<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness]]></title><description><![CDATA[<p dir="auto">内存的价格看来是降不了啦！需求越来越大。此配置需要64G内存。12G以上显存。推荐4070 Ti super 12G 以上。</p>
<h1>Strata 项目部署与使用指南</h1>
<h2>一、下载 Strata 项目</h2>
<ul>
<li>夸克网盘：<a href="https://pan.quark.cn/s/80dee6a5b8f3?pwd=xW4g" rel="nofollow ugc">下载地址</a></li>
<li>GitHub：<a href="https://github.com/Niko1221/Strata" rel="nofollow ugc">Niko1221/Strata</a></li>
</ul>
<h2>二、下载模型文件</h2>
<blockquote>
<p dir="auto">非手动下载用户可跳过此步骤。</p>
</blockquote>
<p dir="auto">可通过以下任一渠道下载模型文件：</p>
<ul>
<li>夸克网盘：<a href="https://pan.quark.cn/s/574269f404cd?pwd=kgVy" rel="nofollow ugc">下载地址</a></li>
<li>Hugging Face：<a href="https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/tree/main" rel="nofollow ugc">ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF</a></li>
<li>Hugging Face：<a href="https://huggingface.co/unsloth/DeepSeek-V4.1-Flash-1017-GGUF/tree/main/UD-Q8_K_XL" rel="nofollow ugc">unsloth/DeepSeek-V4.1-Flash-1017-GGUF（UD-Q8_K_XL）</a></li>
</ul>
<h2>三、下载推理引擎包</h2>
<h3>下载地址</h3>
<ul>
<li>夸克网盘：<a href="https://pan.quark.cn/s/03d643b43578?pwd=H8fc" rel="nofollow ugc">下载地址</a></li>
<li>GitHub：<a href="https://github.com/Niko1221/Strata/releases/tag/v0.2.5" rel="nofollow ugc">Strata v0.2.5 Releases</a></li>
</ul>
<h3>引擎包名称</h3>
<p dir="auto">英伟达显卡对应的引擎包名称为：</p>
<pre><code class="language-text">strata-dsv41-0.1.50-windows-x86_64-cuda-cuda13.tar.gz
</code></pre>
<h3>存放位置</h3>
<p dir="auto">将下载的压缩包解压到以下目录：</p>
<pre><code class="language-text">Strata-main\engine-dsv41
</code></pre>
<p dir="auto">其中，<code>Strata-main</code> 是 Strata 项目的根目录。如果该目录下没有 <code>engine-dsv41</code> 文件夹，请手动创建。</p>
<h2>四、下载启动文件</h2>
<blockquote>
<p dir="auto">非手动下载用户可跳过此步骤。</p>
</blockquote>
<h3>下载地址</h3>
<ul>
<li>夸克网盘：<a href="https://pan.quark.cn/s/1ed0fb1f6c62?pwd=bRcX" rel="nofollow ugc">下载地址</a></li>
<li>GitHub：<a href="https://github.com/xchaoge/0108-DeepSeek-V4.1-Flash-1017-Strata-" rel="nofollow ugc">启动文件下载地址</a></li>
</ul>
<h3>存放位置</h3>
<p dir="auto">将下载得到的两个启动文件放入 Strata 项目根目录：</p>
<pre><code class="language-text">Strata-main\
</code></pre>
<h2>五、运行 Strata 项目</h2>
<h3>1. 启动项目</h3>
<p dir="auto">根据下载方式选择对应的启动文件：</p>
<ul>
<li><strong>手动下载用户</strong>：双击项目根目录中的 <code>START-HERE-have-model.bat</code>。</li>
<li><strong>非手动下载用户</strong>：双击项目根目录中的 <code>START-HERE.bat</code>。</li>
</ul>
<h3>2. 访问控制面板</h3>
<p dir="auto">项目启动后，在浏览器中访问：</p>
<pre><code class="language-text">http://127.0.0.1:8080/
</code></pre>
<h3>3. 修改上下文长度</h3>
<p dir="auto">打开 Strata 项目根目录中符合以下命名规则的 JSON 文件：</p>
<pre><code class="language-text">strata-deepseek-v4.1-flash-ist-*.json
</code></pre>
<p dir="auto">找到并修改其中的 <code>--max-context</code> 参数，即可调整上下文长度。</p>
<h2>六、接入 DeepSeek Harness</h2>
