志文件位置</span></span><br><span class="line">logfile "/var/log/redis/sentinel.log"</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 哨兵监控的master，主从配置一样，这里只用输入redis主节点的ip/port和法定人数</span></span><br><span class="line"><span class="meta">#</span><span class="bash"> 其中2是quorum，即权重，代表至少需要两个Sentinel节点认为主节点主观下线，才可判定主节点为客观下线。一般建议将其设置为Sentinel节点的一半加1。不仅如此，quorum还与Sentinel节点的领导者选举有关。为了选出Sentinel的领导者，至少需要max(quorum, num(sentinels) / 2 + 1)个Sentinel节点参与选举。</span></span><br><span class="line">sentinel monitor mymaster ip0 6379 2</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> master或slave多长时间（默认30秒）不能使用后标记为s_down状态</span></span><br><span class="line"><span class="meta">#</span><span class="bash"> 每个Sentinel节点都要通过定期发送ping命令来判断Redis节点和其余Sentinel节点是否可达。</span></span><br><span class="line"><span class="meta">#</span><span class="bash"> 如果在指定的时间内，没有收到主节点的有效回复，则判断其为主观下线。需要注意的是，该参数不仅用来判断主节点状态，同样也用来判断该主节点下面的从节点及其它Sentinel的状态。其默认值为30s。</span></span><br><span class="line">sentinel down-after-milliseconds mymaster 3000</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 在failover期间，允许多少个slave同时指向新的主节点。如果numslaves设置较大的话，虽然复制操作并不会阻塞主节点，但多个节点同时指向新的主节点，会增加主节点的网络和磁盘IO负载</span></span><br><span class="line">sentinel parallel-syncs mymaster 1</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 定义故障切换超时时间。默认180000，单位秒，即3min。需要注意的是，该时间不是总的故障切换的时间，而是适用于故障切换的多个场景。</span></span><br><span class="line"><span class="meta">#</span><span class="bash"> 若sentinel在该配置值内未能完成failover操作（即故障时master/slave自动切换），则认为本次failover失败</span></span><br><span class="line">sentinel failover-timeout mymaster 10000</span><br></pre></td></tr></table></figure><blockquote><p>注意 <code>sentinel monitor mymaster ip0 6379 2</code> 里面的 <code>ip0</code> 为 <code>master</code> 的IP</p></blockquote><h3 id="配置备份"><a href="#配置备份" class="headerlink" title="配置备份"></a>配置备份</h3><blockquote><p>将配置文件 <code>/etc/redis.conf</code> 以及 <code>/etc/redis-sentinel.conf</code> 备份一份，因为当 <code>master</code> 宕机并唤醒哨兵模式后，系统会重写配置文件</p></blockquote><h3 id="启动-1"><a href="#启动-1" class="headerlink" title="启动"></a>启动</h3><p>分别运行三台机器的 <code>redis-sentinel</code> </p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-sentinel /etc/redis-sentinel.conf</span></span><br></pre></td></tr></table></figure><h3 id="测试-1"><a href="#测试-1" class="headerlink" title="测试"></a>测试</h3><p>在 <code>master</code> 上杀死 <code>redis</code> 进程</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> pkill redis</span></span><br></pre></td></tr></table></figure><p>随便登录一个 <code>slave</code> 服务器</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">$</span><span class="bash"> redis-cli -h ip1</span></span><br></pre></td></tr></table></figure><p>打印信息</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">ip0:6379&gt; info Replication</span><br></pre></td></tr></table></figure><p>会发现此时的 <code>master</code> 以及 <code>slave</code> 发生了更改</p><h2 id="Reference"><a href="#Reference" class="headerlink" title="Reference"></a>Reference</h2><p><a href="https://www.cnblogs.com/ddrsql/p/6686282.html" target="_blank" rel="noopener">CentOS7 配置 Redis Sentinel 主从集群配置</a></p><p><a href="https://www.centos.bz/2017/07/centos7-redis-master-slave-replicate/" target="_blank" rel="noopener">CentOS7 配置 Redis 主从复制方法</a></p><h2 id="DONE"><a href="#DONE" class="headerlink" title="DONE"></a>DONE</h2>]]></content>
    
    
      
      
    <summary type="html">&lt;h2 id=&quot;搭建环境&quot;&gt;&lt;a href=&quot;#搭建环境&quot; class=&quot;headerlink&quot; title=&quot;搭建环境&quot;&gt;&lt;/a&gt;搭建环境&lt;/h2&gt;&lt;p&gt;三台CentOS7服务器&lt;/p&gt;
&lt;h2 id=&quot;实现目标&quot;&gt;&lt;a href=&quot;#实现目标&quot; class=&quot;headerli</summary>
      
    
    
    
    <category term="安装记录" scheme="https://www.holmeyoung.com/categories/安装记录/"/>
    
    
    <category term="linux" scheme="https://www.holmeyoung.com/tags/linux/"/>
    
    <category term="redis" scheme="https://www.holmeyoung.com/tags/redis/"/>
    
  </entry>
  
  <entry>
    <title>centos7安装p40显卡驱动 + cuda10.0 + cudnn10.0</title>
    <link href="https://www.holmeyoung.com/blog/centos7-p40-cuda10.0-cudnn10.0/"/>
    <id>https://www.holmeyoung.com/blog/centos7-p40-cuda10.0-cudnn10.0/</id>
    <published>2019-03-13T22:52:46.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h2 id="环境"><a href="#环境" class="headerlink" title="环境"></a>环境</h2><ul><li><p>系统</p><p>CentOS Linux release 7.6.1810 (Core)</p></li><li><p>内核</p><p>3.10.0-957.5.1.el7.x86_64</p></li><li><p>显卡</p><p>NVIDIA Tesla P40</p></li><li><p>cuda</p><p>cuda_10.0.130_410.48_linux.run</p></li><li><p>cudnn</p><p>cudnn-10.0-linux-x64-v7.5.0.56.tgz</p></li></ul><h2 id="显卡驱动"><a href="#显卡驱动" class="headerlink" title="显卡驱动"></a>显卡驱动</h2><h3 id="验证你的系统是否能识别上面的GPU"><a href="#验证你的系统是否能识别上面的GPU" class="headerlink" title="验证你的系统是否能识别上面的GPU"></a>验证你的系统是否能识别上面的GPU</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">lspci | grep -i nvidia</span><br></pre></td></tr></table></figure><p>正常应该显示Nvidia显卡的型号，没有任何显示需要更新pci硬件库<code>update-pciids</code></p><h3 id="验证系统是否是受支持的Linux版本"><a href="#验证系统是否是受支持的Linux版本" class="headerlink" title="验证系统是否是受支持的Linux版本"></a>验证系统是否是受支持的Linux版本</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">uname -m &amp;&amp; cat /etc/*release</span><br></pre></td></tr></table></figure><p>到这里查看受支持的Linux版本：<a href="http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#system-requirements" target="_blank" rel="noopener">http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#system-requirements</a></p><h3 id="验证系统是否有GCC编译环境"><a href="#验证系统是否有GCC编译环境" class="headerlink" title="验证系统是否有GCC编译环境"></a>验证系统是否有GCC编译环境</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">sudo yum install gcc</span><br></pre></td></tr></table></figure><h3 id="验证系统是否安装了正确的内核头文件和开发包"><a href="#验证系统是否安装了正确的内核头文件和开发包" class="headerlink" title="验证系统是否安装了正确的内核头文件和开发包"></a>验证系统是否安装了正确的内核头文件和开发包</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">sudo yum install kernel-devel-$(uname -r) kernel-headers-$(uname -r)</span><br></pre></td></tr></table></figure><h3 id="禁用nouveau"><a href="#禁用nouveau" class="headerlink" title="禁用nouveau"></a>禁用nouveau</h3><p>nouveau是一个第三方开源的Nvidia驱动，一般Linux安装的时候默认会安装这个驱动。 这个驱动会与Nvidia官方的驱动冲突，在安装Nvidia驱动和和CUDA之前应先禁用nouveau</p><p>查看系统是否正在使用nouveau</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">lsmod | grep nouveau</span><br></pre></td></tr></table></figure><p>如果有显示内容，则进行以下的步骤： <strong>Centos7禁用方法</strong></p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="comment">#新建一个配置文件</span></span><br><span class="line">sudo vim /etc/modprobe.d/blacklist.conf</span><br><span class="line"></span><br><span class="line"><span class="comment">#写入以下内容</span></span><br><span class="line">blacklist nouveau</span><br><span class="line">options nouveau modeset=0</span><br><span class="line"></span><br><span class="line"><span class="comment">#保存并退出</span></span><br><span class="line">:wq</span><br><span class="line"></span><br><span class="line"><span class="comment">#备份当前的镜像</span></span><br><span class="line">sudo mv /boot/initramfs-$(uname -r).img /boot/initramfs-$(uname -r).img.bak</span><br><span class="line"></span><br><span class="line"><span class="comment">#建立新的镜像</span></span><br><span class="line">sudo dracut /boot/initramfs-$(uname -r).img $(uname -r)</span><br><span class="line"></span><br><span class="line"><span class="comment">#重启</span></span><br><span class="line">sudo reboot</span><br><span class="line"></span><br><span class="line"><span class="comment">#最后输入上面的命令验证</span></span><br><span class="line">lsmod | grep nouveau</span><br></pre></td></tr></table></figure><p>总之最后一定要保证<code>lsmod | grep nouveau</code>命令没有任何输出内容。 这也是runfile文件安装方式相对rpm/deb方式麻烦的一点，如果本节禁用nouveau有问题，可以尝试rpm/deb方式安装，rpm/deb方式安装会自动禁用</p><h3 id="下载驱动"><a href="#下载驱动" class="headerlink" title="下载驱动"></a>下载驱动</h3><p>首先在官网 <a href="https://www.nvidia.com/Download/index.aspx?lang=cn" target="_blank" rel="noopener">https://www.nvidia.com/Download/index.aspx?lang=cn</a> 查找自己显卡的驱动</p><p>我这里是 <code>NVIDIA Tesla P40</code> ，因为列表中可支持的最新版本cuda是 <code>10.0</code> ，所以下载的cuda为 10.0 的版本：<strong>NVIDIA-Linux-x86_64-410.104.run</strong></p><h3 id="安装驱动"><a href="#安装驱动" class="headerlink" title="安装驱动"></a>安装驱动</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">sudo sh NVIDIA-Linux-x86_64-410.104.run --kernel-source-path=/usr/src/kernels/3.10.0-957.5.1.el7.x86_64</span><br></pre></td></tr></table></figure><blockquote><p>这里必须指明 <code>kernel-source-path</code> ，否则安装过程会报错</p></blockquote><h3 id="检验安装"><a href="#检验安装" class="headerlink" title="检验安装"></a>检验安装</h3><p>执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">nvidia-smi</span><br></pre></td></tr></table></figure><p>如下输出则说明成功</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">Thu Mar 14 18:40:38 2019       </span><br><span class="line">+-----------------------------------------------------------------------------+</span><br><span class="line">| NVIDIA-SMI 410.104      Driver Version: 410.104      CUDA Version: 10.0     |</span><br><span class="line">|-------------------------------+----------------------+----------------------+</span><br><span class="line">| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |</span><br><span class="line">| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |</span><br><span class="line">|===============================+======================+======================|</span><br><span class="line">|   0  Tesla P40           Off  | 00000000:3B:00.0 Off |                    0 |</span><br><span class="line">| N/A   25C    P0    47W / 250W |      0MiB / 22919MiB |      0%      Default |</span><br><span class="line">+-------------------------------+----------------------+----------------------+</span><br><span class="line">                                                                               </span><br><span class="line">+-----------------------------------------------------------------------------+</span><br><span class="line">| Processes:                                                       GPU Memory |</span><br><span class="line">|  GPU       PID   Type   Process name                             Usage      |</span><br><span class="line">|=============================================================================|</span><br><span class="line">|  No running processes found                                                 |</span><br><span class="line">+-----------------------------------------------------------------------------+</span><br></pre></td></tr></table></figure><h2 id="cuda"><a href="#cuda" class="headerlink" title="cuda"></a>cuda</h2><h3 id="下载cuda"><a href="#下载cuda" class="headerlink" title="下载cuda"></a>下载cuda</h3><p>因为 NVIDIA 官网上，P40的cuda最多支持到10.0版本，驱动下载的也是10.0版本，所以这里下载<code>cuda10.0</code></p><p>各版本下载链接 <a href="https://developer.nvidia.com/cuda-toolkit-archive" target="_blank" rel="noopener">https://developer.nvidia.com/cuda-toolkit-archive</a></p><p>我这里下载的是 <code>runfile</code></p><h3 id="安装cuda"><a href="#安装cuda" class="headerlink" title="安装cuda"></a>安装cuda</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">sudo sh cuda_10.0.130_410.48_linux.run --kernel-source-path=/usr/src/kernels/3.10.0-957.5.1.el7.x86_64</span><br></pre></td></tr></table></figure><blockquote><p>这里必须指明 <code>kernel-source-path</code> ，否则安装过程会报错</p></blockquote><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="comment">#刚开始安装会进入more模式，一直按`空格`即可</span></span><br><span class="line"></span><br><span class="line"><span class="comment">#Do you accept the previously read EULA?</span></span><br><span class="line"><span class="comment">#是否接受协议</span></span><br><span class="line">accept</span><br><span class="line"></span><br><span class="line"><span class="comment">#Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 387.26?</span></span><br><span class="line"><span class="comment">#是否安装Nvida驱动，不需要</span></span><br><span class="line">n</span><br><span class="line"></span><br><span class="line"><span class="comment">#Install the CUDA 10.0 Toolkit?</span></span><br><span class="line">y</span><br><span class="line"></span><br><span class="line"><span class="comment">#Enter Toolkit Location</span></span><br><span class="line"><span class="comment">#输入Toolkit的安装目录</span></span><br><span class="line"><span class="comment">#一般默认即可</span></span><br><span class="line"><span class="comment">#如果更改的话，注意是不可以改成 /usr/local/cuda 的，其余都可以</span></span><br><span class="line">回车</span><br><span class="line"></span><br><span class="line"><span class="comment">#Do you want to install a symbolic link at /usr/local/cuda?</span></span><br><span class="line"><span class="comment">#创建一个软连接，我选择是</span></span><br><span class="line">y</span><br><span class="line"></span><br><span class="line"><span class="comment">#Install the CUDA 10.0 Samples?</span></span><br><span class="line"><span class="comment">#安装CUDA官方示例包</span></span><br><span class="line"><span class="comment">#这里选择 no，因为安装目录下就有</span></span><br><span class="line">n</span><br><span class="line"></span><br><span class="line"></span><br><span class="line"></span><br><span class="line"></span><br><span class="line"><span class="comment">#如果版本不同，提示也可能不同，根据情况输入或选择</span></span><br></pre></td></tr></table></figure><h2 id="cudnn"><a href="#cudnn" class="headerlink" title="cudnn"></a>cudnn</h2><h3 id="下载cudnn"><a href="#下载cudnn" class="headerlink" title="下载cudnn"></a>下载cudnn</h3><blockquote><p>注意和 <code>cuda</code> 版本匹配</p></blockquote><h3 id="安装cudnn"><a href="#安装cudnn" class="headerlink" title="安装cudnn"></a>安装cudnn</h3><p>按照官网命令</p><p><a href="https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html" target="_blank" rel="noopener">https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html</a></p><ul><li><p>Navigate to your <code>cudnnpath</code> directory containing the cuDNN Tar file.</p></li><li><p>Unzip the cuDNN package.</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">tar -xzvf cudnn-10.0-linux-x64-v7.5.0.56.tgz</span><br></pre></td></tr></table></figure></li><li><p>Copy the following files into the CUDA Toolkit directory, and change the file permissions.</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">sudo cp cuda/include/cudnn.h /usr/<span class="built_in">local</span>/cuda/include</span><br><span class="line">sudo cp cuda/lib64/libcudnn* /usr/<span class="built_in">local</span>/cuda/lib64</span><br><span class="line">sudo chmod a+r /usr/<span class="built_in">local</span>/cuda/include/cudnn.h /usr/<span class="built_in">local</span>/cuda/lib64/libcudnn*</span><br></pre></td></tr></table></figure></li></ul>]]></content>
    
