{"id":118588,"date":"2026-10-27T18:09:17","date_gmt":"2026-10-27T10:09:17","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/118588.html"},"modified":"2026-10-27T18:09:17","modified_gmt":"2026-10-27T10:09:17","slug":"8x910b4-32g-npu%e6%9c%8d%e5%8a%a1%e5%99%a8-vllm-ascend%e9%83%a8%e7%bd%b2docker%e5%ae%89%e8%a3%85%e6%8a%a5%e5%91%8a","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/118588.html","title":{"rendered":"8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a"},"content":{"rendered":"<p>        \/* MD \/ \u5bcc\u6587\u672c\u4e2d\u7684 .toc\uff08\u542b\u535a\u5ba2\u56ed\u642c\u5bb6\u7b49\u5d4c\u5957\u7ed3\u6784\uff09\uff1b.toc-box \u5728\u4fa7\u680f\uff0c\u4e0d\u53d7\u5f71\u54cd *\/<br \/>\n        #content_views .toc,<br \/>\n        \/* \u7f16\u8f91\u5668\u5e38\u5728\u76ee\u5f55\u524d\u540e\u63d2\u5165\u7a7a p\uff08:empty \u4ecd\u5360 20px\uff09\uff0c\u4e00\u5e76\u53bb\u6389\u907f\u514d\u9876\u7a7a\u9699 *\/<br \/>\n        #content_views.markdown_views &gt; p:empty:has(+ .toc),<br \/>\n        #content_views.markdown_views &gt; .toc + p:empty,<br \/>\n        \/* \u5bcc\u6587\u672c\u65e7\u7248\u76ee\u5f55\u6807\u8bb0 *\/<br \/>\n        #content_views.htmledit_views #main-toc,<br \/>\n        #content_views.htmledit_views #hr-toc,<br \/>\n        #content_views.htmledit_views p[id*=\\&#8221;-toc\\&#8221;] {<br \/>\n          display: none !important;<br \/>\n        }<br \/>\n        \/* \u76ee\u5f55\u53bb\u6389\u540e\uff0c\u7d27\u8ddf\u7684\u9996\u4e2a\u6807\u9898\u4e0d\u518d\u591a\u51fa\u4e00\u5757\u4e0a\u8fb9\u8ddd *\/<br \/>\n        #content_views.markdown_views &gt; .toc + h1,<br \/>\n        #content_views.markdown_views &gt; .toc + h2,<br \/>\n        #content_views.markdown_views &gt; .toc + h3,<br \/>\n        #content_views.markdown_views &gt; .toc + h4,<br \/>\n        #content_views.markdown_views &gt; .toc + p:empty + h1,<br \/>\n        #content_views.markdown_views &gt; .toc + p:empty + h2,<br \/>\n        #content_views.markdown_views &gt; .toc + p:empty + h3,<br \/>\n        #content_views.markdown_views &gt; .toc + p:empty + h4 {<br \/>\n          margin-top: 0 !important;<br \/>\n        }<\/p>\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u4e00\u3001\u9879\u76ee\u6982\u8ff0<\/span><\/h3>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:currentcolor;vertical-align:top\">\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u4e00\u3001\u9879\u76ee\u6982\u8ff0<\/span><\/h3>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u9879\u76ee<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u5185\u5bb9<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u670d\u52a1\u5668IP<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">10.255.254.65<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u786c\u4ef6\u914d\u7f6e<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">8\u00d7 Huawei Ascend 910B4-32GB<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">NUMA\u67b6\u6784<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">2 Nodes (Node 0: 4 NPU, Node 1: 4 NPU)<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u64cd\u4f5c\u7cfb\u7edf<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Ubuntu 22.04 LTS<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u90e8\u7f72\u6a21\u578b<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Qwen3.6-27B-W8A8 (Eco-Tech\u5b98\u65b9\u9a8c\u8bc1)<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u63a8\u7406\u6846\u67b6<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">vLLM-Ascend v0.18.0<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">Docker\u955c\u50cf<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">m.daocloud.io\/quay.io\/ascend\/vllm-ascend:v0.18.0<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u90e8\u7f72\u65e5\u671f<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">2026-06-12<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u90e8\u7f72\u4eba\u5458<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">hmtsgpu<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u4e8c\u3001\u786c\u4ef6\u73af\u5883\u786e\u8ba4<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">2.1 NPU\u4fe1\u606f<\/span><\/h4>\n<h4 style=\"text-align:center\"><img decoding=\"async\" alt=\"\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/10\/20261027100914-6ae0784a8bebe.png\" \/><\/h4>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">2.2 NUMA\u62d3\u6251<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">$\u00a0numactl\u00a0<span style=\"color:#7d9029\">&#8211;hardware<\/span><\/span><\/p>\n<p style=\"text-align:center\"><span style=\"color:#000000\"><span style=\"color:#7d9029\"><img decoding=\"async\" alt=\"\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/10\/20261027100915-6ae0784ba48d7.png\" \/><\/span><\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">2.3 NPU\u4e0eNUMA\u5bf9\u5e94\u5173\u7cfb<\/span><\/h4>\n<table cellspacing=\"0\" style=\"width:784px\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top;width:73.5pt\">\n<p style=\"text-align:left\">NUMA Node<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:80.75pt\">\n<p style=\"text-align:left\">CPU\u8303\u56f4<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:53pt\">\n<p style=\"text-align:left\">\u5185\u5b58<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:62.05pt\">\n<p style=\"text-align:left\">\u5206\u914dNPU<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:173.5pt\">\n<p style=\"text-align:left\">\u670d\u52a1<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top;width:73.5pt\">\n<p style=\"text-align:left\">0<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:80.75pt\">\n<p style=\"text-align:left\">0-31, 64-95<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:53pt\">\n<p style=\"text-align:left\">193GB<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62.05pt\">\n<p style=\"text-align:left\">NPU 0-3<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:173.5pt\">\n<p style=\"text-align:left\">Service 1 (NPU 0-1), Service 2 (NPU 2-3)<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top;width:73.5pt\">\n<p style=\"text-align:left\">1<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:80.75pt\">\n<p style=\"text-align:left\">32-63, 96-127<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:53pt\">\n<p style=\"text-align:left\">258GB<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62.05pt\">\n<p style=\"text-align:left\">NPU 4-7<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:173.5pt\">\n<p style=\"text-align:left\">Service 3 (NPU 4-5), Service 4 (NPU 6-7)<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u4e09\u3001Docker\u5b89\u88c5\u8fc7\u7a0b<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">3.1\u00a0\u5b89\u88c5Docker CE<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#06287e\">sudo<\/span>\u00a0apt update<br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0apt install\u00a0<span style=\"color:#7d9029\">-y<\/span>\u00a0apt-transport-https ca-certificates curl gnupg