https://genffy.com/series/rag%E7%B3%BB%E5%88%97/ Recent content in RAG系列 on 进击的页面仔 Hugo -- 0.147.0 zh 2026 进击的页面仔 Fri, 26 Jan 2024 18:29:40 +0800 https://genffy.com/posts/nvidia-cuda-driver-install/ Fri, 26 Jan 2024 18:29:40 +0800 https://genffy.com/posts/nvidia-cuda-driver-install/ <p>最近想着将公司里 <code>Confluence</code> 的内容拿来跑个智能知识库系统练练手，学习的是 <a href="https://twitter.com/Tisoga">Jiayuan (Forrest)</a> 大大的 <a href="https://twitter.com/Tisoga/status/1731478506465636749">http://devv.ai 是如何构建高效的 RAG 系统的</a> 这篇文章，找了一圈最终直接用的项目是 <a href="https://github.com/chatchat-space/Langchain-Chatchat">Langchain-Chatchat</a>，结果在一启动的时候就报错了个错，<code>UserWarning: CUDA initialization: The NVIDIA driver on your system is too old (found version 11080).</code>，很明显，它是告诉我显卡驱动版本太低了，需要更新。这里其实有两种方式：</p> <ul> <li>降低 <code>pytorch</code> 的版本</li> <li>升级显卡驱动</li> </ul> <p>先确定本机的驱动信息，以及项目的 pytorch 版本所对应的 cuda 版本信息。</p> <h2 id="基本信息">基本信息</h2> <h3 id="nvidia-smi">nvidia-smi</h3> <blockquote> <p>最开始的时候是单卡，初始的信息记录</p></blockquote> <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-bash" data-lang="bash"><span class="line"><span class="cl">$ lspci <span class="p">|</span> grep -i nvidia </span></span><span class="line"><span class="cl">af:00.0 3D controller: NVIDIA Corporation GP102GL <span class="o">[</span>Tesla P40<span class="o">]</span> <span class="o">(</span>rev a1<span class="o">)</span> </span></span><span class="line"><span class="cl"> </span></span><span class="line"><span class="cl">$ nvidia-smi </span></span><span class="line"><span class="cl">+-----------------------------------------------------------------------------+ </span></span><span class="line"><span class="cl"><span class="p">|</span> NVIDIA-SMI 520.61.05 Driver Version: 520.61.05 CUDA Version: 11.8 <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span>-------------------------------+----------------------+----------------------+ </span></span><span class="line"><span class="cl"><span class="p">|</span> GPU Name Persistence-M<span class="p">|</span> Bus-Id Disp.A <span class="p">|</span> Volatile Uncorr. ECC <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> Fan Temp Perf Pwr:Usage/Cap<span class="p">|</span> Memory-Usage <span class="p">|</span> GPU-Util Compute M. <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> <span class="p">|</span> <span class="p">|</span> MIG M. <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span><span class="o">===============================</span>+<span class="o">======================</span>+<span class="o">======================</span><span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> <span class="m">0</span> Tesla P40 Off <span class="p">|</span> 00000000:AF:00.0 Off <span class="p">|</span> <span class="m">0</span> <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> N/A 28C P8 9W / 250W <span class="p">|</span> 9MiB / 23040MiB <span class="p">|</span> 0% Default <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> <span class="p">|</span> <span class="p">|</span> N/A <span class="p">|</span> </span></span><span class="line"><span class="cl">+-------------------------------+----------------------+----------------------+ </span></span><span class="line"><span class="cl"> </span></span><span class="line"><span class="cl">+-----------------------------------------------------------------------------+ </span></span><span class="line"><span class="cl"><span class="p">|</span> Processes: <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> GPU GI CI PID Type Process name GPU Memory <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> ID ID Usage <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span><span class="o">=============================================================================</span><span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> <span class="m">0</span> N/A N/A <span class="m">2113</span> G /usr/lib/xorg/Xorg 4MiB <span class="p">|</span> </span></span><span class="line"><span class="cl"><span class="p">|</span> <span class="m">0</span> N/A N/A <span class="m">7618</span> G /usr/lib/xorg/Xorg 4MiB <span class="p">|</span> </span></span><span class="line"><span class="cl">+-----------------------------------------------------------------------------+ </span></span><span class="line"><span class="cl"> </span></span><span class="line"><span class="cl">$ nvcc --version </span></span><span class="line"><span class="cl">nvcc: NVIDIA <span class="o">(</span>R<span class="o">)</span> Cuda compiler driver </span></span><span class="line"><span class="cl">Copyright <span class="o">(</span>c<span class="o">)</span> 2005-2022 NVIDIA Corporation </span></span><span class="line"><span class="cl">Built on Wed_Sep_21_10:33:58_PDT_2022 </span></span><span class="line"><span class="cl">Cuda compilation tools, release 11.8, V11.8.89 </span></span><span class="line"><span class="cl">Build cuda_11.8.r11.8/compiler.31833905_0 </span></span></code></pre></div><h3 id="项目依赖的-cuda-版本">项目依赖的 CUDA 版本</h3> <p>通过简单的脚本能看到项目依赖的 <code>cuda</code> 版本为 <code>12.1</code></p>