    
      
      
    <summary type="html">&lt;h2 id=&quot;环境&quot;&gt;&lt;a href=&quot;#环境&quot; class=&quot;headerlink&quot; title=&quot;环境&quot;&gt;&lt;/a&gt;环境&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;p&gt;系统&lt;/p&gt;
&lt;p&gt;CentOS Linux release 7.6.1810 (Core)&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;</summary>
      
    
    
    
    <category term="安装记录" scheme="https://www.holmeyoung.com/categories/安装记录/"/>
    
    
    <category term="linux" scheme="https://www.holmeyoung.com/tags/linux/"/>
    
    <category term="gpu" scheme="https://www.holmeyoung.com/tags/gpu/"/>
    
    <category term="机器学习" scheme="https://www.holmeyoung.com/tags/机器学习/"/>
    
  </entry>
  
  <entry>
    <title>centos7 anaconda3 安装 youcompleteme</title>
    <link href="https://www.holmeyoung.com/blog/centos7-anaconda3-youcompleteme/"/>
    <id>https://www.holmeyoung.com/blog/centos7-anaconda3-youcompleteme/</id>
    <published>2018-11-07T02:27:06.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h3 id="环境"><a href="#环境" class="headerlink" title="环境"></a>环境</h3><ul><li><strong><code>CentOS7</code></strong></li><li><strong><code>Anaconda3</code></strong></li><li><strong><code>oh-my-zsh</code></strong></li></ul><h3 id="更新yum"><a href="#更新yum" class="headerlink" title="更新yum"></a>更新yum</h3><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo yum upgrade</span><br><span class="line">sudo yum update</span><br></pre></td></tr></table></figure><h3 id="安装可能需要的"><a href="#安装可能需要的" class="headerlink" title="安装可能需要的"></a>安装可能需要的</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">sudo yum install python-devel</span><br><span class="line">sudo yum install cmake</span><br></pre></td></tr></table></figure><h3 id="升级gcc"><a href="#升级gcc" class="headerlink" title="升级gcc"></a>升级gcc</h3><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo yum install centos-release-scl -y</span><br><span class="line">sudo yum install devtoolset-3-toolchain -y</span><br><span class="line">sudo yum install gcc-c++</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 如果安装了 oh-my—zsh 则不执行下面这句，否则会切换为bash</span></span><br><span class="line">sudo scl enable devtoolset-3 bash</span><br></pre></td></tr></table></figure><h3 id="禁用Anaconda3"><a href="#禁用Anaconda3" class="headerlink" title="禁用Anaconda3"></a>禁用Anaconda3</h3><p>将 </p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">export</span> PATH=<span class="string">"/root/anaconda3/bin:<span class="variable">$PATH</span>"</span></span><br></pre></td></tr></table></figure><p>从 <code>.zshrc</code> 、<code>.bashrc</code> 中注释掉</p><p>注释后<strong>根目录下</strong>执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">source</span> .zshrc</span><br><span class="line"><span class="built_in">source</span> .bashrc</span><br></pre></td></tr></table></figure><blockquote><p>此时执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">&gt; python -V</span><br><span class="line">&gt;</span><br></pre></td></tr></table></figure></blockquote><blockquote><p>依旧会显示 <code>Python 3.6.5 :: Anaconda, Inc.</code> </p><p><strong>所以我们需要新开一个终端运行下面的命令</strong></p></blockquote><h3 id="卸载系统原生vim（vim7）"><a href="#卸载系统原生vim（vim7）" class="headerlink" title="卸载系统原生vim（vim7）"></a>卸载系统原生vim（vim7）</h3><figure class="highlight plain"><table><tr><td class="code"><pre><span class="line">yum remove vim</span><br></pre></td></tr></table></figure><h3 id="自己源码安装vim（vim8）"><a href="#自己源码安装vim（vim8）" class="headerlink" title="自己源码安装vim（vim8）"></a>自己源码安装vim（vim8）</h3><h4 id="安装命令"><a href="#安装命令" class="headerlink" title="安装命令"></a>安装命令</h4><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">git clone https://github.com/vim/vim.git</span><br><span class="line">cd vim/src</span><br></pre></td></tr></table></figure><blockquote><p><strong><code>下面是最关键的一步，多少次失败都是因为这个命令不对！！！</code></strong></p></blockquote><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">./configure --with-features=huge --<span class="built_in">enable</span>-pythoninterp=yes --<span class="built_in">enable</span>-python3interp=yes --with-python3-config-dir=/root/anaconda3/lib/python3.6/config-3.6m-x86_64-linux-gnu</span><br></pre></td></tr></table></figure><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">sudo make</span><br><span class="line">sudo make install</span><br></pre></td></tr></table></figure><blockquote><p><code>--with-python3-config-dir=/root/anaconda3/lib/python3.6/config-3.6m-x86_64-linux-gnu</code></p><p>这条命令根据不同人 <code>anaconda3</code> 安装位置不同会有变动</p></blockquote><h4 id="环境设置"><a href="#环境设置" class="headerlink" title="环境设置"></a>环境设置</h4><p>将</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">export</span> PATH=<span class="string">"/usr/local/bin:<span class="variable">$PATH</span>"</span></span><br></pre></td></tr></table></figure><p>配置了 <code>oh-my-zsh</code> 的，添加到 <code>.zshrc</code></p><p>没配置 <code>oh-my-zsh</code> 的，添加到 <code>.bashrc</code></p><p>添加后<strong>根目录下</strong>执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">source</span> .zshrc</span><br><span class="line"><span class="comment"># source .bashrc</span></span><br></pre></td></tr></table></figure><h4 id="关于安装"><a href="#关于安装" class="headerlink" title="关于安装"></a>关于安装</h4><h5 id="命令参数"><a href="#命令参数" class="headerlink" title="命令参数"></a>命令参数</h5><p>这里列举一些<code>./configure</code> 后面的配置选项</p><table><thead><tr><th>参数</th><th>含义</th></tr></thead><tbody><tr><td><strong><code>--with-features=huge</code></strong></td><td>支持最大特性</td></tr><tr><td><strong><code>--enable-rubyinterp=yes</code></strong></td><td>打开对ruby编写的插件的支持</td></tr><tr><td><strong><code>--enable-pythoninterp=yes</code></strong></td><td>打开对python编写的插件的支持</td></tr><tr><td><strong><code>--enable-python3interp=yes</code></strong></td><td>打开对python3编写的插件的支持</td></tr><tr><td><strong><code>--enable-luainterp=yes</code></strong></td><td>打开对lua编写的插件的支持</td></tr><tr><td><strong><code>--enable-perlinterp=yes</code></strong></td><td>打开对perl编写的插件的支持</td></tr><tr><td><strong><code>--enable-multibyte=yes</code></strong></td><td>打开多字节支持，可以在Vim中输入中文</td></tr><tr><td><strong><code>--enable-cscope=yes</code></strong></td><td>打开对cscope的支持</td></tr><tr><td><strong><code>--with-python-config-dir=/usr/lib/python2.7/config-x86_64-linux-gnu</code></strong></td><td>指定python 路径</td></tr><tr><td><strong><code>--with-python3-config-dir=/usr/lib/python3.5/config-3.5m-x86_64-linux-gnu</code></strong></td><td>指定python3路径</td></tr><tr><td><strong><code>--prefix=/usr/local/vim</code></strong></td><td>指定将要安装到的路径(自行创建)</td></tr></tbody></table><h5 id="安装路径"><a href="#安装路径" class="headerlink" title="安装路径"></a>安装路径</h5><ul><li><p>通过 <code>yum install xxx</code> 安装的 <strong>软件</strong> 在 <code>usr/bin/</code> 下</p></li><li><p>通过 <strong>源码</strong> 安装的 <strong>软件</strong> 在 <code>usr/local/bin/</code> 下</p></li></ul><p>当我们通过 <code>yum remove vim</code> 卸载 <code>vim</code> 后再通过源码安装，安装好之后再输入 <code>vim</code> 默认指向的就是 <code>usr/local/bin/</code> 下的自己安装的</p><p>如下</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">➜  /root whereis vim</span><br><span class="line">vim: /usr/local/bin/vim /usr/share/vim /usr/share/man/man1/vim.1.gz</span><br></pre></td></tr></table></figure><h3 id="安装YouCompleteMe"><a href="#安装YouCompleteMe" class="headerlink" title="安装YouCompleteMe"></a>安装YouCompleteMe</h3><p>根目录下执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git <span class="built_in">clone</span> https://github.com/VundleVim/Vundle.vim.git ~/.vim/bundle/Vundle.vim</span><br></pre></td></tr></table></figure><p>编辑 <code>.vimrc</code> ，加入</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="string">" Vundle Setting</span></span><br><span class="line"><span class="string">set nocompatible</span></span><br><span class="line"><span class="string">filetype off</span></span><br><span class="line"><span class="string">set rtp+=~/.vim/bundle/Vundle.vim</span></span><br><span class="line"><span class="string">call vundle#begin()</span></span><br><span class="line"><span class="string"></span></span><br><span class="line"><span class="string">"</span> <span class="built_in">let</span> Vundle manage Vundle, required</span><br><span class="line">Plugin <span class="string">'VundleVim/Vundle.vim'</span></span><br><span class="line"></span><br><span class="line"><span class="string">" Plugins</span></span><br><span class="line"><span class="string">Plugin 'Valloric/YouCompleteMe'</span></span><br><span class="line"><span class="string"></span></span><br><span class="line"><span class="string">"</span> All of your Plugins must be added before the following line</span><br><span class="line">call vundle<span class="comment">#end()</span></span><br><span class="line">filetype plugin indent on</span><br></pre></td></tr></table></figure><p>保存退出</p><p>在终端执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">vim +PluginInstall +qall</span><br></pre></td></tr></table></figure><p>或者在 <code>vim</code> 中执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">:PluginInstall</span><br></pre></td></tr></table></figure><p>等待插件安装完成自动退出</p><p>此时会提示 <code>YouCompleteMe</code> 未编译</p><p>执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">cd</span> ~/.vim/bundle/YouCompleteMe</span><br><span class="line">./install.py --clang-completer</span><br></pre></td></tr></table></figure><p>这里只编译了 <code>YouCompleteMe</code> 对 <code>C</code> 族的补全，更多语言的补全命令可见 <a href="https://github.com/Valloric/YouCompleteMe" target="_blank" rel="noopener">YouCompleteMe</a> </p><h4 id="关于另一个版本的安装方法"><a href="#关于另一个版本的安装方法" class="headerlink" title="关于另一个版本的安装方法"></a>关于另一个版本的安装方法</h4><ol><li><p>如果已经安装 <code>Vundle</code> 直接跳到 <code>2</code></p><p>根目录下执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git <span class="built_in">clone</span> https://github.com/VundleVim/Vundle.vim.git ~/.vim/bundle/Vundle.vim</span><br></pre></td></tr></table></figure><p>编辑 <code>.vimrc</code> ，加入</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="string">" Vundle Setting</span></span><br><span class="line"><span class="string">set nocompatible</span></span><br><span class="line"><span class="string">filetype off</span></span><br><span class="line"><span class="string">set rtp+=~/.vim/bundle/Vundle.vim</span></span><br><span class="line"><span class="string">call vundle#begin()</span></span><br><span class="line"><span class="string"></span></span><br><span class="line"><span class="string">"</span> <span class="built_in">let</span> Vundle manage Vundle, required</span><br><span class="line">Plugin <span class="string">'VundleVim/Vundle.vim'</span></span><br><span class="line"></span><br><span class="line"><span class="string">" Plugins</span></span><br><span class="line"><span class="string">"</span> Plugin <span class="string">'Valloric/YouCompleteMe'</span></span><br><span class="line"></span><br><span class="line"><span class="string">" All of your Plugins must be added before the following line</span></span><br><span class="line"><span class="string">call vundle#end()</span></span><br><span class="line"><span class="string">filetype plugin indent on</span></span><br></pre></td></tr></table></figure><p>保存退出执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">vim +PluginInstall +qall</span><br></pre></td></tr></table></figure><p>或者在 <code>vim</code> 中执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">:PluginInstall</span><br></pre></td></tr></table></figure></li><li><p>安装 <code>YouCompleteMe</code></p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git <span class="built_in">clone</span> https://github.com/Valloric/YouCompleteMe.git ~/.vim/bundle/YouCompleteMe</span><br><span class="line"><span class="built_in">cd</span> ~/.vim/bundle/YouCompleteMe</span><br><span class="line">git submodule update --init --recursive</span><br></pre></td></tr></table></figure><p>编辑 <code>.vimrc</code> ，取消第 <code>11</code> 行的注释，即变为</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">Plugin <span class="string">'Valloric/YouCompleteMe'</span></span><br></pre></td></tr></table></figure><p>保存退出后执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">cd</span> ~/.vim/bundle/YouCompleteMe</span><br><span class="line">./install.py --clang-completer</span><br></pre></td></tr></table></figure><p><strong>DONE</strong></p></li></ol><blockquote><p><strong><code>这个安装方法我试了是没有成功的，可能是我环境的原因</code></strong></p></blockquote><blockquote><p>亲测在 <code>.vimrc</code> 中添加</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">&gt; Plugin <span class="string">'Valloric/YouCompleteMe'</span></span><br><span class="line">&gt;</span><br></pre></td></tr></table></figure></blockquote><blockquote><p>并执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">&gt; vim +PluginInstall +qall</span><br><span class="line">&gt;</span><br></pre></td></tr></table></figure></blockquote><blockquote><p>后</p><p>命令</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">&gt; git submodule update --init --recursive</span><br><span class="line">&gt;</span><br></pre></td></tr></table></figure></blockquote><blockquote><p>会失效，故采用 <code>Vundle</code> 安装可直接执行命令</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">&gt; ./install.py --clang-completer</span><br><span class="line">&gt;</span><br></pre></td></tr></table></figure></blockquote><blockquote><p>进行编译</p></blockquote><h3 id="放生Anaconda3"><a href="#放生Anaconda3" class="headerlink" title="放生Anaconda3"></a>放生Anaconda3</h3><p>将</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">export</span> PATH=<span class="string">"/root/anaconda3/bin:<span class="variable">$PATH</span>"</span></span><br></pre></td></tr></table></figure><p>在 <code>.zshrc</code> 中的注释去掉</p><p>没有安装 <code>oh-my-zsh</code> 的则去掉 <code>.bashrc</code> 中的注释</p><p>保存退出后<strong>根目录下</strong>执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">source</span> .zshrc</span><br><span class="line"><span class="built_in">source</span> .bashrc</span><br></pre></td></tr></table></figure><blockquote><p>此时执行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">&gt; python -V</span><br><span class="line">&gt;</span><br></pre></td></tr></table></figure></blockquote><blockquote><p>会显示 <code>Python 3.6.5 :: Anaconda, Inc.</code> </p></blockquote><p><strong>DONE</strong></p><h3 id="Ref"><a href="#Ref" class="headerlink" title="Ref"></a>Ref</h3><blockquote><p><a href="https://www.cnblogs.com/hugb/p/8593012.html" target="_blank" rel="noopener">Centos7安装vim8.0 + YouCompleteMe</a></p></blockquote>]]></content>
    