lsb-release<\/p>\n<p>curl\u00a0<span style=\"color:#7d9029\">-fsSL<\/span>\u00a0https:\/\/download.docker.com\/linux\/ubuntu\/gpg\u00a0<span style=\"color:#007020\">|<\/span>\u00a0<span style=\"color:#06287e\">sudo<\/span>\u00a0gpg\u00a0<span style=\"color:#7d9029\">&#8211;dearmor<\/span>\u00a0<span style=\"color:#7d9029\">-o<\/span>\u00a0\/usr\/share\/keyrings\/docker-archive-keyring.gpg<\/p>\n<p><span style=\"color:#008000\">echo<\/span>\u00a0<span style=\"color:#4070a0\">&#034;deb [arch&#061;<\/span><span style=\"color:#19177c\">$(<\/span>dpkg\u00a0<span style=\"color:#7d9029\">&#8211;print-architecture<\/span><span style=\"color:#19177c\">)<\/span><span style=\"color:#4070a0\">\u00a0signed-by&#061;\/usr\/share\/keyrings\/docker-archive-keyring.gpg] https:\/\/download.docker.com\/linux\/ubuntu\u00a0<\/span><span style=\"color:#19177c\">$(<\/span>lsb_release\u00a0<span style=\"color:#7d9029\">-cs<\/span><span style=\"color:#19177c\">)<\/span><span style=\"color:#4070a0\">\u00a0stable&#034;<\/span>\u00a0<span style=\"color:#007020\">|<\/span>\u00a0<span style=\"color:#06287e\">sudo<\/span>\u00a0tee \/etc\/apt\/sources.list.d\/docker.list\u00a0<span style=\"color:#666666\">&gt;<\/span>\u00a0\/dev\/null<\/p>\n<p><span style=\"color:#06287e\">sudo<\/span>\u00a0apt update<br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0apt install\u00a0<span style=\"color:#7d9029\">-y<\/span>\u00a0docker-ce docker-ce-cli containerd.io docker-compose-plugin<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u5b89\u88c5\u7ed3\u679c&#xff1a;\u00a0Docker version 29.5.3, build d1c06ef<\/span><\/p>\n<p style=\"text-align:start\">\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">3.2\u00a0\u914d\u7f6e\u7528\u6237\u6743\u9650<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#06287e\">sudo<\/span>\u00a0usermod\u00a0<span style=\"color:#7d9029\">-aG<\/span>\u00a0docker\u00a0<span style=\"color:#19177c\">$USER<\/span><br \/>\nnewgrp\u00a0docker<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u9a8c\u8bc1&#xff1a;\u00a0docker run hello-world\u00a0\u6210\u529f\u8f93\u51fa<\/span><\/p>\n<p style=\"text-align:start\">\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">3.3\u00a0\u914d\u7f6e\u56fd\u5185\u955c\u50cf\u52a0\u901f<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#06287e\">sudo<\/span>\u00a0mkdir\u00a0<span style=\"color:#7d9029\">-p<\/span>\u00a0\/etc\/docker<br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0tee \/etc\/docker\/daemon.json\u00a0<span style=\"color:#666666\">&lt;&lt; &#039;EOF&#039;<\/span><br \/>\n<span style=\"color:#4070a0\">{<\/span><br \/>\n<span style=\"color:#4070a0\">\u00a0\u00a0\u00a0\u00a0&#034;registry-mirrors&#034;: [<\/span><br \/>\n<span style=\"color:#4070a0\">\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0&#034;https:\/\/docker.m.daocloud.io&#034;,<\/span><br \/>\n<span style=\"color:#4070a0\">\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0&#034;https:\/\/docker.1panel.live&#034;,<\/span><br \/>\n<span style=\"color:#4070a0\">\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0&#034;https:\/\/hub.rat.dev&#034;,<\/span><br \/>\n<span style=\"color:#4070a0\">\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0&#034;https:\/\/docker.mirrors.ustc.edu.cn&#034;,<\/span><br \/>\n<span style=\"color:#4070a0\">\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0&#034;https:\/\/dockerpull.org&#034;,<\/span><br \/>\n<span style=\"color:#4070a0\">\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0&#034;https:\/\/docker.1panel.dev&#034;<\/span><br \/>\n<span style=\"color:#4070a0\">\u00a0\u00a0\u00a0\u00a0]<\/span><br \/>\n<span style=\"color:#4070a0\">}<\/span><br \/>\n<span style=\"color:#666666\">EOF<\/span><\/p>\n<p><span style=\"color:#06287e\">sudo<\/span>\u00a0systemctl daemon-reload<br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0systemctl restart docker<\/span><\/p>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u56db\u3001vLLM-Ascend Docker\u955c\u50cf<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">4.1\u00a0\u955c\u50cf\u4fe1\u606f<\/span><\/h4>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u5c5e\u6027<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u503c<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u955c\u50cf\u6e90<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">quay.io\/ascend\/vllm-ascend:v0.18.0<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u56fd\u5185\u52a0\u901f<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">m.daocloud.io\/quay.io\/ascend\/vllm-ascend:v0.18.0<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u955c\u50cf\u5927\u5c0f<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">~15GB<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u5185\u7f6ePython<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">3.11.14<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">vLLM\u7248\u672c<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">0.18.0&#043;empty<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u5305\u542b\u7ec4\u4ef6<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">vllm-ascend plugin, CANN\u9a71\u52a8, PyTorch-NPU<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4 style=\"text-align:start\"><\/h4>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">4.2\u00a0\u62c9\u53d6\u547d\u4ee4<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">docker\u00a0pull m.daocloud.io\/quay.io\/ascend\/vllm-ascend:v0.18.0<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u7ed3\u679c&#xff1a;\u00a0Status: Image is up to date<\/span><\/p>\n<h4 style=\"text-align:start\"><\/h4>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">4.3\u00a0\u955c\u50cf\u9a8c\u8bc1<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">docker\u00a0images\u00a0<span style=\"color:#007020\">|<\/span>\u00a0<span style=\"color:#06287e\">grep<\/span>\u00a0vllm-ascend<br \/>\n<span style=\"color:#60a0b0\"># m.daocloud.io\/quay.io\/ascend\/vllm-ascend \u00a0\u00a0v0.18.0 \u00a0\u00a0&#8230;<\/span><\/span><\/p>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u4e94\u3001\u5bb9\u5668\u90e8\u7f72\u65b9\u6848<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">5.1\u00a0\u90e8\u7f72\u67b6\u6784<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u91c7\u75284\u5bb9\u5668\u00d72NPU\u67b6\u6784&#xff0c;\u4e25\u683c\u6309NUMA\u62d3\u6251\u9694\u79bb&#xff1a;<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510<br \/>\n\u2502 \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a08\u00d7 910B4 NPU Server \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u2502<br \/>\n\u251c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2524<br \/>\n\u2502 \u00a0NUMA 0 (CPU: 0-31,64-95 | Mem: 193GB) \u00a0\u00a0\u00a0\u2502<br \/>\n\u2502 \u00a0\u251c\u2500 Container 1: NPU 0,1 \u00a0\u2192 Port 8000 \u00a0\u00a0\u00a0\u00a0\u2502<br \/>\n\u2502 \u00a0\u2514\u2500 Container 2: NPU 2,3 \u00a0\u2192 Port 8001 \u00a0\u00a0\u00a0\u00a0\u2502<br \/>\n\u251c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2524<br \/>\n\u2502 \u00a0NUMA 1 (CPU: 32-63,96-127 | Mem: 258GB) \u00a0\u2502<br \/>\n\u2502 \u00a0\u251c\u2500 Container 3: NPU 4,5 \u00a0\u2192 Port 8002 \u00a0\u00a0\u00a0\u00a0\u2502<br \/>\n\u2502 \u00a0\u2514\u2500 Container 4: NPU 6,7 \u00a0\u2192 Port 8003 \u00a0\u00a0\u00a0\u00a0\u2502<br \/>\n\u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">5.2\u00a0\u5bb9\u5668\u521b\u5efa\u53c2\u6570<\/span><\/h4>\n<p style=\"text-align:start\">\n<p style=\"text-align:start\">\n<p style=\"text-align:start\">\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u6240\u6709\u5bb9\u5668\u5171\u4eab\u4ee5\u4e0b\u516c\u5171\u914d\u7f6e&#xff1a;<\/span><\/p>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u53c2\u6570<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u503c<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8bf4\u660e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;net&#061;host<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u7f51\u7edc\u6a21\u5f0f<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u4f7f\u7528\u5bbf\u4e3b\u673a\u7f51\u7edc\u6808<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;shm-size&#061;25g<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u5171\u4eab\u5185\u5b58<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">4\u5bb9\u5668\u5171\u4eab100GB\u7269\u7406\u5185\u5b58<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;cap-add SYS_NICE<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u7279\u6743<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">numactl\u7ed1\u5b9aCPU\/\u5185\u5b58\u9700\u8981<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;device \/dev\/davinciX<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">NPU\u8bbe\u5907<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u6bcf\u5bb9\u5668\u5206\u914d2\u5f20NPU<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;device \/dev\/davinci_manager<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u7ba1\u7406\u8bbe\u5907<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">NPU\u7ba1\u7406\u63a5\u53e3<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;device \/dev\/devmm_svm<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u5185\u5b58\u8bbe\u5907<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8bbe\u5907\u5185\u5b58\u7ba1\u7406<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;device \/dev\/hisi_hdc<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u4e3b\u673a\u8bbe\u5907<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u4e3b\u673a\u8bbe\u5907\u63a7\u5236\u5668<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">-v \/usr\/local\/dcmi<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u9a71\u52a8\u6302\u8f7d<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">DCMI\u63a5\u53e3<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">-v \/usr\/local\/Ascend\/driver<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u9a71\u52a8\u6302\u8f7d<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">CANN\u9a71\u52a8\u5e93<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">-v \/data\/models<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u6a21\u578b\u76ee\u5f55<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u5bbf\u4e3b\u673a\u6a21\u578b\u6302\u8f7d<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">5.3 NUMA\u9694\u79bb\u914d\u7f6e<\/span><\/h4>\n<table cellspacing=\"0\" style=\"width:783.828125px\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top;width:37.95pt\">\n<p style=\"text-align:center\">\u670d\u52a1<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:62pt\">\n<p style=\"text-align:center\">\u5bb9\u5668\u540d<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:71.15pt\">\n<p style=\"text-align:center\">cpuset-cpus<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:77.1pt\">\n<p style=\"text-align:center\">cpuset-mems<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:67.45pt\">\n<p style=\"text-align:center\">NPU\u8bbe\u5907<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:127.1pt\">\n<p style=\"text-align:center\">numactl\u7ed1\u5b9a<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top;width:37.95pt\">\n<p style=\"text-align:center\">1<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62pt\">\n<p style=\"text-align:center\">s1-npu01<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:71.15pt\">\n<p style=\"text-align:center\">0-31,64-95<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:77.1pt\">\n<p style=\"text-align:center\">0<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:67.45pt\">\n<p style=\"text-align:center\">davinci0,1<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:127.1pt\">\n<p style=\"text-align:center\">\u2013cpunodebind&#061;0 \u2013membind&#061;0<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top;width:37.95pt\">\n<p style=\"text-align:center\">2<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62pt\">\n<p style=\"text-align:center\">s2-npu23<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:71.15pt\">\n<p style=\"text-align:center\">0-31,64-95<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:77.1pt\">\n<p style=\"text-align:center\">0<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:67.45pt\">\n<p style=\"text-align:center\">davinci2,3<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:127.1pt\">\n<p style=\"text-align:center\">\u2013cpunodebind&#061;0 \u2013membind&#061;0<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top;width:37.95pt\">\n<p style=\"text-align:center\">3<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62pt\">\n<p style=\"text-align:center\">s3-npu45<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:71.15pt\">\n<p style=\"text-align:center\">32-63,96-127<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:77.1pt\">\n<p style=\"text-align:center\">1<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:67.45pt\">\n<p style=\"text-align:center\">davinci4,5<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:127.1pt\">\n<p style=\"text-align:center\">\u2013cpunodebind&#061;1 \u2013membind&#061;1<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top;width:37.95pt\">\n<p style=\"text-align:center\">4<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62pt\">\n<p style=\"text-align:center\">s4-npu67<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:71.15pt\">\n<p style=\"text-align:center\">32-63,96-127<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:77.1pt\">\n<p style=\"text-align:center\">1<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:67.45pt\">\n<p style=\"text-align:center\">davinci6,7<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:127.1pt\">\n<p style=\"text-align:center\">\u2013cpunodebind&#061;1 \u2013membind&#061;1<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">5.4\u00a0\u5bb9\u5668\u521b\u5efa\u547d\u4ee4\u793a\u4f8b&#xff08;Service 1&#xff09;<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">docker\u00a0run\u00a0<span style=\"color:#7d9029\">-itd<\/span>\u00a0\u00a0\u00a0<span style=\"color:#7d9029\">&#8211;name<\/span>\u00a0vllm-ascend-s1-npu01 