<p dir="auto">DeepSeek 官网：<a href="https://www.deepseek.com/" rel="nofollow ugc">https://www.deepseek.com/</a></p>
<p dir="auto">按照下表填写接入参数：</p>
<table class="table table-bordered table-striped">
<thead>
<tr>
<th>参数</th>
<th>填写内容</th>
</tr>
</thead>
<tbody>
<tr>
<td>Provider ID</td>
<td><code>deepseek-strata</code></td>
</tr>
<tr>
<td>显示名称</td>
<td><code>deepseek-strata</code></td>
</tr>
<tr>
<td>API 地址</td>
<td><code>http://127.0.0.1:8080/v1</code></td>
</tr>
<tr>
<td>API 协议</td>
<td><code>OpenAI Chat Completions</code></td>
</tr>
<tr>
<td>API Key</td>
<td>任意填写</td>
</tr>
</tbody>
</table>
<p dir="auto">配置完成后，点击 <strong>“获取可用模型”</strong>，添加并选择需要使用的模型。</p>
]]></description><link>https://lcz.me/topic/2596</link><generator>RSS for Node</generator><lastBuildDate>Mon, 26 Oct 2026 21:42:28 GMT</lastBuildDate><atom:link href="https://lcz.me/topic/2596.rss" rel="self" type="application/rss+xml"/><pubDate>Mon, 26 Oct 2026 08:33:46 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 13:24:47 GMT]]></title><description><![CDATA[<p dir="auto">Strata 0.2.6 提速了<br />
<img src="https://upload.lcz.me/uploads/c72c9896-d8ae-47f5-8573-f1718735753a.jpeg" alt="image.jpeg" class="img-markdown  img-fluid" /></p>
]]></description><link>https://lcz.me/post/28012</link><guid isPermaLink="true">https://lcz.me/post/28012</guid><dc:creator><![CDATA[perter]]></dc:creator><pubDate>Mon, 26 Oct 2026 13:24:47 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 13:14:52 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/j-z-0" aria-label="Profile: J-Z-0">@<bdi>J-Z-0</bdi></a> 對啊 我們就一直用Qwen，梁子不出甜品級模型我們就繼續用Qwen用到別家也投入這戰場</p>
]]></description><link>https://lcz.me/post/28007</link><guid isPermaLink="true">https://lcz.me/post/28007</guid><dc:creator><![CDATA[dardeaw feng]]></dc:creator><pubDate>Mon, 26 Oct 2026 13:14:52 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 13:13:24 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/dardeaw-feng" aria-label="Profile: dardeaw-feng">@<bdi>dardeaw-feng</bdi></a> 非常同意，我现在就在用unsloth的q4kxl，单卡60tps。就这么跑着，跑到qwen4 出来</p>
]]></description><link>https://lcz.me/post/28005</link><guid isPermaLink="true">https://lcz.me/post/28005</guid><dc:creator><![CDATA[J Z 0]]></dc:creator><pubDate>Mon, 26 Oct 2026 13:13:24 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 12:39:52 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/j-z-0" aria-label="Profile: J-Z-0">@<bdi>J-Z-0</bdi></a> 這種單併發的推論引擎 千萬別花大錢 說真的DS V4.1不是太貴的模型 為了這個再砸錢下去也頂多跑跑Q2<br />
我用QFN一個多月了，對這模型只要上到4bit其能力就非常夠用了，玩strata就是為了性價比，當你偏離了性價比一切都顯得為了花錢而花錢</p>
]]></description><link>https://lcz.me/post/27983</link><guid isPermaLink="true">https://lcz.me/post/27983</guid><dc:creator><![CDATA[dardeaw feng]]></dc:creator><pubDate>Mon, 26 Oct 2026 12:39:52 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 12:19:20 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/williamlouis" aria-label="Profile: williamlouis">@<bdi>williamlouis</bdi></a> 研究半天，预估7900xtx单卡跑最小的iq2，大概10tps， 再等等吧，指不定什么时候又有黑科技出来呢。买现在这套就没想到会有strata，以为27b就是本地小破机器的极限了，买了老旧的5900xt和ddr4。不然肯定上最新的ddr5 pcie5了，感觉以后IO速度是一切的瓶颈</p>