    
      
      
    <summary type="html">&lt;h3 id=&quot;环境&quot;&gt;&lt;a href=&quot;#环境&quot; class=&quot;headerlink&quot; title=&quot;环境&quot;&gt;&lt;/a&gt;环境&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;CentOS7&lt;/code&gt;&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;&lt;code&gt;Ana</summary>
      
    
    
    
    <category term="安装记录" scheme="https://www.holmeyoung.com/categories/安装记录/"/>
    
    
    <category term="linux" scheme="https://www.holmeyoung.com/tags/linux/"/>
    
  </entry>
  
  <entry>
    <title>python 进程间管道通信</title>
    <link href="https://www.holmeyoung.com/blog/python-process-fifo/"/>
    <id>https://www.holmeyoung.com/blog/python-process-fifo/</id>
    <published>2018-11-05T06:04:13.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<p>实现了两个 <code>Linux</code> 进程之间通过管道进行通讯的代码逻辑。分为 <code>client.py</code> 与 <code>server.py</code> 运行时需要先运行 <code>server</code> 再运行 <code>client</code></p><p><code>client.py</code></p><figure class="highlight python"><table><tr><td class="code"><pre><span class="line"><span class="keyword">import</span> os</span><br><span class="line"><span class="keyword">import</span> time</span><br><span class="line"></span><br><span class="line">write_path = <span class="string">"pipe.in"</span></span><br><span class="line">read_path = <span class="string">"pipe.out"</span></span><br><span class="line"></span><br><span class="line">wf = os.open(write_path, os.O_SYNC | os.O_CREAT | os.O_RDWR)</span><br><span class="line">rf = <span class="literal">None</span></span><br><span class="line"><span class="comment"># rf = os.open(read_path, os.O_RDONLY)</span></span><br><span class="line"><span class="keyword">for</span> i <span class="keyword">in</span> range(<span class="number">1</span>, <span class="number">11</span>):</span><br><span class="line">    msg = <span class="string">"msg "</span> + str(i)</span><br><span class="line">    msg = msg.encode()</span><br><span class="line">    len_send = os.write(wf, msg)</span><br><span class="line">    <span class="keyword">print</span> (<span class="string">"sent msg: %s"</span> % msg)</span><br><span class="line"></span><br><span class="line">    <span class="keyword">if</span> rf <span class="keyword">is</span> <span class="literal">None</span>:</span><br><span class="line">        rf = os.open(read_path, os.O_RDONLY)</span><br><span class="line"></span><br><span class="line">    s = os.read(rf, <span class="number">1024</span>)</span><br><span class="line">    <span class="keyword">if</span> len(s) == <span class="number">0</span>:</span><br><span class="line">        <span class="keyword">break</span></span><br><span class="line">    <span class="keyword">print</span> (<span class="string">"received msg: %s"</span> % s)</span><br><span class="line"></span><br><span class="line">    time.sleep(<span class="number">1</span>)</span><br><span class="line"></span><br><span class="line">os.write(wf, <span class="string">'exit'</span>.encode())</span><br><span class="line"></span><br><span class="line">os.close(rf)</span><br><span class="line">os.close(wf)</span><br></pre></td></tr></table></figure><p><code>server.py</code></p><figure class="highlight python"><table><tr><td class="code"><pre><span class="line"><span class="keyword">import</span> os, time</span><br><span class="line"></span><br><span class="line">read_path = <span class="string">"pipe.in"</span></span><br><span class="line">write_path = <span class="string">"pipe.out"</span></span><br><span class="line"></span><br><span class="line"><span class="keyword">if</span> os.path.exists(read_path):</span><br><span class="line">    os.remove(read_path)</span><br><span class="line"><span class="keyword">if</span> os.path.exists(write_path):</span><br><span class="line">    os.remove(write_path)</span><br><span class="line"></span><br><span class="line">os.mkfifo(write_path)</span><br><span class="line">os.mkfifo(read_path)</span><br><span class="line"></span><br><span class="line">rf = os.open(read_path, os.O_RDONLY)</span><br><span class="line">wf = os.open(write_path, os.O_SYNC | os.O_CREAT | os.O_RDWR)</span><br><span class="line"></span><br><span class="line"><span class="keyword">while</span> <span class="literal">True</span>:</span><br><span class="line">    s = os.read(rf, <span class="number">1024</span>)</span><br><span class="line">    <span class="keyword">print</span> (<span class="string">"received msg: %s"</span> % s)</span><br><span class="line">    <span class="keyword">if</span> len(s) == <span class="number">0</span>:</span><br><span class="line">        time.sleep(<span class="number">1</span>)</span><br><span class="line">        <span class="keyword">continue</span></span><br><span class="line"></span><br><span class="line">    <span class="keyword">if</span> <span class="string">"exit"</span>.encode() <span class="keyword">in</span> s:</span><br><span class="line">        <span class="keyword">break</span></span><br><span class="line"></span><br><span class="line">    os.write(wf, s)</span><br><span class="line"></span><br><span class="line">os.close(rf)</span><br><span class="line">os.close(wf)</span><br></pre></td></tr></table></figure>]]></content>
    
    
      
      
    <summary type="html">&lt;p&gt;实现了两个 &lt;code&gt;Linux&lt;/code&gt; 进程之间通过管道进行通讯的代码逻辑。分为 &lt;code&gt;client.py&lt;/code&gt; 与 &lt;code&gt;server.py&lt;/code&gt; 运行时需要先运行 &lt;code&gt;server&lt;/code&gt; 再运行 &lt;code&gt;clie</summary>
      
    
    
    
    <category term="python" scheme="https://www.holmeyoung.com/categories/python/"/>
    
    
    <category term="python" scheme="https://www.holmeyoung.com/tags/python/"/>
    
  </entry>
  
  <entry>
    <title>修改 pip 源为清华大学开源镜像</title>
    <link href="https://www.holmeyoung.com/blog/change-pip-mirror/"/>
    <id>https://www.holmeyoung.com/blog/change-pip-mirror/</id>
    <published>2018-06-14T11:19:17.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h3 id="pypi-镜像使用帮助"><a href="#pypi-镜像使用帮助" class="headerlink" title="pypi 镜像使用帮助"></a>pypi 镜像使用帮助</h3><p><code>pypi</code> 镜像每 5 分钟同步一次。</p><h3 id="临时使用"><a href="#临时使用" class="headerlink" title="临时使用"></a>临时使用</h3><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">pip install -i https://pypi.tuna.tsinghua.edu.cn/simple some-package</span><br></pre></td></tr></table></figure><p>注意， <code>simple</code> 不能少, 是 <code>https</code> 而不是 <code>http</code> </p><h3 id="设为默认"><a href="#设为默认" class="headerlink" title="设为默认"></a>设为默认</h3><p>修改 <code>~/.config/pip/pip.conf</code> <strong>(Linux)</strong>, <code>%APPDATA%\pip\pip.ini</code> <strong>(Windows 10)</strong> 或 <code>$HOME/Library/Application Support/pip/pip.conf</code> <strong>(macOS)</strong> (没有就创建一个)， 修改 <code>index-url</code> 至 <code>tuna</code> ，例如</p><figure class="highlight plain"><table><tr><td class="code"><pre><span class="line">[global]</span><br><span class="line">index-url = https://pypi.tuna.tsinghua.edu.cn/simple</span><br></pre></td></tr></table></figure><p><code>pip</code> 和 <code>pip3</code> 并存时，只需修改 <code>~/.pip/pip.conf</code> </p>]]></content>
    
    
      
      
    <summary type="html">&lt;h3 id=&quot;pypi-镜像使用帮助&quot;&gt;&lt;a href=&quot;#pypi-镜像使用帮助&quot; class=&quot;headerlink&quot; title=&quot;pypi 镜像使用帮助&quot;&gt;&lt;/a&gt;pypi 镜像使用帮助&lt;/h3&gt;&lt;p&gt;&lt;code&gt;pypi&lt;/code&gt; 镜像每 5 分钟同步一次。&lt;/p</summary>
      
    
    