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;net<\/span><span style=\"color:#666666\">&#061;<\/span>host \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;shm-size<\/span><span style=\"color:#666666\">&#061;<\/span>25g \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;cap-add<\/span>\u00a0SYS_NICE \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;cpuset-cpus<\/span><span style=\"color:#666666\">&#061;<\/span><span style=\"color:#4070a0\">&#034;0-31,64-95&#034;<\/span>\u00a0\u00a0\u00a0<span style=\"color:#7d9029\">&#8211;cpuset-mems<\/span><span style=\"color:#666666\">&#061;<\/span><span style=\"color:#4070a0\">&#034;0&#034;<\/span>\u00a0\u00a0\u00a0<span style=\"color:#7d9029\">&#8211;device<\/span>\u00a0\/dev\/davinci0 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;device<\/span>\u00a0\/dev\/davinci1 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;device<\/span>\u00a0\/dev\/davinci_manager \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;device<\/span>\u00a0\/dev\/devmm_svm \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;device<\/span>\u00a0\/dev\/hisi_hdc \u00a0\u00a0<span style=\"color:#7d9029\">-v<\/span>\u00a0\/usr\/local\/dcmi:\/usr\/local\/dcmi \u00a0\u00a0<span style=\"color:#7d9029\">-v<\/span>\u00a0\/usr\/local\/bin\/npu-smi:\/usr\/local\/bin\/npu-smi \u00a0\u00a0<span style=\"color:#7d9029\">-v<\/span>\u00a0\/usr\/local\/Ascend\/driver\/lib64\/:\/usr\/local\/Ascend\/driver\/lib64\/ \u00a0\u00a0<span style=\"color:#7d9029\">-v<\/span>\u00a0\/usr\/local\/Ascend\/driver\/version.info:\/usr\/local\/Ascend\/driver\/version.info \u00a0\u00a0<span style=\"color:#7d9029\">-v<\/span>\u00a0\/etc\/ascend_install.info:\/etc\/ascend_install.info \u00a0\u00a0<span style=\"color:#7d9029\">-v<\/span>\u00a0\/data\/models:\/data\/models \u00a0\u00a0m.daocloud.io\/quay.io\/ascend\/vllm-ascend:v0.18.0 \u00a0\u00a0bash<\/span><\/p>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u516d\u3001\u6a21\u578b\u90e8\u7f72<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">6.1\u00a0\u6a21\u578b\u4fe1\u606f<\/span><\/h4>\n<table cellspacing=\"0\" style=\"width:784px\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">\u6a21\u578b<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">\u6765\u6e90<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">\u5927\u5c0f<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:104.2pt\">\n<p style=\"text-align:left\">\u91cf\u5316\u65b9\u5f0f<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top;width:104.2pt\">\n<p style=\"text-align:left\">\u9a8c\u8bc1\u72b6\u6001<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">Qwen3.5-27B-w8a8-mtp<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">Eco-Tech (ModelScope)<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">~35GB<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:104.2pt\">\n<p style=\"text-align:left\">W8A8 (ModelSlim)<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:104.2pt\">\n<p style=\"text-align:left\">vllm-ascend v0.18.0rc1\u9a8c\u8bc1<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">Qwen3.6-27B-w8a8<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">Eco-Tech (ModelScope)<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:62.5pt\">\n<p style=\"text-align:left\">~35GB<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:104.2pt\">\n<p style=\"text-align:left\">W8A8 (ModelSlim)<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top;width:104.2pt\">\n<p style=\"text-align:left\">vllm-ascend v0.18.0rc1\u9a8c\u8bc1<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4 style=\"text-align:start\"><\/h4>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">6.2\u00a0\u4e0b\u8f7d\u547d\u4ee4<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#60a0b0\">#\u00a0\u5728\u5bb9\u5668\u5185\u6267\u884c<\/span><br \/>\n<span style=\"color:#008000\">export<\/span>\u00a0<span style=\"color:#19177c\">VLLM_USE_MODELSCOPE<\/span><span style=\"color:#666666\">&#061;<\/span>true<br \/>\npip\u00a0install modelscope<\/p>\n<p>modelscope\u00a0download\u00a0<span style=\"color:#7d9029\">&#8211;model<\/span>\u00a0Eco-Tech\/Qwen3.5-27B-w8a8-mtp\u00a0<span style=\"color:#7d9029\">&#8211;local_dir<\/span>\u00a0\/data\/models\/Qwen3.5-27B-w8a8-mtp<br \/>\nmodelscope\u00a0download\u00a0<span style=\"color:#7d9029\">&#8211;model<\/span>\u00a0Eco-Tech\/Qwen3.6-27B-w8a8\u00a0<span style=\"color:#7d9029\">&#8211;local_dir<\/span>\u00a0\/data\/models\/Qwen3.6-27B-w8a8<\/span><\/p>\n<h4 style=\"text-align:start\"><\/h4>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">6.3\u00a0\u6a21\u578b\u6587\u4ef6\u7ed3\u6784<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\/data\/models\/Qwen3.6-27B-w8a8\/<br \/>\n\u251c\u2500\u2500 config.json \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0#\u00a0\u6a21\u578b\u914d\u7f6e\u00a0(4308 bytes)<br \/>\n\u251c\u2500\u2500 quant_model_description.json \u00a0\u00a0#\u00a0\u91cf\u5316\u63cf\u8ff0\u00a0(124KB)<br \/>\n\u251c\u2500\u2500 quant_model_weights-00001-of-00009.safetensors \u00a0#\u00a0\u6743\u91cd\u5206\u72471 (4.0GB)<br \/>\n\u251c\u2500\u2500 quant_model_weights-00002-of-00009.safetensors \u00a0#\u00a0\u6743\u91cd\u5206\u72472 (4.0GB)<br \/>\n\u251c\u2500\u2500 &#8230; (\u51719\u4e2a\u5206\u7247)<br \/>\n\u251c\u2500\u2500 quant_model_weights.safetensors.index.json \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0#\u00a0\u7d22\u5f15\u6587\u4ef6<br \/>\n\u251c\u2500\u2500 tokenizer.json \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0#\u00a0\u5206\u8bcd\u5668<br \/>\n\u251c\u2500\u2500 tokenizer_config.json \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0#\u00a0\u5206\u8bcd\u5668\u914d\u7f6e<br \/>\n\u2514\u2500\u2500 chat_template.jinja \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0#\u00a0\u5bf9\u8bdd\u6a21\u677f<\/span><\/p>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h4 style=\"text-align:start\"><\/h4>\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u4e03\u3001vLLM\u670d\u52a1\u542f\u52a8\u914d\u7f6e<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">7.1\u00a0\u542f\u52a8\u547d\u4ee4<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\/usr\/bin\/numactl\u00a0<span style=\"color:#7d9029\">&#8211;cpunodebind<\/span><span style=\"color:#666666\">&#061;<\/span>0\u00a0<span style=\"color:#7d9029\">&#8211;membind<\/span><span style=\"color:#666666\">&#061;<\/span>0 \u00a0\u00a0\/usr\/local\/python3.11.14\/bin\/vllm serve \/data\/models\/Qwen3.6-27B-w8a8 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;tensor-parallel-size<\/span>\u00a02 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;port<\/span>\u00a08000 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;host<\/span>\u00a00.0.0.0 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;dtype<\/span>\u00a0bfloat16 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;max-model-len<\/span>\u00a0262144 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;quantization<\/span>\u00a0ascend \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;enforce-eager<\/span>\u00a0\u00a0\u00a0<span style=\"color:#7d9029\">&#8211;gpu-memory-utilization<\/span>\u00a00.95 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;disable-custom-all-reduce<\/span>\u00a0\u00a0\u00a0<span style=\"color:#7d9029\">&#8211;enable-prefix-caching<\/span>\u00a0\u00a0\u00a0<span style=\"color:#7d9029\">&#8211;served-model-name<\/span>\u00a0qwen3.6-27b-w8a8 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;enable-auto-tool-choice<\/span>\u00a0\u00a0\u00a0<span style=\"color:#7d9029\">&#8211;tool-call-parser<\/span>\u00a0qwen3_xml \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;reasoning-parser<\/span>\u00a0qwen3 \u00a0\u00a0<span style=\"color:#7d9029\">&#8211;api-key<\/span>\u00a0sk-qwen-27b-w8a8-2026<\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">7.2\u00a0\u53c2\u6570\u8bf4\u660e<\/span><\/h4>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u53c2\u6570<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u503c<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8bf4\u660e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;tensor-parallel-size<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">2<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">2\u5f20NPU\u5f20\u91cf\u5e76\u884c<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;dtype<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">bfloat16<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8ba1\u7b97\u6570\u636e\u7c7b\u578b<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;max-model-len<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">262144<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u6700\u5927\u4e0a\u4e0b\u6587\u957f\u5ea6\u00a0(Qwen3.6)<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;quantization<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">ascend<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u6607\u817eModelSlim W8A8\u91cf\u5316<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;enforce-eager<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u7981\u7528graph\u7f16\u8bd1&#xff0c;\u786e\u4fdd\u7a33\u5b9a\u6027<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;gpu-memory-utilization<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">0.95<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">GPU\u5185\u5b58\u5229\u7528\u738795%<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;disable-custom-all-reduce<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u7981\u7528\u81ea\u5b9a\u4e49all-reduce<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;enable-prefix-caching<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u542f\u7528\u524d\u7f00\u7f13\u5b58\u52a0\u901f<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;enable-auto-tool-choice<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u81ea\u52a8\u5de5\u5177\u9009\u62e9<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;tool-call-parser<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">qwen3_xml<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Qwen3 XML\u5de5\u5177\u8c03\u7528\u683c\u5f0f<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;reasoning-parser<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">qwen3<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Qwen3\u63a8\u7406\u6a21\u5f0f\u89e3\u6790<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">&#8211;api-key<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">sk-qwen-27b-w8a8-2026<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">API\u8ba4\u8bc1\u5bc6\u94a5<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h4 style=\"text-align:start\"><\/h4>\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u516b\u3001Nginx\u8d1f\u8f7d\u5747\u8861\u914d\u7f6e<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">8.1\u00a0\u5b89\u88c5Nginx<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#06287e\">sudo<\/span>\u00a0apt update<br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0apt install\u00a0<span style=\"color:#7d9029\">-y<\/span>\u00a0nginx<\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">8.2\u00a0\u914d\u7f6e\u6587\u4ef6<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">upstream vllm_backend {<br \/>\n\u00a0\u00a0\u00a0\u00a0server 10.255.254.65:8000;<br \/>\n\u00a0\u00a0\u00a0\u00a0server 10.255.254.65:8001;<br \/>\n\u00a0\u00a0\u00a0\u00a0server 10.255.254.65:8002;<br \/>\n\u00a0\u00a0\u00a0\u00a0server 10.255.254.65:8003;<br \/>\n\u00a0\u00a0\u00a0\u00a0keepalive 64;<br \/>\n}<\/p>\n<p>server {<br \/>\n\u00a0\u00a0\u00a0\u00a0listen 2199;<br \/>\n\u00a0\u00a0\u00a0\u00a0client_max_body_size 0;<br \/>\n\u00a0\u00a0\u00a0\u00a0client_body_timeout 3600s;<br \/>\n\u00a0\u00a0\u00a0\u00a0client_header_timeout 3600s;<\/p>\n<p>\u00a0\u00a0\u00a0\u00a0location \/v1\/ {<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_pass http:\/\/vllm_backend;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_http_version 1.1;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_set_header Connection &#034;&#034;;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_connect_timeout 30s;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_send_timeout 3600s;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_read_timeout 3600s;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_buffering off;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0chunked_transfer_encoding on;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_set_header Host $host;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_set_header X-Real-IP $remote_addr;<br \/>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;<br \/>\n\u00a0\u00a0\u00a0\u00a0}<br \/>\n}<\/span><\/p>\n<p style=\"text-align:start\">\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">8.3\u00a0\u8d85\u65f6\u914d\u7f6e\u6c47\u603b<\/span><\/h4>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u8d85\u65f6\u7c7b\u578b<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u503c<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8bf4\u660e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">client_body_timeout<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">3600s<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8bfb\u53d6\u5ba2\u6237\u7aef\u8bf7\u6c42\u4f53<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">client_header_timeout<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">3600s<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8bfb\u53d6\u5ba2\u6237\u7aef\u8bf7\u6c42\u5934<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">proxy_connect_timeout<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">30s<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8fde\u63a5\u540e\u7aef\u8d85\u65f6<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">proxy_send_timeout<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">3600s<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u5411\u540e\u7aef\u53d1\u9001\u6570\u636e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">proxy_read_timeout<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">3600s<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u7b49\u5f85\u540e\u7aef\u54cd\u5e94<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h4 