]]></description><link>https://lcz.me/post/27962</link><guid isPermaLink="true">https://lcz.me/post/27962</guid><dc:creator><![CDATA[J Z 0]]></dc:creator><pubDate>Mon, 26 Oct 2026 12:19:20 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 11:52:11 GMT]]></title><description><![CDATA[<p dir="auto">看起來是Strata 控制 DSV4.1 在專家模型那邊不是很適配, 等下一版再看看?</p>
]]></description><link>https://lcz.me/post/27956</link><guid isPermaLink="true">https://lcz.me/post/27956</guid><dc:creator><![CDATA[Peace Love]]></dc:creator><pubDate>Mon, 26 Oct 2026 11:52:11 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 11:32:37 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/kop-wang" aria-label="Profile: kop-wang">@<bdi>kop-wang</bdi></a> 不知道PREFILL 4卡有多少 ？</p>
]]></description><link>https://lcz.me/post/27949</link><guid isPermaLink="true">https://lcz.me/post/27949</guid><dc:creator><![CDATA[Joseph Chen]]></dc:creator><pubDate>Mon, 26 Oct 2026 11:32:37 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 11:20:26 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/kop-wang" aria-label="Profile: kop-wang">@<bdi>kop-wang</bdi></a>  3090 大概 20t/s</p>
]]></description><link>https://lcz.me/post/27945</link><guid isPermaLink="true">https://lcz.me/post/27945</guid><dc:creator><![CDATA[johnnybegood]]></dc:creator><pubDate>Mon, 26 Oct 2026 11:20:26 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 10:00:56 GMT]]></title><description><![CDATA[<p dir="auto">256有了 4卡臣妾有点做不到啊</p>
]]></description><link>https://lcz.me/post/27938</link><guid isPermaLink="true">https://lcz.me/post/27938</guid><dc:creator><![CDATA[ka ka]]></dc:creator><pubDate>Mon, 26 Oct 2026 10:00:56 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 09:26:29 GMT]]></title><description><![CDATA[<p dir="auto">128G内存玩玩QFN算了，要不了那么大的本地模型。</p>
]]></description><link>https://lcz.me/post/27927</link><guid isPermaLink="true">https://lcz.me/post/27927</guid><dc:creator><![CDATA[terry]]></dc:creator><pubDate>Mon, 26 Oct 2026 09:26:29 GMT</pubDate></item><item><title><![CDATA[Reply to DeepSeek-V4.1-Flash-1017 Strata版 AI 大模型能本地部署｜接入 Harness on Mon, 26 Oct 2026 09:23:33 GMT]]></title><description><![CDATA[<p dir="auto">我看strata官方测试，4 x 9700 + 256RAM，才能做到decode 80t/s。</p>
<p dir="auto">如果真是最低配置，是不是decode就已经没法看了。配上deepseek喜欢雷霆思考……</p>
]]></description><link>https://lcz.me/post/27926</link><guid isPermaLink="true">https://lcz.me/post/27926</guid><dc:creator><![CDATA[kop wang]]></dc:creator><pubDate>Mon, 26 Oct 2026 09:23:33 GMT</pubDate></item></channel></rss>