    
    <category term="linux" scheme="https://www.holmeyoung.com/categories/linux/"/>
    
    
    <category term="python" scheme="https://www.holmeyoung.com/tags/python/"/>
    
  </entry>
  
  <entry>
    <title>ubuntu14.04 安装 caffe</title>
    <link href="https://www.holmeyoung.com/blog/ubuntu14-caffe/"/>
    <id>https://www.holmeyoung.com/blog/ubuntu14-caffe/</id>
    <published>2018-04-30T09:01:05.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h3 id="组件-amp-版本"><a href="#组件-amp-版本" class="headerlink" title="组件&amp;版本"></a>组件&amp;版本</h3><ul><li><p><strong>Ubuntu</strong></p><ul><li>14.04</li></ul></li><li><p><strong>CUDA</strong></p><ul><li>8.0</li></ul></li><li><strong>cuDNN</strong><ul><li>cudnn-7.0-linux-x64-<strong>v3.0</strong>-prod.tgz</li></ul></li><li><strong>opencv</strong><ul><li>3.2</li></ul></li><li><strong>python</strong><ul><li>Python 2.7.6</li></ul></li><li><strong>caffe</strong><ul><li>release candidate 3</li></ul></li></ul><h3 id="安装过程"><a href="#安装过程" class="headerlink" title="安装过程"></a>安装过程</h3><h4 id="安装开发所需依赖包"><a href="#安装开发所需依赖包" class="headerlink" title="安装开发所需依赖包"></a>安装开发所需依赖包</h4><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo apt-get install build-essential  </span><br><span class="line">sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libboost-all-dev libhdf5-serial-dev libgflags-dev libgoogle-glog-dev liblmdb-dev protobuf-compiler</span><br></pre></td></tr></table></figure><h4 id="安装CUDA-8-0"><a href="#安装CUDA-8-0" class="headerlink" title="安装CUDA 8.0"></a>安装CUDA 8.0</h4><ul><li><p>确定 <code>GPU</code> 支持 <code>CUDA</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">lspci | grep -i nvidia</span><br></pre></td></tr></table></figure><p>显示</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">02:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)</span><br><span class="line">02:00.1 Audio device: NVIDIA Corporation Device 10ef (rev a1)</span><br><span class="line">03:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)</span><br><span class="line">03:00.1 Audio device: NVIDIA Corporation Device 10ef (rev a1)</span><br><span class="line">82:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)</span><br><span class="line">82:00.1 Audio device: NVIDIA Corporation Device 10ef (rev a1)</span><br><span class="line">83:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1)</span><br><span class="line">83:00.1 Audio device: NVIDIA Corporation Device 10ef (rev a1)</span><br></pre></td></tr></table></figure><p>根据 <strong><a href="http://developer.nvidia.com/cuda-gpus" target="_blank" rel="noopener">http://developer.nvidia.com/cuda-gpus</a></strong> 去验证，发现支持 <code>CUDA</code></p></li><li><p>确定 <code>linux</code> 版本支持 <code>CUDA</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">uname -m &amp;&amp; cat /etc/*release</span><br></pre></td></tr></table></figure><p>显示</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">x86_64</span><br><span class="line">DISTRIB_ID=Ubuntu</span><br><span class="line">DISTRIB_RELEASE=14.04</span><br><span class="line">DISTRIB_CODENAME=trusty</span><br><span class="line">DISTRIB_DESCRIPTION="Ubuntu 14.04.5 LTS"</span><br><span class="line">NAME="Ubuntu"</span><br><span class="line">VERSION="14.04.5 LTS, Trusty Tahr"</span><br><span class="line">ID=ubuntu</span><br><span class="line">ID_LIKE=debian</span><br><span class="line">PRETTY_NAME="Ubuntu 14.04.5 LTS"</span><br><span class="line">VERSION_ID="14.04"</span><br><span class="line">HOME_URL="http://www.ubuntu.com/"</span><br><span class="line">SUPPORT_URL="http://help.ubuntu.com/"</span><br><span class="line">BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"</span><br></pre></td></tr></table></figure></li><li><p>确定系统已经安装了 <code>gcc</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">gcc --version</span><br></pre></td></tr></table></figure><p>显示</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">gcc (Ubuntu 4.8.5-2ubuntu1~14.04.1) 4.8.5</span><br><span class="line">Copyright (C) 2015 Free Software Foundation, Inc.</span><br><span class="line">This is free software; see the source for copying conditions.  There is NO</span><br><span class="line">warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.</span><br></pre></td></tr></table></figure></li><li><p>确定系统已经安装了正确的 <code>Kernel Headers</code> 和开发包</p><ul><li><p>查看系统正在运行的 <code>kernel</code> 版本</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">uname -r</span><br></pre></td></tr></table></figure><p>显示</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">3.13.0-142-generic</span><br></pre></td></tr></table></figure></li><li><p>安装对应的 <code>kernels header</code> 和开发包</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo apt-get install linux-headers-$(uname -r)</span><br></pre></td></tr></table></figure></li></ul></li><li><p>安装 <code>CUDA</code></p><ul><li>下载对应版本的cuda，从 <strong><a href="https://developer.nvidia.com/cuda-downloads" target="_blank" rel="noopener">https://developer.nvidia.com/cuda-downloads</a></strong> 下载 <code>deb(network)</code> 版</li></ul></li></ul><ul><li><p>进行 <code>md5</code> 校验</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">md5sum cuda-repo-ubuntu1404_8.0.61-1_amd64.deb</span><br></pre></td></tr></table></figure></li><li><p>对比相同，然后使用 <code>deb</code> 文件进行安装</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo dpkg -i cuda-repo-ubuntu1404_8.0.61-1_amd64.deb</span><br><span class="line">sudo apt-get update</span><br><span class="line">sudo apt-get install cuda</span><br></pre></td></tr></table></figure></li><li><p>重启，完成 <code>cuda</code> 安装</p></li></ul><h4 id="安装-cuDNN"><a href="#安装-cuDNN" class="headerlink" title="安装 cuDNN"></a>安装 cuDNN</h4><ul><li><p>从 <strong><a href="https://developer.nvidia.com/rdp/cudnn-download" target="_blank" rel="noopener">https://developer.nvidia.com/rdp/cudnn-download</a></strong> 下载 <code>cuDNN</code> ，版本为 <code>cudnn-7.0-linux-x64-v3.0-prod.tgz</code></p></li><li><p>安装</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">tar -zxvf cudnn-7.0-linux-x64-v3.0-prod.tgz</span><br><span class="line">cd cuda</span><br><span class="line">sudo cp lib64/* /usr/local/cuda/lib64/</span><br><span class="line">sudo cp include/cudnn.h /usr/local/cuda/include/</span><br></pre></td></tr></table></figure></li><li><p>更新软链接</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">cd /usr/local/cuda/lib64</span><br><span class="line">sudo rm -rf libcudnn.so libcudnn.so.7.0</span><br><span class="line">sudo ln -s libcudnn.so.7.0.64 libcudnn.so.7.0</span><br><span class="line">sudo ln -s libcudnn.so.7.0 libcudnn.so</span><br></pre></td></tr></table></figure></li></ul><h4 id="设置-CUDA-环境变量"><a href="#设置-CUDA-环境变量" class="headerlink" title="设置 CUDA 环境变量"></a>设置 CUDA 环境变量</h4><ul><li><p><strong>VERSION1</strong></p><ul><li><p>在 <code>/etc/profile</code> 中添加 <code>CUDA</code> 环境变量</p></li><li><p>执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo vim /etc/profile</span><br></pre></td></tr></table></figure><p>在打开的文件中加入如下两句话</p><figure class="highlight bash"><table><tr><td class="code"><pre><span class="line"><span class="built_in">export</span> PATH=/usr/<span class="built_in">local</span>/cuda/bin:<span class="variable">$PATH</span></span><br><span class="line"><span class="built_in">export</span> LD_LIBRARY_PATH=/usr/<span class="built_in">local</span>/cuda/lib64:<span class="variable">$LD_LIBRARY_PATH</span></span><br></pre></td></tr></table></figure></li><li><p>保存后，使环境变量立即生效，执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">source /etc/profile</span><br></pre></td></tr></table></figure></li></ul></li><li><p><strong>VERSION2</strong></p><ul><li><p>在 <code>~/.bashrc</code> 中添加 <code>CUDA</code> 环境变量</p></li><li><p>执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo vim ~/.bashrc</span><br></pre></td></tr></table></figure><p>在打开的文件中加入如下两句话</p><figure class="highlight bash"><table><tr><td class="code"><pre><span class="line"><span class="built_in">export</span> PATH=/usr/<span class="built_in">local</span>/cuda/bin:<span class="variable">$PATH</span></span><br><span class="line"><span class="built_in">export</span> LD_LIBRARY_PATH=/usr/<span class="built_in">local</span>/cuda/lib64:<span class="variable">$LD_LIBRARY_PATH</span></span><br></pre></td></tr></table></figure></li><li><p>保存后，使环境变量立即生效，执行</p></li></ul></li><li><p>我用的 <code>VERSION2</code></p></li></ul><h4 id="安装cuda-samples"><a href="#安装cuda-samples" class="headerlink" title="安装cuda samples"></a>安装cuda samples</h4><ul><li><p>进入 <code>/usr/local/cuda/samples</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">cd /usr/local/cuda/samples</span><br></pre></td></tr></table></figure></li><li><p>执行下面的命令来 <code>build samples</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo make all -j4</span><br></pre></td></tr></table></figure></li><li><p>全部编译完成后，进入 <code>samples/bin/x86_64/linux/release</code> ，运行 <code>deviceQuery</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">cd /usr/local/cuda/samples/bin/x86_64/linux/release</span><br><span class="line">./deviceQuery</span><br></pre></td></tr></table></figure><p>如果出现显卡信息，则驱动及显卡安装成功，结果如下</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">./deviceQuery Starting...</span><br><span class="line"></span><br><span class="line"> CUDA Device Query (Runtime API) version (CUDART static linking)</span><br><span class="line"></span><br><span class="line">Detected 4 CUDA Capable device(s)</span><br><span class="line"></span><br><span class="line">Device 0: "GeForce GTX 1080 Ti"</span><br><span class="line">.</span><br><span class="line">.</span><br><span class="line">.</span><br><span class="line">Device 3: "GeForce GTX 1080 Ti"</span><br><span class="line">  CUDA Driver Version / Runtime Version          9.0 / 8.0</span><br><span class="line">  CUDA Capability Major/Minor version number:    6.1</span><br><span class="line">  Total amount of global memory:                 11172 MBytes (11715084288 bytes)</span><br><span class="line">  (28) Multiprocessors, (128) CUDA Cores/MP:     3584 CUDA Cores</span><br><span class="line">  GPU Max Clock rate:                            1582 MHz (1.58 GHz)</span><br><span class="line">  Memory Clock rate:                             5505 Mhz</span><br><span class="line">  Memory Bus Width:                              352-bit</span><br><span class="line">  L2 Cache Size:                                 2883584 bytes</span><br><span class="line">  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)</span><br><span class="line">  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers</span><br><span class="line">  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers</span><br><span class="line">  Total amount of constant memory:               65536 bytes</span><br><span class="line">  Total amount of shared memory per block:       49152 bytes</span><br><span class="line">  Total number of registers available per block: 65536</span><br><span class="line">  Warp size:                                     32</span><br><span class="line">  Maximum number of threads per multiprocessor:  2048</span><br><span class="line">  Maximum number of threads per block:           1024</span><br><span class="line">  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)</span><br><span class="line">  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)</span><br><span class="line">  Maximum memory pitch:                          2147483647 bytes</span><br><span class="line">  Texture alignment:                             512 bytes</span><br><span class="line">  Concurrent copy and kernel execution:          Yes with 2 copy engine(s)</span><br><span class="line">  Run time limit on kernels:                     No</span><br><span class="line">  Integrated GPU sharing Host Memory:            No</span><br><span class="line">  Support host page-locked memory mapping:       Yes</span><br><span class="line">  Alignment requirement for Surfaces:            Yes</span><br><span class="line">  Device has ECC support:                        Disabled</span><br><span class="line">  Device supports Unified Addressing (UVA):      Yes</span><br><span class="line">  Device PCI Domain ID / Bus ID / location ID:   0 / 131 / 0</span><br><span class="line">  Compute Mode:</span><br><span class="line">     &lt; Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) &gt;</span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU0) -&gt; GeForce GTX 1080 Ti (GPU1) : Yes</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU0) -&gt; GeForce GTX 1080 Ti (GPU2) : No</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU0) -&gt; GeForce GTX 1080 Ti (GPU3) : No</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU1) -&gt; GeForce GTX 1080 Ti (GPU0) : Yes</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU1) -&gt; GeForce GTX 1080 Ti (GPU2) : No</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU1) -&gt; GeForce GTX 1080 Ti (GPU3) : No</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU2) -&gt; GeForce GTX 1080 Ti (GPU0) : No</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU2) -&gt; GeForce GTX 1080 Ti (GPU1) : No</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU2) -&gt; GeForce GTX 1080 Ti (GPU3) : Yes</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU3) -&gt; GeForce GTX 1080 Ti (GPU0) : No</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU3) -&gt; GeForce GTX 1080 Ti (GPU1) : No</span></span><br><span class="line"><span class="meta">&gt;</span><span class="bash"> Peer access from GeForce GTX 1080 Ti (GPU3) -&gt; GeForce GTX 1080 Ti (GPU2) : Yes</span></span><br><span class="line"></span><br><span class="line">deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 9.0, CUDA Runtime Version = 8.0, NumDevs = 4, Device0 = GeForce GTX 1080 Ti, Device1 = GeForce GTX 1080 Ti, Device2 = GeForce GTX 1080 Ti, Device3 = GeForce GTX 1080 Ti</span><br><span class="line">Result = PASS</span><br></pre></td></tr></table></figure><blockquote><p>因为我是 4 个 GPU ，所以会有多个 Device 信息</p></blockquote><p>安装成功！