style=\"text-align:start\"><\/h4>\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u4e5d\u3001systemd\u670d\u52a1\u5316<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">9.1\u00a0\u670d\u52a1\u6587\u4ef6<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#007020\">[Unit]<\/span><br \/>\n<span style=\"color:#902000\">Description<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">vLLM Ascend Service 1 (NPU 0-1, NUMA 0, Port 8000)<\/span><br \/>\n<span style=\"color:#902000\">After<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">docker.service network.target<\/span><br \/>\n<span style=\"color:#902000\">Requires<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">docker.service<\/span><\/p>\n<p><span style=\"color:#007020\">[Service]<\/span><br \/>\n<span style=\"color:#902000\">Type<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">simple<\/span><br \/>\n<span style=\"color:#902000\">Restart<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">always<\/span><br \/>\n<span style=\"color:#902000\">RestartSec<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#40a070\">10<\/span><br \/>\n<span style=\"color:#902000\">User<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">root<\/span><br \/>\n<span style=\"color:#902000\">ExecStart<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">\/usr\/local\/bin\/vllm-ascend-s1-start.sh<\/span><br \/>\n<span style=\"color:#902000\">ExecStop<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">\/usr\/bin\/docker stop -t 30 vllm-ascend-s1-npu01<\/span><br \/>\n<span style=\"color:#902000\">ExecStopPost<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">\/usr\/bin\/docker rm -f vllm-ascend-s1-npu01<\/span><\/p>\n<p><span style=\"color:#007020\">[Install]<\/span><br \/>\n<span style=\"color:#902000\">WantedBy<\/span><span style=\"color:#007020\">&#061;<\/span><span style=\"color:#4070a0\">multi-user.target<\/span><\/span><\/p>\n<h4 style=\"text-align:start\"><\/h4>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">9.2\u00a0\u542f\u52a8\u811a\u672c\u529f\u80fd<\/span><\/h4>\n<li>\u5bb9\u5668\u521b\u5efa&#xff1a;\u5982\u5bb9\u5668\u4e0d\u5b58\u5728&#xff0c;\u81ea\u52a8\u521b\u5efa\u5e76\u914d\u7f6eNUMA\u9694\u79bb<\/li>\n<li>\u5bb9\u5668\u542f\u52a8&#xff1a;\u5982\u5bb9\u5668\u672a\u8fd0\u884c&#xff0c;\u81ea\u52a8\u542f\u52a8<\/li>\n<li>vLLM\u542f\u52a8&#xff1a;\u5982vllm serve\u672a\u8fd0\u884c&#xff0c;\u81ea\u52a8\u5728\u5bb9\u5668\u5185\u542f\u52a8<\/li>\n<li>\u8fdb\u7a0b\u76d1\u63a7&#xff1a;while true\u5faa\u73af&#xff0c;\u6bcf10\u79d2\u68c0\u6d4b&#xff0c;\u5d29\u6e83\u81ea\u52a8\u91cd\u542f<\/li>\n<li>\u6e05\u7406\u505c\u6b62&#xff1a;ExecStop\u65f6\u505c\u6b62\u5e76\u5220\u9664\u5bb9\u5668<\/li>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">9.3\u00a0\u670d\u52a1\u7ba1\u7406\u547d\u4ee4<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#60a0b0\">#\u00a0\u67e5\u770b\u72b6\u6001<\/span><br \/>\nsystemctl\u00a0status vllm-ascend-s1<\/p>\n<p><span style=\"color:#60a0b0\">#\u00a0\u67e5\u770b\u65e5\u5fd7<\/span><br \/>\njournalctl\u00a0<span style=\"color:#7d9029\">-u<\/span>\u00a0vllm-ascend-s1\u00a0<span style=\"color:#7d9029\">-f<\/span><\/p>\n<p><span style=\"color:#60a0b0\">#\u00a0\u91cd\u542f\u670d\u52a1<\/span><br \/>\nsystemctl\u00a0restart vllm-ascend-s1<\/p>\n<p><span style=\"color:#60a0b0\">#\u00a0\u5f00\u673a\u81ea\u542f<\/span><br \/>\nsystemctl\u00a0enable vllm-ascend-s1<\/span><\/p>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h4 style=\"text-align:start\"><\/h4>\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u5341\u3001\u95ee\u9898\u4e0e\u89e3\u51b3\u65b9\u6848<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">10.1\u00a0\u95ee\u98981&#xff1a;\u6a21\u578b\u8f93\u51fa\u5168\u611f\u53f9\u53f7<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u73b0\u8c61&#xff1a;\u00a0curl\u00a0\u8fd4\u56de\u00a0&#034;content&#034;:&#034;!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!&#034;<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u6839\u56e0&#xff1a;\u00a01.\u00a0config.json\u00a0\u88ab\u624b\u52a8\u7be1\u6539&#xff08;\u67b6\u6784\u4ece\u00a0Qwen3_5ForConditionalGeneration\u00a0\u6539\u4e3a\u00a0Qwen3ForCausalLM&#xff09;\u00a02.\u00a0safetensors\u00a0\u6743\u91cd\u6587\u4ef6\u7f3a\u5931\u00a0weight_scale\u00a0\u548c\u00a0weight_offset\u00a0\u91cf\u5316\u53c2\u6570<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u89e3\u51b3&#xff1a;\u00a0\u91cd\u65b0\u4e0b\u8f7d\u5b98\u65b9\u9a8c\u8bc1\u6a21\u578b\u00a0Eco-Tech\/Qwen3.6-27B-w8a8<\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">10.2\u00a0\u95ee\u98982&#xff1a;numactl\u6743\u9650\u4e0d\u8db3<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u73b0\u8c61&#xff1a;\u00a0set_mempolicy: Operation not permitted<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u89e3\u51b3&#xff1a;\u00a0\u521b\u5efa\u5bb9\u5668\u65f6\u6dfb\u52a0\u00a0&#8211;cap-add SYS_NICE<\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">10.3\u00a0\u95ee\u98983&#xff1a;Nginx\u91cd\u590dupstream<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u73b0\u8c61&#xff1a;\u00a0duplicate upstream &#034;vllm_backend&#034;<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u89e3\u51b3&#xff1a;\u00a0\u6e05\u7406\u00a0\/etc\/nginx\/conf.d\/\u00a0\u548c\u00a0\/etc\/nginx\/sites-enabled\/\u00a0\u4e2d\u91cd\u590d\u5b9a\u4e49&#xff0c;\u5408\u5e76\u4e3a\u5355\u4e00\u914d\u7f6e\u6587\u4ef6<\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">10.4\u00a0\u95ee\u98984&#xff1a;\u811a\u672c\u8bed\u6cd5\u9519\u8bef<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u73b0\u8c61&#xff1a;\u00a0&#8211;api-key: command not found<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u6839\u56e0&#xff1a;\u00a0sed\u00a0\u63d2\u5165\u00a0&#8211;api-key\u00a0\u65f6\u4f7f\u7528\u4e86\u00a0\\\\n\u00a0\u5b57\u9762\u91cf&#xff0c;\u5bfc\u81f4shell\u89e3\u6790\u9519\u8bef<\/span><\/p>\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u89e3\u51b3&#xff1a;\u00a0\u91cd\u65b0\u751f\u6210\u5b8c\u6574\u811a\u672c&#xff0c;\u786e\u4fdd\u00a0&#8211;api-key\u00a0\u5728\u540c\u4e00\u884c\u6216\u7528\u6b63\u786e\u7eed\u884c\u7b26<\/span><\/p>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u5341\u4e00\u3001\u9a8c\u8bc1\u7ed3\u679c<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">11.1\u00a0\u670d\u52a1\u72b6\u6001<\/span><\/h4>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u670d\u52a1<\/p>\n<\/td>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u5bb9\u5668<\/p>\n<\/td>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u72b6\u6001<\/p>\n<\/td>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">PID<\/p>\n<\/td>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u7aef\u53e3<\/p>\n<\/td>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">NPU<\/p>\n<\/td>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">NUMA<