https://www.knightli.com/en/tags/huggingface-cli/ Recent content in Huggingface-Cli on KnightLi Blog Hugo -- gohugo.io en Fri, 17 Apr 2026 14:48:04 +0800 https://www.knightli.com/en/2026/04/17/huggingface-cli-download-default-cache-path/ Fri, 17 Apr 2026 14:48:04 +0800 https://www.knightli.com/en/2026/04/17/huggingface-cli-download-default-cache-path/ <p>If you download a model with the Hugging Face command-line tool, for example:</p> <div class="highlight"><div class="chroma"> <table class="lntable"><tr><td class="lntd"> <pre tabindex="0" class="chroma"><code><span class="lnt">1 </span></code></pre></td> <td class="lntd"> <pre tabindex="0" class="chroma"><code class="language-bash" data-lang="bash"><span class="line"><span class="cl">huggingface-cli download &lt;repo-id&gt; </span></span></code></pre></td></tr></table> </div> </div><p>the downloaded model files usually do not appear in the current directory. Instead, they are saved into Hugging Face&rsquo;s default cache directory.</p> <h2 id="default-cache-locations">Default cache locations </h2><p><code>huggingface-cli download</code> follows the Hugging Face Hub cache mechanism. The default paths are:</p> <table> <thead> <tr> <th>System</th> <th>Default cache directory</th> </tr> </thead> <tbody> <tr> <td>Linux / macOS</td> <td><code>~/.cache/huggingface/hub</code></td> </tr> <tr> <td>Windows</td> <td><code>C:\Users\用户名\.cache\huggingface\hub</code></td> </tr> </tbody> </table> <p>On Windows, if your username is <code>knightli</code>, the default path is roughly:</p> <div class="highlight"><div class="chroma"> <table class="lntable"><tr><td class="lntd"> <pre tabindex="0" class="chroma"><code><span class="lnt">1 </span></code></pre></td> <td class="lntd"> <pre tabindex="0" class="chroma"><code class="language-text" data-lang="text"><span class="line"><span class="cl">C:\Users\knightli\.cache\huggingface\hub </span></span></code></pre></td></tr></table> </div> </div><h2 id="why-the-model-is-not-in-the-current-directory">Why the model is not in the current directory </h2><p>This is expected behavior. By default, <code>huggingface-cli download</code> stores repository files in the cache directory so they can be reused later without downloading them again.</p> <p>Inside the cache directory, repositories are usually stored in a structure like this:</p> <div class="highlight"><div class="chroma"> <table class="lntable"><tr><td class="lntd"> <pre tabindex="0" class="chroma"><code><span class="lnt">1 </span></code></pre></td> <td class="lntd"> <pre tabindex="0" class="chroma"><code class="language-text" data-lang="text"><span class="line"><span class="cl">models--组织名--模型名 </span></span></code></pre></td></tr></table> </div> </div><p>For example, a model repository may be cached as:</p> <div class="highlight"><div class="chroma"> <table class="lntable"><tr><td class="lntd"> <pre tabindex="0" class="chroma"><code><span class="lnt">1 </span></code></pre></td> <td class="lntd"> <pre tabindex="0" class="chroma"><code class="language-text" data-lang="text"><span class="line"><span class="cl">models--unsloth--gemma-4-E4B-it-GGUF </span></span></code></pre></td></tr></table> </div> </div><p>The actual model files are usually under the <code>snapshots</code> subdirectory inside that cache folder.</p> <h2 id="how-to-download-to-a-specific-directory">How to download to a specific directory </h2><p>If you want the model files to appear in a directory you choose, add <code>--local-dir</code> to <code>huggingface-cli download</code>:</p> <div class="highlight"><div class="chroma"> <table class="lntable"><tr><td class="lntd"> <pre tabindex="0" class="chroma"><code><span class="lnt">1 </span></code></pre></td> <td class="lntd"> <pre tabindex="0" class="chroma"><code class="language-bash" data-lang="bash"><span class="line"><span class="cl">huggingface-cli download &lt;repo-id&gt; --local-dir D:<span class="se">\m</span>odels<span class="se">\&lt;</span>model-name&gt; </span></span></code></pre></td></tr></table> </div> </div><p>This is more convenient when working with local inference tools such as <code>llama.cpp</code>, <code>Ollama</code>, or <code>LM Studio</code>, because the model files are easier to find later.</p> <h2 id="summary">Summary </h2><ul> <li>Default cache directory: <code>~/.cache/huggingface/hub</code></li> <li>Default Windows directory: <code>C:\Users\用户名\.cache\huggingface\hub</code></li> <li>Use <code>--local-dir</code> if you want to save files to a specific location</li> <li>If you are only looking for a <code>.gguf</code> file, first check the <code>snapshots</code> subdirectory under the corresponding cached model repository</li> </ul>