</p></li></ul><h4 id="安装Intel-MKL，openBlas-或Atlas"><a href="#安装Intel-MKL，openBlas-或Atlas" class="headerlink" title="安装Intel MKL，openBlas 或Atlas"></a>安装Intel MKL，openBlas 或Atlas</h4><p>我选择的是 <code>Atlas</code> ，为 <code>caffe</code> 默认使用的，不要额外配置，安装命令</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo apt-get install libatlas-base-dev</span><br></pre></td></tr></table></figure><h4 id="安装opencv"><a href="#安装opencv" class="headerlink" title="安装opencv"></a>安装opencv</h4><ul><li><p>首先安装必须的包</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo apt-get install cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev</span><br></pre></td></tr></table></figure></li><li><p>从 <code>opencv</code> 官网下载最新版本 <code>opencv</code> ，<code>3.2.0.zip</code></p><p>解压</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">unzip 3.2.0.zip</span><br></pre></td></tr></table></figure></li><li><p>编译</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">cd  opencv-3.2.0</span><br><span class="line">mkdir release</span><br><span class="line">cd release</span><br><span class="line">cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local .. -DWITH_LAPACK=OFF</span><br><span class="line">make -j4</span><br><span class="line">sudo make install</span><br></pre></td></tr></table></figure></li><li><p>测试</p><p>因为我的 <code>GPU</code> 服务器走的别人的网关，这里显示不出来图片，也就没有测试。</p><p>但是是需要测试的，自行 <code>Google</code> 吧</p></li></ul><h4 id="安装python依赖库"><a href="#安装python依赖库" class="headerlink" title="安装python依赖库"></a>安装python依赖库</h4><ul><li><p>下载 <code>caffe</code> 源码，解压后进入 <code>caffe-master</code> 下的 <code>python</code> 目录</p><blockquote><p>这里一定要注意当前下载的 caffe 版本是否支持自己安装的 cuDNN 版本</p></blockquote></li><li><p>安装 <code>python-pip</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">sudo apt-get install python-pip</span><br></pre></td></tr></table></figure></li><li><p>执行如下命令安装相关依赖</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">for req in $(cat requirements.txt); do pip install $req; done</span><br></pre></td></tr></table></figure></li></ul><h4 id="编译Caffe"><a href="#编译Caffe" class="headerlink" title="编译Caffe"></a>编译Caffe</h4><ul><li><p>进入 <code>caffe-master</code> 目录，复制一份 <code>Makefile.config.examples</code> </p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">cp Makefile.config.example Makefile.config</span><br></pre></td></tr></table></figure></li><li><p>修改 <code>Makefile.config</code></p><p>开启对 <code>CuDNN</code> 和 <code>Opencv3</code> 的支持</p><figure class="highlight makefile"><table><tr><td class="code"><pre><span class="line"><span class="comment">## Refer to http://caffe.berkeleyvision.org/installation.html</span></span><br><span class="line"><span class="comment"># Contributions simplifying and improving our build system are welcome!</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># cuDNN acceleration switch (uncomment to build with cuDNN).</span></span><br><span class="line">USE_CUDNN := 1</span><br><span class="line"></span><br><span class="line"><span class="comment"># CPU-only switch (uncomment to build without GPU support).</span></span><br><span class="line"><span class="comment"># CPU_ONLY := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># uncomment to disable IO dependencies and corresponding data layers</span></span><br><span class="line"><span class="comment"># USE_OPENCV := 0</span></span><br><span class="line"><span class="comment"># USE_LEVELDB := 0</span></span><br><span class="line"><span class="comment"># USE_LMDB := 0</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># uncomment to allow MDB_NOLOCK when reading LMDB files (only if necessary)</span></span><br><span class="line"><span class="comment">#You should not set this flag if you will be reading LMDBs with any</span></span><br><span class="line"><span class="comment">#possibility of simultaneous read and write</span></span><br><span class="line"><span class="comment"># ALLOW_LMDB_NOLOCK := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment if you're using OpenCV 3</span></span><br><span class="line">OPENCV_VERSION := 3</span><br><span class="line"></span><br><span class="line"><span class="comment"># To customize your choice of compiler, uncomment and set the following.</span></span><br><span class="line"><span class="comment"># N.B. the default for Linux is g++ and the default for OSX is clang++</span></span><br><span class="line"><span class="comment"># CUSTOM_CXX := g++</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># CUDA directory contains bin/ and lib/ directories that we need.</span></span><br><span class="line">CUDA_DIR := /usr/local/cuda</span><br><span class="line"><span class="comment"># On Ubuntu 14.04, if cuda tools are installed via</span></span><br><span class="line"><span class="comment"># "sudo apt-get install nvidia-cuda-toolkit" then use this instead:</span></span><br><span class="line"><span class="comment"># CUDA_DIR := /usr</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># CUDA architecture setting: going with all of them.</span></span><br><span class="line"><span class="comment"># For CUDA &lt; 6.0, comment the *_50 lines for compatibility.</span></span><br><span class="line">CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \</span><br><span class="line">-gencode arch=compute_20,code=sm_21 \</span><br><span class="line">-gencode arch=compute_30,code=sm_30 \</span><br><span class="line">-gencode arch=compute_35,code=sm_35 \</span><br><span class="line">-gencode arch=compute_50,code=sm_50 \</span><br><span class="line">-gencode arch=compute_50,code=compute_50</span><br><span class="line"></span><br><span class="line"><span class="comment"># BLAS choice:</span></span><br><span class="line"><span class="comment"># atlas for ATLAS (default)</span></span><br><span class="line"><span class="comment"># mkl for MKL</span></span><br><span class="line"><span class="comment"># open for OpenBlas</span></span><br><span class="line">BLAS := atlas</span><br><span class="line"><span class="comment"># Custom (MKL/ATLAS/OpenBLAS) include and lib directories.</span></span><br><span class="line"><span class="comment"># Leave commented to accept the defaults for your choice of BLAS</span></span><br><span class="line"><span class="comment"># (which should work)!</span></span><br><span class="line"><span class="comment"># BLAS_INCLUDE := /path/to/your/blas</span></span><br><span class="line"><span class="comment"># BLAS_LIB := /path/to/your/blas</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Homebrew puts openblas in a directory that is not on the standard search path</span></span><br><span class="line"><span class="comment"># BLAS_INCLUDE := $(shell brew --prefix openblas)/include</span></span><br><span class="line"><span class="comment"># BLAS_LIB := $(shell brew --prefix openblas)/lib</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># This is required only if you will compile the matlab interface.</span></span><br><span class="line"><span class="comment"># MATLAB directory should contain the mex binary in /bin.</span></span><br><span class="line"><span class="comment"># MATLAB_DIR := /usr/local</span></span><br><span class="line"><span class="comment"># MATLAB_DIR := /Applications/MATLAB_R2012b.app</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># <span class="doctag">NOTE:</span> this is required only if you will compile the python interface.</span></span><br><span class="line"><span class="comment"># We need to be able to find Python.h and numpy/arrayobject.h.</span></span><br><span class="line">PYTHON_INCLUDE := /usr/<span class="keyword">include</span>/python2.7 \</span><br><span class="line">/usr/lib/python2.7/dist-packages/numpy/core/<span class="keyword">include</span></span><br><span class="line"><span class="comment"># Anaconda Python distribution is quite popular. Include path:</span></span><br><span class="line"><span class="comment"># Verify anaconda location, sometimes it's in root.</span></span><br><span class="line"><span class="comment"># ANACONDA_HOME := $(HOME)/anaconda</span></span><br><span class="line"><span class="comment"># PYTHON_INCLUDE := $(ANACONDA_HOME)/include \</span></span><br><span class="line"><span class="comment"># $(ANACONDA_HOME)/include/python2.7 \</span></span><br><span class="line"><span class="comment"># $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include \</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment to use Python 3 (default is Python 2)</span></span><br><span class="line"><span class="comment"># PYTHON_LIBRARIES := boost_python3 python3.5m</span></span><br><span class="line"><span class="comment"># PYTHON_INCLUDE := /usr/include/python3.5m \</span></span><br><span class="line"><span class="comment">#                 /usr/lib/python3.5/dist-packages/numpy/core/include</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># We need to be able to find libpythonX.X.so or .dylib.</span></span><br><span class="line">PYTHON_LIB := /usr/lib</span><br><span class="line"><span class="comment"># PYTHON_LIB := $(ANACONDA_HOME)/lib</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Homebrew installs numpy in a non standard path (keg only)</span></span><br><span class="line"><span class="comment"># PYTHON_INCLUDE += $(dir $(shell python -c 'import numpy.core; print(numpy.core.__file__)'))/include</span></span><br><span class="line"><span class="comment"># PYTHON_LIB += $(shell brew --prefix numpy)/lib</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment to support layers written in Python (will link against Python libs)</span></span><br><span class="line"><span class="comment"># WITH_PYTHON_LAYER := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Whatever else you find you need goes here.</span></span><br><span class="line">INCLUDE_DIRS := <span class="variable">$(PYTHON_INCLUDE)</span> /usr/local/<span class="keyword">include</span></span><br><span class="line">LIBRARY_DIRS := <span class="variable">$(PYTHON_LIB)</span> /usr/local/lib /usr/lib</span><br><span class="line"></span><br><span class="line"><span class="comment"># If Homebrew is installed at a non standard location (for example your home directory) and you use it for general dependencies</span></span><br><span class="line"><span class="comment"># INCLUDE_DIRS += $(shell brew --prefix)/include</span></span><br><span class="line"><span class="comment"># LIBRARY_DIRS += $(shell brew --prefix)/lib</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment to use `pkg-config` to specify OpenCV library paths.</span></span><br><span class="line"><span class="comment"># (Usually not necessary -- OpenCV libraries are normally installed in one of the above $LIBRARY_DIRS.)</span></span><br><span class="line"><span class="comment"># USE_PKG_CONFIG := 1</span></span><br><span class="line"></span><br><span class="line">BUILD_DIR := build</span><br><span class="line">DISTRIBUTE_DIR := distribute</span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment for debugging. Does not work on OSX due to https://github.com/BVLC/caffe/issues/171</span></span><br><span class="line"><span class="comment"># DEBUG := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># The ID of the GPU that 'make runtest' will use to run unit tests.</span></span><br><span class="line">TEST_GPUID := 0</span><br><span class="line"></span><br><span class="line"><span class="comment"># enable pretty build (comment to see full commands)</span></span><br><span class="line">Q ?= @</span><br></pre></td></tr></table></figure></li><li><p>保存退出，编译</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">make all -j4</span><br><span class="line">make test -j4</span><br><span class="line">make runtest -j4</span><br></pre></td></tr></table></figure></li><li><p>结果</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">[ FAILED ] 5 tests, listed below:</span><br><span class="line">[ FAILED ] CuDNNConvolutionLayerTest/0.TestSimpleConvolutionCuDNN, where TypeParam = float</span><br><span class="line">[ FAILED ] CuDNNConvolutionLayerTest/0.TestGradientCuDNN, where TypeParam = float</span><br><span class="line">[ FAILED ] CuDNNConvolutionLayerTest/0.TestGradientGroupCuDNN, where TypeParam = float</span><br><span class="line">[ FAILED ] CuDNNConvolutionLayerTest/0.TestSobelConvolutionCuDNN, where TypeParam = float</span><br><span class="line">[ FAILED ] CuDNNConvolutionLayerTest/0.TestSimpleConvolutionGroupCuDNN, where TypeParam = float</span><br></pre></td></tr></table></figure><p>这个是因为我们使用的是 <code>cuDNN v3</code> 版本，而这部分只在 <code>cuDNN v4</code> 下 <code>work</code></p></li></ul><h3 id="更换cuDNN与caffe版本"><a href="#更换cuDNN与caffe版本" class="headerlink" title="更换cuDNN与caffe版本"></a>更换cuDNN与caffe版本</h3><h4 id="为啥要换"><a href="#为啥要换" class="headerlink" title="为啥要换"></a>为啥要换</h4><p>我当然不想换！因为我老大不太满意！也没有说不满意，就是素质三连了一下</p><p>都行</p><p>随你</p><p>都可以</p><p>。。。</p><h4 id="确定依赖关系"><a href="#确定依赖关系" class="headerlink" title="确定依赖关系"></a>确定依赖关系</h4><ul><li><p><code>CuDNN</code> 是专门针对 <code>Deep Learning</code> 框架设计的一套 <code>GPU</code> 计算加速方案，目前支持的 <code>DL</code> 库包括 <code>Caffe</code> ，<code>ConvNet</code> ，<code>Torch7</code> 等</p><p>基本原理是把 <code>lib</code> 文件加入到系统能找到的 <code>lib</code> 文件夹里， 把头文件加到系统能找到的 <code>include</code> 文件夹里就可以。这里把他们加到 <code>CUDA</code> 的文件夹下</p></li><li><p><code>CUDA</code> 是直接安装的、<code>OpenCV</code> 安装的时候是依赖于 <code>CUDA</code> 的、<code>cuDNN</code> 的安装就是将其 <code>lib文件加入到系统能找到的lib文件夹里， 把头文件加到系统能找到的include文件夹里，这里把他们加到CUDA的文件夹下</code></p></li><li><p>所以说 <code>cuDNN</code> 无关于 <code>CUDA</code> ，无关于 <code>OpenCV</code> ，它只和 <code>caffe</code> 有关系，因为 <code>caffe</code> 安装的时候要依赖它。</p></li></ul><h4 id="重装过程"><a href="#重装过程" class="headerlink" title="重装过程"></a>重装过程</h4><ul><li><p><code>cuDNN</code> 文件结构</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">root@Jedy:/home/jedy/deeplearning/cuda# tree -N</span><br><span class="line">.