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">vllm-ascend-s1<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">s1-npu01<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">active<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">86<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">8000<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">0,1<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">0<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">vllm-ascend-s2<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">s2-npu23<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">active<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">86<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">8001<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">2,3<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">0<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">vllm-ascend-s3<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">s3-npu45<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">active<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">86<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">8002<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">4,5<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">1<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">vllm-ascend-s4<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">s4-npu67<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">active<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">86<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">8003<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">6,7<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">1<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">11.2 API\u6d4b\u8bd5<\/span><\/h4>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u6d4b\u8bd5\u9879<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8bf7\u6c42<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u9884\u671f<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u7ed3\u679c<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u65e0Key\u8bbf\u95ee<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u65e0Authorization<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">401<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u2705 401 Unauthorized<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u9519\u8befKey<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Bearer wrong-key<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">401<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u2705 401 Unauthorized<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u6b63\u786eKey<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Bearer sk-qwen-27b-w8a8-2026<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">200<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u2705 200 OK<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u6d41\u5f0f\u8f93\u51fa<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">stream: true<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">SSE<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u2705 data: {\u2026} &#043; [DONE]<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u6a21\u578b\u5217\u8868<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">GET \/v1\/models<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">qwen3.6-27b-w8a8<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u2705\u00a0\u8fd4\u56de\u6b63\u786e<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4 style=\"text-align:start\"><\/h4>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">11.3\u00a0\u63a8\u7406\u9a8c\u8bc1<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\">curl\u00a0http:\/\/10.255.254.65:2199\/v1\/chat\/completions \u00a0\u00a0<span style=\"color:#7d9029\">-H<\/span>\u00a0<span style=\"color:#4070a0\">&#034;Authorization: Bearer sk-qwen-27b-w8a8-2026&#034;<\/span>\u00a0\u00a0\u00a0<span style=\"color:#7d9029\">-d<\/span>\u00a0<span style=\"color:#4070a0\">&#039;{&#034;model&#034;:&#034;qwen3.6-27b-w8a8&#034;,&#034;messages&#034;:[{&#034;role&#034;:&#034;user&#034;,&#034;content&#034;:&#034;What is the capital of France?&#034;}],&#034;max_tokens&#034;:50}&#039;<\/span><\/p>\n<p><span style=\"color:#60a0b0\">#\u00a0\u8f93\u51fa: thinking process &#043; &#034;Expected answer: Paris&#034;<\/span><\/span><\/p>\n<p style=\"text-align:start\">\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u5341\u4e8c\u3001\u6587\u4ef6\u6e05\u5355<\/span><\/h3>\n<table cellspacing=\"0\">\n<tbody>\n<tr>\n<td style=\"border-color:windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\u6587\u4ef6\u8def\u5f84<\/p>\n<\/td>\n<td style=\"border-color:windowtext windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">\u8bf4\u660e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/usr\/local\/bin\/vllm-ascend-s1-start.sh<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Service 1\u542f\u52a8\u811a\u672c<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/usr\/local\/bin\/vllm-ascend-s2-start.sh<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Service 2\u542f\u52a8\u811a\u672c<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/usr\/local\/bin\/vllm-ascend-s3-start.sh<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Service 3\u542f\u52a8\u811a\u672c<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/usr\/local\/bin\/vllm-ascend-s4-start.sh<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Service 4\u542f\u52a8\u811a\u672c<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/etc\/systemd\/system\/vllm-ascend-s1.service<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Service 1 systemd\u914d\u7f6e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/etc\/systemd\/system\/vllm-ascend-s2.service<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Service 2 systemd\u914d\u7f6e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/etc\/systemd\/system\/vllm-ascend-s3.service<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Service 3 systemd\u914d\u7f6e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/etc\/systemd\/system\/vllm-ascend-s4.service<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Service 4 systemd\u914d\u7f6e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/etc\/nginx\/sites-available\/vllm<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Nginx\u8d1f\u8f7d\u5747\u8861\u914d\u7f6e<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/data\/models\/Qwen3.6-27B-w8a8\/<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Qwen3.6\u6a21\u578b\u6587\u4ef6<\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border-color:currentcolor windowtext windowtext;vertical-align:top\">\n<p style=\"text-align:left\">\/data\/models\/Qwen3.5-27B-w8a8-mtp\/<\/p>\n<\/td>\n<td style=\"border-color:currentcolor windowtext windowtext currentcolor;vertical-align:top\">\n<p style=\"text-align:left\">Qwen3.5\u6a21\u578b\u6587\u4ef6<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<h4 style=\"text-align:start\"><\/h4>\n<h3 