</span><br><span class="line">├── include</span><br><span class="line">│   └── cudnn.h</span><br><span class="line">└── lib64</span><br><span class="line">    ├── libcudnn.so</span><br><span class="line">    ├── libcudnn.so.5</span><br><span class="line">    ├── libcudnn.so.5.0.5</span><br><span class="line">    └── libcudnn_static.a</span><br><span class="line"></span><br><span class="line">2 directories, 5 files</span><br></pre></td></tr></table></figure><p>而安装 <code>cuDNN</code> 的命令是</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">tar -zxvf cudnn-7.0-linux-x64-v3.0-prod.tgz </span><br><span class="line">cd cuda</span><br><span class="line">sudo cp lib64/* /usr/local/cuda/lib64/</span><br><span class="line">sudo cp include/cudnn.h /usr/local/cuda/include/</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash">更新软链接</span></span><br><span class="line">cd /usr/local/cuda/lib64</span><br><span class="line">sudo rm -rf libcudnn.so libcudnn.so.7.0</span><br><span class="line">sudo ln -s libcudnn.so.7.0.64 libcudnn.so.7.0</span><br><span class="line">sudo ln -s libcudnn.so.7.0 libcudnn.so</span><br></pre></td></tr></table></figure></li><li><p>重装 <code>cuDNN</code> </p><ul><li><p>因为 <code>cuDNN</code> 只是将文件拷贝进去，所以我们只需要下载更新版本的 <code>cuDNN</code> 然后覆盖安装即可。我是打算一步到位，使用最新的 <code>caffe</code> ，又怕 <code>cuDNN</code> 向下兼容不好，所以下载了 <code>v5</code> 版本，版本为 <code>cudnn-8.0-linux-x64-v5.0-ga.tgz</code></p></li><li><p>进入到 <code>/usr/local/cuda/lib64</code> 中，删除旧版本的 <code>cuDNN</code></p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">cd /usr/local/cuda/lib64</span><br><span class="line">sudo rm -rf libcudnn.so libcudnn.so.7.0 libcudnn.so.7.0.64</span><br></pre></td></tr></table></figure></li><li><p>解压新版本的 <code>cuDNN</code> 并拷贝</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">tar -zxvf cudnn-8.0-linux-x64-v5.0-ga.tgz</span><br><span class="line">cd cuda</span><br><span class="line">sudo cp lib64/* /usr/local/cuda/lib64/</span><br><span class="line">sudo cp include/cudnn.h /usr/local/cuda/include/</span><br></pre></td></tr></table></figure><blockquote><p>这里的 include/cudnn.h 我们并没有删除旧版本的，而是采用了覆盖安装的方式</p></blockquote></li><li><p>更新软链接</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">cd /usr/local/cuda/lib64</span><br><span class="line">sudo rm -rf libcudnn.so libcudnn.so.5</span><br><span class="line">sudo ln -s libcudnn.so.5.0.5 libcudnn.so.5</span><br><span class="line">sudo ln -s libcudnn.so.5 libcudnn.so</span><br></pre></td></tr></table></figure></li></ul></li><li><p>重装 <code>caffe</code></p><ul><li><p>在 <code>caffe</code> 的 <code>root</code> 目录执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">make clean</span><br></pre></td></tr></table></figure></li><li><p>删除旧版本的 <code>caffe</code> </p></li><li><p>下载新版本的 <code>caffe</code> ，我这里直接下载了 <code>1.0</code> 版本的 <code>caffe</code> </p></li><li><p>进入 <code>caffe-master</code> 目录，复制一份 <code>Makefile.config.examples</code> </p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">cp Makefile.config.example Makefile.config</span><br></pre></td></tr></table></figure></li><li><p>修改 <code>Makefile.config</code> ，开启 <code>CuDNN</code> 和 <code>Opencv3</code> 的支持即可，修改好的如下</p><figure class="highlight makefile"><table><tr><td class="code"><pre><span class="line"><span class="comment">## Refer to http://caffe.berkeleyvision.org/installation.html</span></span><br><span class="line"><span class="comment"># Contributions simplifying and improving our build system are welcome!</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># cuDNN acceleration switch (uncomment to build with cuDNN).</span></span><br><span class="line">USE_CUDNN := 1</span><br><span class="line"></span><br><span class="line"><span class="comment"># CPU-only switch (uncomment to build without GPU support).</span></span><br><span class="line"><span class="comment"># CPU_ONLY := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># uncomment to disable IO dependencies and corresponding data layers</span></span><br><span class="line"><span class="comment"># USE_OPENCV := 0</span></span><br><span class="line"><span class="comment"># USE_LEVELDB := 0</span></span><br><span class="line"><span class="comment"># USE_LMDB := 0</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># uncomment to allow MDB_NOLOCK when reading LMDB files (only if necessary)</span></span><br><span class="line"><span class="comment">#You should not set this flag if you will be reading LMDBs with any</span></span><br><span class="line"><span class="comment">#possibility of simultaneous read and write</span></span><br><span class="line"><span class="comment"># ALLOW_LMDB_NOLOCK := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment if you're using OpenCV 3</span></span><br><span class="line">OPENCV_VERSION := 3</span><br><span class="line"></span><br><span class="line"><span class="comment"># To customize your choice of compiler, uncomment and set the following.</span></span><br><span class="line"><span class="comment"># N.B. the default for Linux is g++ and the default for OSX is clang++</span></span><br><span class="line"><span class="comment"># CUSTOM_CXX := g++</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># CUDA directory contains bin/ and lib/ directories that we need.</span></span><br><span class="line">CUDA_DIR := /usr/local/cuda</span><br><span class="line"><span class="comment"># On Ubuntu 14.04, if cuda tools are installed via</span></span><br><span class="line"><span class="comment"># "sudo apt-get install nvidia-cuda-toolkit" then use this instead:</span></span><br><span class="line"><span class="comment"># CUDA_DIR := /usr</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># CUDA architecture setting: going with all of them.</span></span><br><span class="line"><span class="comment"># For CUDA &lt; 6.0, comment the *_50 through *_61 lines for compatibility.</span></span><br><span class="line"><span class="comment"># For CUDA &lt; 8.0, comment the *_60 and *_61 lines for compatibility.</span></span><br><span class="line">CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \</span><br><span class="line">-gencode arch=compute_20,code=sm_21 \</span><br><span class="line">-gencode arch=compute_30,code=sm_30 \</span><br><span class="line">-gencode arch=compute_35,code=sm_35 \</span><br><span class="line">-gencode arch=compute_50,code=sm_50 \</span><br><span class="line">-gencode arch=compute_52,code=sm_52 \</span><br><span class="line">-gencode arch=compute_60,code=sm_60 \</span><br><span class="line">-gencode arch=compute_61,code=sm_61 \</span><br><span class="line">-gencode arch=compute_61,code=compute_61</span><br><span class="line"></span><br><span class="line"><span class="comment"># BLAS choice:</span></span><br><span class="line"><span class="comment"># atlas for ATLAS (default)</span></span><br><span class="line"><span class="comment"># mkl for MKL</span></span><br><span class="line"><span class="comment"># open for OpenBlas</span></span><br><span class="line">BLAS := atlas</span><br><span class="line"><span class="comment"># Custom (MKL/ATLAS/OpenBLAS) include and lib directories.</span></span><br><span class="line"><span class="comment"># Leave commented to accept the defaults for your choice of BLAS</span></span><br><span class="line"><span class="comment"># (which should work)!</span></span><br><span class="line"><span class="comment"># BLAS_INCLUDE := /path/to/your/blas</span></span><br><span class="line"><span class="comment"># BLAS_LIB := /path/to/your/blas</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Homebrew puts openblas in a directory that is not on the standard search path</span></span><br><span class="line"><span class="comment"># BLAS_INCLUDE := $(shell brew --prefix openblas)/include</span></span><br><span class="line"><span class="comment"># BLAS_LIB := $(shell brew --prefix openblas)/lib</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># This is required only if you will compile the matlab interface.</span></span><br><span class="line"><span class="comment"># MATLAB directory should contain the mex binary in /bin.</span></span><br><span class="line"><span class="comment"># MATLAB_DIR := /usr/local</span></span><br><span class="line"><span class="comment"># MATLAB_DIR := /Applications/MATLAB_R2012b.app</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># <span class="doctag">NOTE:</span> this is required only if you will compile the python interface.</span></span><br><span class="line"><span class="comment"># We need to be able to find Python.h and numpy/arrayobject.h.</span></span><br><span class="line">PYTHON_INCLUDE := /usr/<span class="keyword">include</span>/python2.7 \</span><br><span class="line">/usr/lib/python2.7/dist-packages/numpy/core/<span class="keyword">include</span></span><br><span class="line"><span class="comment"># Anaconda Python distribution is quite popular. Include path:</span></span><br><span class="line"><span class="comment"># Verify anaconda location, sometimes it's in root.</span></span><br><span class="line"><span class="comment"># ANACONDA_HOME := $(HOME)/anaconda</span></span><br><span class="line"><span class="comment"># PYTHON_INCLUDE := $(ANACONDA_HOME)/include \</span></span><br><span class="line"><span class="comment"># $(ANACONDA_HOME)/include/python2.7 \</span></span><br><span class="line"><span class="comment"># $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment to use Python 3 (default is Python 2)</span></span><br><span class="line"><span class="comment"># PYTHON_LIBRARIES := boost_python3 python3.5m</span></span><br><span class="line"><span class="comment"># PYTHON_INCLUDE := /usr/include/python3.5m \</span></span><br><span class="line"><span class="comment">#                 /usr/lib/python3.5/dist-packages/numpy/core/include</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># We need to be able to find libpythonX.X.so or .dylib.</span></span><br><span class="line">PYTHON_LIB := /usr/lib</span><br><span class="line"><span class="comment"># PYTHON_LIB := $(ANACONDA_HOME)/lib</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Homebrew installs numpy in a non standard path (keg only)</span></span><br><span class="line"><span class="comment"># PYTHON_INCLUDE += $(dir $(shell python -c 'import numpy.core; print(numpy.core.__file__)'))/include</span></span><br><span class="line"><span class="comment"># PYTHON_LIB += $(shell brew --prefix numpy)/lib</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment to support layers written in Python (will link against Python libs)</span></span><br><span class="line"><span class="comment"># WITH_PYTHON_LAYER := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Whatever else you find you need goes here.</span></span><br><span class="line">INCLUDE_DIRS := <span class="variable">$(PYTHON_INCLUDE)</span> /usr/local/<span class="keyword">include</span></span><br><span class="line">LIBRARY_DIRS := <span class="variable">$(PYTHON_LIB)</span> /usr/local/lib /usr/lib</span><br><span class="line"></span><br><span class="line"><span class="comment"># If Homebrew is installed at a non standard location (for example your home directory) and you use it for general dependencies</span></span><br><span class="line"><span class="comment"># INCLUDE_DIRS += $(shell brew --prefix)/include</span></span><br><span class="line"><span class="comment"># LIBRARY_DIRS += $(shell brew --prefix)/lib</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># NCCL acceleration switch (uncomment to build with NCCL)</span></span><br><span class="line"><span class="comment"># https://github.com/NVIDIA/nccl (last tested version: v1.2.3-1+cuda8.0)</span></span><br><span class="line"><span class="comment"># USE_NCCL := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment to use `pkg-config` to specify OpenCV library paths.</span></span><br><span class="line"><span class="comment"># (Usually not necessary -- OpenCV libraries are normally installed in one of the above $LIBRARY_DIRS.)</span></span><br><span class="line"><span class="comment"># USE_PKG_CONFIG := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># N.B. both build and distribute dirs are cleared on `make clean`</span></span><br><span class="line">BUILD_DIR := build</span><br><span class="line">DISTRIBUTE_DIR := distribute</span><br><span class="line"></span><br><span class="line"><span class="comment"># Uncomment for debugging. Does not work on OSX due to https://github.com/BVLC/caffe/issues/171</span></span><br><span class="line"><span class="comment"># DEBUG := 1</span></span><br><span class="line"></span><br><span class="line"><span class="comment"># The ID of the GPU that 'make runtest' will use to run unit tests.</span></span><br><span class="line">TEST_GPUID := 0</span><br><span class="line"></span><br><span class="line"><span class="comment"># enable pretty build (comment to see full commands)</span></span><br><span class="line">Q ?= @</span><br></pre></td></tr></table></figure></li><li><p>保存退出，编译</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">make all -j4</span><br><span class="line">make test -j4</span><br><span class="line">make runtest -j4</span><br></pre></td></tr></table></figure></li><li><p>没有报错，成功</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">[----------] Global test environment tear-down</span><br><span class="line">[==========] 2101 tests from 277 test cases ran. (371248 ms total)</span><br><span class="line">[  PASSED  ] 2101 tests.</span><br></pre></td></tr></table></figure></li></ul></li></ul><h4 id="修改后的组件版本"><a href="#修改后的组件版本" class="headerlink" title="修改后的组件版本"></a>修改后的组件版本</h4><ul><li><strong>Ubuntu</strong><ul><li>14.04</li></ul></li></ul><ul><li><strong>CUDA</strong><ul><li>8.0</li></ul></li><li><strong>cuDNN</strong><ul><li>cudnn-8.0-linux-x64-v5.0-ga.tgz</li></ul></li><li><strong>opencv</strong><ul><li>3.2</li></ul></li><li><strong>python</strong><ul><li>Python 2.7.6</li></ul></li><li><strong>caffe</strong><ul><li>1.0</li></ul></li></ul><h3 id="一点心得"><a href="#一点心得" class="headerlink" title="一点心得"></a>一点心得</h3><ul><li><p>make 编译的时候可以通过</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">make -j4</span><br></pre></td></tr></table></figure><p>指令指定多线程编译，可以很大程度加快编译速度。具体可以开多少看自己机器配置</p></li><li><p>看好OpenCV支持的CUDA版本</p></li><li><p>看好caffe支持的cuDNN版本</p></li><li><p>每个人硬件、环境的不同都会导致踩到不同的坑。这个就多 Google 吧！</p></li></ul><blockquote><p>部分参考 <a href="http://www.cnblogs.com/wm123/p/5385940.html" target="_blank" rel="noopener"><strong>Ubuntu 14.04上安装caffe</strong></a></p></blockquote>]]></content>
    