style=\"text-align:start\"><span style=\"color:#000000\">\u5341\u4e09\u3001\u540e\u7eed\u7ef4\u62a4<\/span><\/h3>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">13.1\u00a0\u65e5\u5e38\u68c0\u67e5<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#60a0b0\"># NPU\u72b6\u6001<\/span><br \/>\nnpu-smi\u00a0info<\/p>\n<p><span style=\"color:#60a0b0\">#\u00a0\u5bb9\u5668\u72b6\u6001<\/span><br \/>\ndocker\u00a0ps\u00a0<span style=\"color:#007020\">|<\/span>\u00a0<span style=\"color:#06287e\">grep<\/span>\u00a0vllm-ascend<\/p>\n<p><span style=\"color:#60a0b0\">#\u00a0\u670d\u52a1\u72b6\u6001<\/span><br \/>\nsystemctl\u00a0status vllm-ascend-s1 s2 s3 s4<\/p>\n<p><span style=\"color:#60a0b0\"># Nginx\u72b6\u6001<\/span><br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0systemctl status nginx<\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">13.2\u00a0\u65e5\u5fd7\u67e5\u770b<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#60a0b0\"># vLLM\u65e5\u5fd7<\/span><br \/>\njournalctl\u00a0<span style=\"color:#7d9029\">-u<\/span>\u00a0vllm-ascend-s1\u00a0<span style=\"color:#7d9029\">-f<\/span><\/p>\n<p><span style=\"color:#60a0b0\"># Nginx\u8bbf\u95ee\u65e5\u5fd7<\/span><br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0tail\u00a0<span style=\"color:#7d9029\">-f<\/span>\u00a0\/var\/log\/nginx\/access.log<\/p>\n<p><span style=\"color:#60a0b0\"># Nginx\u9519\u8bef\u65e5\u5fd7<\/span><br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0tail\u00a0<span style=\"color:#7d9029\">-f<\/span>\u00a0\/var\/log\/nginx\/error.log<\/span><\/p>\n<h4 style=\"text-align:start\"><span style=\"color:#000000\">13.3\u00a0\u91cd\u542f\u64cd\u4f5c<\/span><\/h4>\n<p style=\"text-align:start\"><span style=\"color:#000000\"><span style=\"color:#60a0b0\">#\u00a0\u91cd\u542f\u5355\u4e2a\u670d\u52a1<\/span><br \/>\nsystemctl\u00a0restart vllm-ascend-s1<\/p>\n<p><span style=\"color:#60a0b0\">#\u00a0\u91cd\u542f\u6240\u6709\u670d\u52a1<\/span><br \/>\nsystemctl\u00a0restart vllm-ascend-s1 vllm-ascend-s2 vllm-ascend-s3 vllm-ascend-s4<\/p>\n<p><span style=\"color:#60a0b0\">#\u00a0\u91cd\u8f7dNginx<\/span><br \/>\n<span style=\"color:#06287e\">sudo<\/span>\u00a0nginx\u00a0<span style=\"color:#7d9029\">-s<\/span>\u00a0reload<\/span><\/p>\n<p style=\"text-align:start\">\n<hr \/>\n<p style=\"text-align:start\">\n<p style=\"text-align:start\"><span style=\"color:#000000\">\u62a5\u544a\u751f\u6210\u65f6\u95f4&#xff1a;\u00a02026-06-12\u00a0\u62a5\u544a\u7248\u672c&#xff1a;\u00a0v1.0\u00a0\u670d\u52a1\u5668&#xff1a;\u00a08\u00d7910B4-32G (10.255.254.65)<\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n","protected":false},"excerpt":{"rendered":"<p>\/* MD \/ \u5bcc\u6587\u672c\u4e2d\u7684 .toc\uff08\u542b\u535a\u5ba2\u56ed\u642c\u5bb6\u7b49\u5d4c\u5957\u7ed3\u6784\uff09\uff1b.toc-box \u5728\u4fa7\u680f\uff0c\u4e0d\u53d7\u5f71\u54cd *\/#content_views .toc,\/* \u7f16\u8f91\u5668\u5e38\u5728\u76ee\u5f55\u524d\u540e\u63d2\u5165\u7a7a p\uff08:empty \u4ecd\u5360 20px\uff09\uff0c\u4e00\u5e76\u53bb\u6389\u907f\u514d\u9876\u7a7a\u9699 *\/#content_views.markdown_views &gt; p:empty:has(+ .toc),#content_views.markdown_views &gt; .toc + p:empty,\/* \u5bcc\u6587\u672c\u65e7\u7248\u76ee\u5f55\u6807\u8bb0 *\/#content<\/p>\n","protected":false},"author":2,"featured_media":118586,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[27,2681,43],"topic":[],"class_list":["post-118588","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-docker","tag-vllm","tag-43"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v20.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/118588.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\/* MD \/ \u5bcc\u6587\u672c\u4e2d\u7684 .toc\uff08\u542b\u535a\u5ba2\u56ed\u642c\u5bb6\u7b49\u5d4c\u5957\u7ed3\u6784\uff09\uff1b.toc-box \u5728\u4fa7\u680f\uff0c\u4e0d\u53d7\u5f71\u54cd *\/#content_views .toc,\/* \u7f16\u8f91\u5668\u5e38\u5728\u76ee\u5f55\u524d\u540e\u63d2\u5165\u7a7a p\uff08:empty \u4ecd\u5360 20px\uff09\uff0c\u4e00\u5e76\u53bb\u6389\u907f\u514d\u9876\u7a7a\u9699 *\/#content_views.markdown_views &gt; p:empty:has(+ .toc),#content_views.markdown_views &gt; .toc + p:empty,\/* \u5bcc\u6587\u672c\u65e7\u7248\u76ee\u5f55\u6807\u8bb0 *\/#content\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/118588.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-27T10:09:17+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/10\/20261027100914-6ae0784a8bebe.png\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"6 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/118588.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/118588.html\",\"name\":\"8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-10-27T10:09:17+00:00\",\"dateModified\":\"2026-10-27T10:09:17+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/118588.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/118588.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/118588.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/118588.html","og_locale":"zh_CN","og_type":"article","og_title":"8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\/* MD \/ \u5bcc\u6587\u672c\u4e2d\u7684 .toc\uff08\u542b\u535a\u5ba2\u56ed\u642c\u5bb6\u7b49\u5d4c\u5957\u7ed3\u6784\uff09\uff1b.toc-box \u5728\u4fa7\u680f\uff0c\u4e0d\u53d7\u5f71\u54cd *\/#content_views .toc,\/* \u7f16\u8f91\u5668\u5e38\u5728\u76ee\u5f55\u524d\u540e\u63d2\u5165\u7a7a p\uff08:empty \u4ecd\u5360 20px\uff09\uff0c\u4e00\u5e76\u53bb\u6389\u907f\u514d\u9876\u7a7a\u9699 *\/#content_views.markdown_views &gt; p:empty:has(+ .toc),#content_views.markdown_views &gt; .toc + p:empty,\/* \u5bcc\u6587\u672c\u65e7\u7248\u76ee\u5f55\u6807\u8bb0 *\/#content","og_url":"https:\/\/www.wsisp.com\/helps\/118588.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-10-27T10:09:17+00:00","og_image":[{"url":"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/10\/20261027100914-6ae0784a8bebe.png"}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"6 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/118588.html","url":"https:\/\/www.wsisp.com\/helps\/118588.html","name":"8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-10-27T10:09:17+00:00","dateModified":"2026-10-27T10:09:17+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/118588.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/118588.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/118588.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"8\u00d7910B4-32G NPU\u670d\u52a1\u5668 vLLM-Ascend\u90e8\u7f72Docker\u5b89\u88c5\u62a5\u544a"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/118588","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=118588"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/118588\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media\/118586"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=118588"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=118588"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=118588"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=118588"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}