    
      
      
    <summary type="html">&lt;h3 id=&quot;组件-amp-版本&quot;&gt;&lt;a href=&quot;#组件-amp-版本&quot; class=&quot;headerlink&quot; title=&quot;组件&amp;amp;版本&quot;&gt;&lt;/a&gt;组件&amp;amp;版本&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Ubuntu&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
</summary>
      
    
    
    
    <category term="安装记录" scheme="https://www.holmeyoung.com/categories/安装记录/"/>
    
    
    <category term="linux" scheme="https://www.holmeyoung.com/tags/linux/"/>
    
    <category term="机器学习" scheme="https://www.holmeyoung.com/tags/机器学习/"/>
    
    <category term="caffe" scheme="https://www.holmeyoung.com/tags/caffe/"/>
    
  </entry>
  
  <entry>
    <title>git 使用小结</title>
    <link href="https://www.holmeyoung.com/blog/git-summary/"/>
    <id>https://www.holmeyoung.com/blog/git-summary/</id>
    <published>2018-03-23T12:05:20.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h2 id="新建-GitLab-仓库时官方给出的命令"><a href="#新建-GitLab-仓库时官方给出的命令" class="headerlink" title="新建 GitLab 仓库时官方给出的命令"></a>新建 GitLab 仓库时官方给出的命令</h2><h3 id="Git-global-setup"><a href="#Git-global-setup" class="headerlink" title="Git global setup"></a><strong>Git global setup</strong></h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git config --global user.name <span class="string">"YourName"</span></span><br><span class="line">git config --global user.email <span class="string">"YourEmail"</span></span><br></pre></td></tr></table></figure><h3 id="Create-a-new-repository"><a href="#Create-a-new-repository" class="headerlink" title="Create a new repository"></a>Create a new repository</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git <span class="built_in">clone</span> https://gitlab.com/holmeyoung/num_recog.git</span><br><span class="line"><span class="built_in">cd</span> num_recog</span><br><span class="line">touch README.md</span><br><span class="line">git add README.md</span><br><span class="line">git commit -m <span class="string">"add README"</span></span><br><span class="line">git push -u origin master</span><br></pre></td></tr></table></figure><h3 id="Existing-folder"><a href="#Existing-folder" class="headerlink" title="Existing folder"></a><strong>Existing folder</strong></h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">cd</span> existing_folder</span><br><span class="line">git init</span><br><span class="line">git remote add origin https://gitlab.com/holmeyoung/num_recog.git</span><br><span class="line">git add .</span><br><span class="line">git commit -m <span class="string">"Initial commit"</span></span><br><span class="line">git push -u origin master</span><br></pre></td></tr></table></figure><h3 id="Existing-Git-repository"><a href="#Existing-Git-repository" class="headerlink" title="Existing Git repository"></a><strong>Existing Git repository</strong></h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">cd</span> existing_repo</span><br><span class="line">git remote rename origin old-origin</span><br><span class="line">git remote add origin https://gitlab.com/holmeyoung/num_recog.git</span><br><span class="line">git push -u origin --all</span><br><span class="line">git push -u origin --tags</span><br></pre></td></tr></table></figure><h2 id="新建-GitHab-仓库时官方给出的命令"><a href="#新建-GitHab-仓库时官方给出的命令" class="headerlink" title="新建 GitHab 仓库时官方给出的命令"></a>新建 GitHab 仓库时官方给出的命令</h2><h3 id="…or-create-a-new-repository-on-the-command-line"><a href="#…or-create-a-new-repository-on-the-command-line" class="headerlink" title="…or create a new repository on the command line"></a>…or create a new repository on the command line</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">echo</span> <span class="string">"# tmp"</span> &gt;&gt; README.md</span><br><span class="line">git init</span><br><span class="line">git add README.md</span><br><span class="line">git commit -m <span class="string">"first commit"</span></span><br><span class="line">git remote add origin git@github.com:Holmeyoung/tmp.git</span><br><span class="line">git push -u origin master</span><br></pre></td></tr></table></figure><h3 id="…or-push-an-existing-repository-from-the-command-line"><a href="#…or-push-an-existing-repository-from-the-command-line" class="headerlink" title="…or push an existing repository from the command line"></a>…or push an existing repository from the command line</h3><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git remote add origin git@github.com:Holmeyoung/tmp.git</span><br><span class="line">git push -u origin master</span><br></pre></td></tr></table></figure><h3 id="…or-import-code-from-another-repository"><a href="#…or-import-code-from-another-repository" class="headerlink" title="…or import code from another repository"></a>…or import code from another repository</h3><p>You can initialize this repository with code from a Subversion, Mercurial, or TFS project.</p><h2 id="Git-免密码"><a href="#Git-免密码" class="headerlink" title="Git 免密码"></a>Git 免密码</h2><h3 id="Git-HTTPS-方式传输文件免密码"><a href="#Git-HTTPS-方式传输文件免密码" class="headerlink" title="Git HTTPS 方式传输文件免密码"></a><strong>Git <code>HTTPS</code> 方式传输文件免密码</strong></h3><h4 id="cache"><a href="#cache" class="headerlink" title="cache"></a><strong>cache</strong></h4><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git config --global credential.helper cache</span><br></pre></td></tr></table></figure><blockquote><p>或直接修改 <strong><code>~/.gitconfig</code></strong></p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">&gt; [credential]</span><br><span class="line">&gt;     helper = cache --timeout=3600</span><br><span class="line">&gt;</span><br></pre></td></tr></table></figure></blockquote><blockquote><p>采用 <code>cache</code> 形式，密码会临时保存在 <code>~/git-credential-cache</code> 文件夹下</p></blockquote><h4 id="store"><a href="#store" class="headerlink" title="store"></a><strong>store</strong></h4><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git config --global credential.helper store</span><br></pre></td></tr></table></figure><blockquote><p>或直接修改 <strong><code>~/.gitconfig</code></strong></p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">&gt; [credential]</span><br><span class="line">&gt;     helper = store</span><br><span class="line">&gt;</span><br></pre></td></tr></table></figure></blockquote><blockquote><p>采用 <code>store</code> 形式，密码会永久保存在 <code>~/.git-credentials</code> 文件中</p></blockquote><h3 id="Git-SSH-方式传输文件免密码"><a href="#Git-SSH-方式传输文件免密码" class="headerlink" title="Git SSH 方式传输文件免密码"></a>Git <code>SSH</code> 方式传输文件免密码</h3><p><strong>配置密钥即可</strong></p><h2 id="分支管理"><a href="#分支管理" class="headerlink" title="分支管理"></a>分支管理</h2><h3 id="远程服务器（远程仓库）"><a href="#远程服务器（远程仓库）" class="headerlink" title="远程服务器（远程仓库）"></a>远程服务器（远程仓库）</h3><h4 id="远程服务器（远程仓库）是什么"><a href="#远程服务器（远程仓库）是什么" class="headerlink" title="远程服务器（远程仓库）是什么"></a>远程服务器（远程仓库）是什么</h4><p>有命令<br><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git push origin master:master</span><br></pre></td></tr></table></figure></p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git pull origin master:master</span><br></pre></td></tr></table></figure><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git fetch origin</span><br></pre></td></tr></table></figure><p><strong>这里的 <code>origin</code> 究竟是什么</strong></p><p>有命令<br><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git remote add origin git@code.ziroom.com:shuly/wiki.git</span><br></pre></td></tr></table></figure></p><blockquote><p>所以 </p><p><code>origin --&gt; git@code.ziroom.com:shuly/wiki.git</code></p><p><code>origin</code> 即为指向远程服务器 <a href="mailto:`git@code.ziroom.com" target="_blank" rel="noopener">`git@code.ziroom.com</a>:shuly/wiki.git` 的远程服务器名</p></blockquote><h4 id="多个远程服务器"><a href="#多个远程服务器" class="headerlink" title="多个远程服务器"></a>多个远程服务器</h4><p>假设还有另一个服务器 <a href="mailto:`git@github.com" target="_blank" rel="noopener">`git@github.com</a><code>。可以用第二章中提到的</code>git remote add<code>命令把它加为当前项目的远程分支之一。我们把它命名为</code>origin_dev` ，以便代替完整的 Git URL 以方便使用。</p><p>执行命令<br><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git remote add origin_dev git@github.com:Holmeyoung/wiki.git</span><br></pre></td></tr></table></figure></p><p>现在你可以用 <code>git fetch origin_dev</code> 来获取另一台服务器上你还没有的数据了。执行该命令会创建一个名为 <code>origin_dev/master</code> 的远程分支，指向 <code>origin_dev</code> 服务器上 <code>master</code> 分支所在的提交对象</p><h3 id="远程分支"><a href="#远程分支" class="headerlink" title="远程分支"></a>远程分支</h3><h4 id="删除远程分支"><a href="#删除远程分支" class="headerlink" title="删除远程分支"></a>删除远程分支</h4><p>如果不再需要某个远程分支了，比如搞定了某个特性并把它合并进了远程的 <code>master</code> 分支（或任何其他存放稳定代码的分支），可以用这个非常无厘头的语法来删除它：<code>git push [远程名] :[分支名]</code> 。如果想在服务器上删除 develop 分支，运行下面的命令：<br><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git push origin :develop</span><br></pre></td></tr></table></figure></p><h3 id="本地分支"><a href="#本地分支" class="headerlink" title="本地分支"></a>本地分支</h3><ul><li><p>创建一个叫做 <code>feature_x</code> 的分支，并切换过去：</p>  <figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git checkout -b feature_x</span><br></pre></td></tr></table></figure></li><li><p>切换回主分支：</p>  <figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git checkout master</span><br></pre></td></tr></table></figure></li><li><p>再把新建的分支删掉：</p>  <figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git branch -d feature_x</span><br></pre></td></tr></table></figure></li><li><p>除非你将分支推送到远端仓库，不然该分支就是 不为他人所见的：</p>  <figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git push origin &lt;branch&gt;</span><br></pre></td></tr></table></figure></li></ul><h2 id="多人协同完整流程"><a href="#多人协同完整流程" class="headerlink" title="多人协同完整流程"></a>多人协同完整流程</h2><ul><li>GitLab 新建远程分支 <code>develop</code></li></ul><blockquote><p>这一步的意义在于防止直接操作 <code>master</code> 分支造成无法挽回的后果</p></blockquote><ul><li><p>下载新的远程分支</p>  <figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git fetch origin</span><br></pre></td></tr></table></figure><blockquote><p>值得注意的是，在 <code>fetch</code> 操作下载好新的远程分支之后，你仍然无法在本地编辑该远程仓库中的分支。换句话说，在本例中，你不会有一个新的 <code>develop</code> 分支，有的只是一个你无法移动的 <code>origin/develop</code> 指针</p></blockquote></li><li><p>操作下载的远程分支</p><ul><li><p>如果要把该远程分支的内容合并到当前分支，可以运行</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git merge origin/develop</span><br></pre></td></tr></table></figure></li><li><p>如果想要一份自己的 serverfix 来开发，可以在远程分支的基础上分化出一个新的分支来</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git checkout -b develop origin/develop</span><br></pre></td></tr></table></figure><p>这会切换到新建的 <code>develop</code> 本地分支，其内容同远程分支 <code>origin/develop</code> 一致，这样你就可以在里面继续开发了。</p></li></ul><blockquote><p>推荐</p></blockquote></li><li><p>新建本地用户分支并开发</p>  <figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git checkout -b holmeyoung</span><br></pre></td></tr></table></figure><p>  在 <code>holmeyoung</code> 分支开发完毕后</p><ul><li><p>提交到缓存区</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git add --all</span><br></pre></td></tr></table></figure></li><li><p>提交到 HEAD</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git commit -m <span class="string">"holmeyoung change"</span></span><br></pre></td></tr></table></figure></li></ul></li><li><p>合并改动到 <code>develop</code> 分支</p><ul><li><p>跳到 <code>develop</code> 分支</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git checkout develop</span><br></pre></td></tr></table></figure></li><li><p>将 <code>holmeyoung</code> 分支的改动 merge 到 <code>develop</code> 分支</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git merge holmeyoung</span><br></pre></td></tr></table></figure></li><li><p>推送本地 <code>develop</code> 分支到远程 <code>develop</code> 分支</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git push origin develop:develop</span><br></pre></td></tr></table></figure></li></ul></li><li><p>合并改动到 <code>master</code> 分支</p><ul><li><p>跳到 <code>master</code> 分支</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git checkout master</span><br></pre></td></tr></table></figure></li><li><p>将 <code>develop</code> 分支的改动 merge 到 <code>master</code> 分支</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git merge develop</span><br></pre></td></tr></table></figure></li><li><p>推送本地 <code>master</code> 分支到远程 <code>master</code> 分支</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line">git push origin master:master</span><br></pre></td></tr></table></figure></li></ul></li></ul><h2 id="REF"><a href="#REF" class="headerlink" title="REF"></a><code>REF</code></h2><blockquote><p><a href="http://www.bootcss.com/p/git-guide/" target="_blank" rel="noopener">git - 简易指南</a></p><p><a href="http://www.ruanyifeng.com/blog/2014/06/git_remote.html" target="_blank" rel="noopener">Git远程操作详解</a></p><p><a href="https://git-scm.com/book/zh/v2" target="_blank" rel="noopener">Git Book</a></p></blockquote>]]></content>
    
    
      
      
    <summary type="html">&lt;h2 id=&quot;新建-GitLab-仓库时官方给出的命令&quot;&gt;&lt;a href=&quot;#新建-GitLab-仓库时官方给出的命令&quot; class=&quot;headerlink&quot; title=&quot;新建 GitLab 仓库时官方给出的命令&quot;&gt;&lt;/a&gt;新建 GitLab 仓库时官方给出的命令&lt;/h2&gt;&lt;</summary>
      
    
    
    
    <category term="技术总结" scheme="https://www.holmeyoung.com/categories/技术总结/"/>
    
    
    <category term="git" scheme="https://www.holmeyoung.com/tags/git/"/>
    
  </entry>
  
  <entry>
    <title>anaconda 安装记录</title>
    <link href="https://www.holmeyoung.com/blog/anaconda-linux/"/>
    <id>https://www.holmeyoung.com/blog/anaconda-linux/</id>
    <published>2018-03-08T06:37:15.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h3 id="Anaconda-是什么？"><a href="#Anaconda-是什么？" class="headerlink" title="Anaconda 是什么？"></a>Anaconda 是什么？</h3><p>Anaconda 是一个可用于科学计算的 Python 发行版，支持 Linux、Mac、Windows系统，内置了常用的科学计算包。它解决了官方 Python 的两大痛点。</p><ul><li>第一：提供了包管理功能，Windows 平台安装第三方包经常失败的场景得以解决</li><li>第二：提供环境管理的功能，功能类似 Virtualenv，解决了多版本Python并存、切换的问题</li></ul><h3 id="下载-Anaconda"><a href="#下载-Anaconda" class="headerlink" title="下载 Anaconda"></a>下载 Anaconda</h3><p>直接在<a href="https://www.anaconda.com/download/#macos" target="_blank" rel="noopener">官网下载</a>安装包，我是下载的 <code>2.7</code> 版本。下载后直接安装并选择默认配置就好。</p><p>安装过程会提示安装 <code>visual studio code</code> 选择安装，如果说为什么的话</p><blockquote><p>比sublime开源、比atom更快、比webstorm更轻</p></blockquote><h3 id="环境变量配置"><a href="#环境变量配置" class="headerlink" title="环境变量配置"></a>环境变量配置</h3><h4 id="如何配置"><a href="#如何配置" class="headerlink" title="如何配置"></a>如何配置</h4><p>刚下载下来在 <code>iterm2</code> 中输入 <code>conda</code> 会显示 <code>zsh command not found</code></p><p>因为是我用的是 <code>iterm2</code> 且配置了 <code>oh my zsh</code> ，错误显示的又是 <code>zsh</code> 所以我们需要在 <code>~/.zsh</code> 即 <code>oh my zsh</code> 的配置文件中添加 <strong>anaconda</strong> 的安装路径。</p><p>即添加</p><figure class="highlight sh"><table><tr><td class="code"><pre><span class="line"><span class="built_in">export</span> PATH=<span class="string">"/anaconda2/bin:<span class="variable">$PATH</span>"</span></span><br></pre></td></tr></table></figure><blockquote><p>需要注意自己用户目录下的是 <code>anaconda2</code> 还是 <code>anaconda</code> </p></blockquote><h4 id="检查配置"><a href="#检查配置" class="headerlink" title="检查配置"></a>检查配置</h4><p>输入 <code>which conda</code> ,返回具体的路径即为成功！</p><h3 id="conda-命令介绍"><a href="#conda-命令介绍" class="headerlink" title="conda 命令介绍"></a>conda 命令介绍</h3><h4 id="关于新建Python"><a href="#关于新建Python" class="headerlink" title="关于新建Python"></a>关于新建Python</h4><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">#</span><span class="bash"> 查看帮助</span></span><br><span class="line">conda -h</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 基于python3.6版本创建一个名字为python36的环境</span></span><br><span class="line">conda create --name python36 python=3.6</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 激活此环境 linux/mac</span></span><br><span class="line">source activate python36 </span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 再来检查python版本，显示是 3.6</span></span><br><span class="line">python -V</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 退出当前环境到系统根目录</span></span><br><span class="line">source deactivate</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 删除该环境</span></span><br><span class="line">conda remove -n python36 --all</span><br><span class="line"></span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 查看所有安装的环境</span></span><br><span class="line">conda info -e</span><br><span class="line"></span><br><span class="line">➜  ~ conda info -e</span><br><span class="line"><span class="meta">#</span><span class="bash"> conda environments:</span></span><br><span class="line"><span class="meta">#</span></span><br><span class="line">base                  *  /Users/holmeyoung/anaconda2</span><br><span class="line">python36                 /Users/holmeyoung/anaconda2/envs/python36</span><br></pre></td></tr></table></figure><h4 id="关于conda-的包管理"><a href="#关于conda-的包管理" class="headerlink" title="关于conda 的包管理"></a>关于conda 的包管理</h4><p>和 <code>pip</code> 是一样的，当然你选择 <code>pip</code> 来安装包也是没问题的</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">#</span><span class="bash"> 安装 matplotlib </span></span><br><span class="line">conda install matplotlib</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 查看已安装的包</span></span><br><span class="line">conda list</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 包更新</span></span><br><span class="line">conda update matplotlib</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 删除包</span></span><br><span class="line">conda remove matplotlib</span><br></pre></td></tr></table></figure><blockquote><p>在 conda 中 <strong>anything is a package</strong>。conda 本身可以看作是一个包，python 环境可以看作是一个包，anaconda 也可以看作是一个包，因此除了普通的第三方包支持更新之外……</p></blockquote><h4 id="conda的自我更新"><a href="#conda的自我更新" class="headerlink" title="conda的自我更新"></a>conda的自我更新</h4><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line"><span class="meta">#</span><span class="bash"> 更新conda本身</span></span><br><span class="line">conda update conda</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 更新anaconda 应用</span></span><br><span class="line">conda update anaconda</span><br><span class="line"></span><br><span class="line"><span class="meta">#</span><span class="bash"> 更新python，假设当前python环境是3.6.1，而最新版本是3.6.2，那么就会升级到3.6.2</span></span><br><span class="line">conda update python</span><br></pre></td></tr></table></figure><h4 id="anaconda卸载"><a href="#anaconda卸载" class="headerlink" title="anaconda卸载"></a>anaconda卸载</h4><p>由于Anaconda的安装文件都包含在一个目录中，所以直接将该目录删除即可</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">rm -rf anaconda2</span><br></pre></td></tr></table></figure><blockquote><p>注意清理 <code>.zshrc</code> 以及 <code>~/.bash_profile</code> 中的Anaconda路径</p></blockquote><h3 id="部分参考"><a href="#部分参考" class="headerlink" title="部分参考"></a>部分参考</h3><p><a href="https://foofish.net/anaconda-install.html" target="_blank" rel="noopener">Anaconda 入门安装教程</a></p><p><a href="http://www.cnblogs.com/caiyutong/p/7002764.html" target="_blank" rel="noopener">Mac下安装Anaconda和环境变量设置，以及升级和卸载</a></p>]]></content>
    
    
      
      
    <summary type="html">&lt;h3 id=&quot;Anaconda-是什么？&quot;&gt;&lt;a href=&quot;#Anaconda-是什么？&quot; class=&quot;headerlink&quot; title=&quot;Anaconda 是什么？&quot;&gt;&lt;/a&gt;Anaconda 是什么？&lt;/h3&gt;&lt;p&gt;Anaconda 是一个可用于科学计算的 Pytho</summary>
      
    
    
    
    <category term="安装记录" scheme="https://www.holmeyoung.com/categories/安装记录/"/>
    
    
    <category term="python" scheme="https://www.holmeyoung.com/tags/python/"/>
    
    <category term="linux" scheme="https://www.holmeyoung.com/tags/linux/"/>
    
  </entry>
  
  <entry>
    <title>实现 ssh 免密码登录服务器</title>
    <link href="https://www.holmeyoung.com/blog/ssh-no-password/"/>
    <id>https://www.holmeyoung.com/blog/ssh-no-password/</id>
    <published>2018-01-05T12:13:52.000Z</published>
    <updated>2022-10-06T07:36:56.304Z</updated>
    
    <content type="html"><![CDATA[<h3 id="修改-ssh-config权限"><a href="#修改-ssh-config权限" class="headerlink" title="修改 ~/.ssh/config权限"></a>修改 <code>~/.ssh/config</code>权限</h3><p><code>~/.ssh/config</code>是 ssh 自己的配置文件，在修改文件之前一定要修改文件的权限以及其所属于的用户，否则会出现问题。</p><p>执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">chmod 600 ~/.ssh/config</span><br></pre></td></tr></table></figure><p>否则可能出现</p><blockquote><p><strong>Bad owner or permissions on /Users/mac/.ssh/config</strong></p></blockquote><p>执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">chown $USER ~/.ssh/config</span><br></pre></td></tr></table></figure><p>否则可能出现</p><blockquote><p><strong>ssh: Could not resolve hostname centos: nodename nor servname provided, or not known</strong></p></blockquote><p>确定 <strong>USER</strong> 的方法是执行 </p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">whoami</span><br></pre></td></tr></table></figure><p> 如我的结果是 <code>mac</code> 即执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">chown mac ~/.ssh/config</span><br></pre></td></tr></table></figure><h3 id="修改-ssh-config"><a href="#修改-ssh-config" class="headerlink" title="修改 ~/.ssh/config"></a>修改 <code>~/.ssh/config</code></h3><p>执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">vim ~/.ssh/config</span><br></pre></td></tr></table></figure><p>将下面内容添加进去</p><figure class="highlight powershell"><table><tr><td class="code"><pre><span class="line">Host 起个名字</span><br><span class="line">  HostName 你的服务器IP</span><br><span class="line">  Port ssh端口</span><br><span class="line">  User ssh用户</span><br><span class="line">  IdentityFile 私钥文件路径(~/.ssh/id_rsa)</span><br></pre></td></tr></table></figure><p>保存退出</p><h3 id="设置免密码登录"><a href="#设置免密码登录" class="headerlink" title="设置免密码登录"></a>设置免密码登录</h3><h4 id="本地生成公钥和私钥文件"><a href="#本地生成公钥和私钥文件" class="headerlink" title="本地生成公钥和私钥文件"></a>本地生成公钥和私钥文件</h4><p>在 Mac OSX 终端（或 iTerm 2 等）中进入 <code>~/.ssh</code> 目录，输入</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">ssh-keygen -t rsa</span><br></pre></td></tr></table></figure><p>会问一些问题，一路回车就好啦。这样就会在 <code>~/.ssh</code> 目录中生成 <code>id_rsa</code>（私钥） 和 <code>id_rsa.pub</code>（公钥）文件</p><h4 id="服务器端配置"><a href="#服务器端配置" class="headerlink" title="服务器端配置"></a>服务器端配置</h4><p>进入服务器端 <code>~/.ssh</code> 目录，执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">vim authorized_keys</span><br></pre></td></tr></table></figure><p>没有的话新建就好了</p><p>打开Mac端 <code>~/.ssh</code> 目录中生成的 <code>id_rsa.pub</code> 文件，将其中的内容复制到服务器端的 <code>authorized_keys</code> 文件中</p><p>保存退出即可</p><blockquote><p>要是有其他用户已经新建了这个文件并且写入了内容，只需要在文件最后追加即可</p></blockquote><h4 id="修改文件权限"><a href="#修改文件权限" class="headerlink" title="修改文件权限"></a>修改文件权限</h4><p>确保下文件权限正常，在服务器端执行</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">chmod 700 ~/.ssh/</span><br><span class="line">chmod 600 ~/.ssh/authorized_keys</span><br></pre></td></tr></table></figure><p>如果无法连接，请在服务器的 <code>/etc/ssh/sshd_config</code> 文件中查看 <code>PubkeyAuthentication</code> 的值是否为 <strong>yes</strong> ，如果不是，请修改为 <strong>yes</strong> ，并使用以下命令重启 ssh 服务</p><figure class="highlight powershell"><table><tr><td class="code"><pre><span class="line">/etc/init.d/ssh restart</span><br></pre></td></tr></table></figure><p>如果需要禁止密码登录，同样可以编辑 <code>/etc/ssh/sshd_config</code> 文件中 <code>PasswordAuthentication</code> 的值为 <strong>no</strong> 并重启 ssh 服务</p><h3 id="最终效果"><a href="#最终效果" class="headerlink" title="最终效果"></a>最终效果</h3><p>例如你在 <code>~/.ssh/config</code> 中写入的是</p><figure class="highlight powershell"><table><tr><td class="code"><pre><span class="line">Host centos</span><br><span class="line">  HostName <span class="number">8.8</span>.<span class="number">8.8</span></span><br><span class="line">  Port <span class="number">22</span></span><br><span class="line">  User root</span><br><span class="line">  IdentityFile ~/.ssh/id_rsa</span><br></pre></td></tr></table></figure><p>那么在终端或者 iterm 中输入</p><figure class="highlight shell"><table><tr><td class="code"><pre><span class="line">ssh centos</span><br></pre></td></tr></table></figure><p>即可直接登录服务器啦</p>]]></content>
    
    
      
      
    <summary type="html">&lt;h3 id=&quot;修改-ssh-config权限&quot;&gt;&lt;a href=&quot;#修改-ssh-config权限&quot; class=&quot;headerlink&quot; title=&quot;修改 ~/.ssh/config权限&quot;&gt;&lt;/a&gt;修改 &lt;code&gt;~/.ssh/config&lt;/code&gt;权限&lt;/h3&gt;&lt;p</summary>
      
    
    
    
    <category term="技术总结" scheme="https://www.holmeyoung.com/categories/技术总结/"/>
    
    
    <category term="linux" scheme="https://www.holmeyoung.com/tags/linux/"/>
    
    <category term="ssh" scheme="https://www.holmeyoung.com/tags/ssh/"/>
    
  </entry>
  
</feed>