{"id":4209,"date":"2024-06-28T14:27:48","date_gmt":"2024-06-28T06:27:48","guid":{"rendered":"https:\/\/www.aqwu.net\/wp\/?p=4209"},"modified":"2024-06-28T14:29:58","modified_gmt":"2024-06-28T06:29:58","slug":"meta-large-language-model-compiler%ef%bc%9a%e7%bc%96%e8%af%91%e5%99%a8%e4%bc%98%e5%8c%96%e7%9a%84%e5%9f%ba%e7%a1%80%e6%a8%a1%e5%9e%8b","status":"publish","type":"post","link":"https:\/\/www.aqwu.net\/wp\/?p=4209","title":{"rendered":"Meta Large Language Model Compiler\uff1a\u7f16\u8bd1\u5668\u4f18\u5316\u7684\u57fa\u7840\u6a21\u578b"},"content":{"rendered":"\n<p><strong>Meta\u63a8\u51fa\u4e86Meta LLM \u7f16\u8bd1\u5668\uff08<\/strong>Meta Large Language Model Compiler<strong>\uff09\uff01<\/strong> <\/p>\n\n\n\n<p>Meta LLM \u662f\u57fa\u4e8e Meta Code Llama \u6784\u5efa\u7684\u6a21\u578b\u7cfb\u5217\uff0c<strong>\u67097B\u548c13B\u4e24\u4e2a\u7248\u672c<\/strong> <\/p>\n\n\n\n<p>Meta LLM\u5177\u6709\u4ee3\u7801\u4f18\u5316\u548c\u7f16\u8bd1\u5668\u529f\u80fd\uff0c\u53ef\u4ee5\u6a21\u62df\u7f16\u8bd1\u5668\u3001\u9884\u6d4b\u4ee3\u7801\u5927\u5c0f\u4f18\u5316\u7684\u6700\u4f73\u8def\u5f84\uff0c\u5e76\u53ef\u4ee5\u53cd\u6c47\u7f16\u4ee3\u7801\u3002<\/p>\n\n\n\n<p>\u8be5\u7f16\u8bd1\u5668\u8fd8\u80fd\u9488\u5bf9\u7279\u5b9a\u7684\u4f18\u5316\u548c\u7f16\u8bd1\u5668\u4efb\u52a1\u8fdb\u884c\u5fae\u8c03 <\/p>\n\n\n\n<p>LLM\u7f16\u8bd1\u5668FTD\uff08\u5fae\u8c03\u7248\u672c\uff09\u5728\u4ee3\u7801\u5927\u5c0f\u4f18\u5316\u4e0a\u5b9e\u73b0\u4e865.24%\u7684\u63d0\u5347\uff08\u8d85\u8fc7\u4f18\u5316\u9009\u9879-Oz\uff09\uff0c<\/p>\n\n\n\n<p>GPT-4 Turbo\u4e3a0.03% \u5728\u53cd\u6c47\u7f16\u80fd\u529b\u4e0a\uff0cLLM\u7f16\u8bd1\u5668FTD\u8fbe\u5230\u4e860.96\u7684\u5f80\u8fd4BLEU\u5f97\u5206\uff0cGPT-4 Turbo\u4e3a0.43<\/p>\n\n\n\n<p>\u5927\u578b\u8bed\u8a00\u6a21\u578b \uff08LLMs\uff09 \u5728\u5404\u79cd\u8f6f\u4ef6\u5de5\u7a0b\u548c\u7f16\u7801\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u5353\u8d8a\u7684\u80fd\u529b\u3002\u7136\u800c\uff0c\u5b83\u4eec\u5728\u4ee3\u7801\u548c\u7f16\u8bd1\u5668\u4f18\u5316\u9886\u57df\u7684\u5e94\u7528\u4ecd\u672a\u5f97\u5230\u5145\u5206\u63a2\u7d22\u3002<\/p>\n\n\n\n<p>\u8bad\u7ec3LLMs\u662f\u8d44\u6e90\u5bc6\u96c6\u578b\u7684\uff0c\u9700\u8981\u5927\u91cf\u7684 GPU \u5c0f\u65f6\u548c\u5927\u91cf\u7684\u6570\u636e\u6536\u96c6\uff0c\u8fd9\u53ef\u80fd\u4ee4\u4eba\u671b\u800c\u5374\u6b65\u3002\u4e3a\u4e86\u5f25\u8865\u8fd9\u4e00\u5dee\u8ddd\uff0c\u6211\u4eec\u5f15\u5165\u4e86 Meta Large Language Model Compiler \uff08LLM Compiler\uff09\uff0c\u8fd9\u662f\u4e00\u5957\u5065\u58ee\u7684\u3001\u516c\u5f00\u53ef\u7528\u7684\u3001\u9884\u8bad\u7ec3\u7684\u6a21\u578b\uff0c\u4e13\u4e3a\u4ee3\u7801\u4f18\u5316\u4efb\u52a1\u800c\u8bbe\u8ba1\u3002Compiler \u5efa\u7acb\u5728 Code Llama \u7684\u57fa\u7840\u4e0a\uff0cLLM\u589e\u5f3a\u4e86\u5bf9\u7f16\u8bd1\u5668\u4e2d\u95f4\u8868\u793a \uff08IR\uff09\u3001\u6c47\u7f16\u8bed\u8a00\u548c\u4f18\u5316\u6280\u672f\u7684\u7406\u89e3\u3002<\/p>\n\n\n\n<p>\u8be5\u6a21\u578b\u5df2\u7ecf\u5728 5460 \u4ebf\u4e2a LLVM-IR \u548c\u6c47\u7f16\u4ee3\u7801\u7684\u5e9e\u5927\u8bed\u6599\u5e93\u4e0a\u8fdb\u884c\u4e86\u8bad\u7ec3\uff0c\u5e76\u7ecf\u8fc7\u4e86\u6307\u4ee4\u5fae\u8c03\u4ee5\u89e3\u91ca\u7f16\u8bd1\u5668\u884c\u4e3a\u3002LLM\u7f16\u8bd1\u5668\u5728\u5b9a\u5236\u7684\u5546\u4e1a\u8bb8\u53ef\u4e0b\u53d1\u5e03\uff0c\u5141\u8bb8\u5e7f\u6cdb\u91cd\u7528\uff0c\u5e76\u63d0\u4f9b\u4e24\u79cd\u5927\u5c0f\uff1a70 \u4ebf\u548c 130 \u4ebf\u4e2a\u53c2\u6570\u3002<\/p>\n\n\n\n<p>\u6211\u4eec\u8fd8\u5c55\u793a\u4e86\u8be5\u6a21\u578b\u7684\u5fae\u8c03\u7248\u672c\uff0c\u5c55\u793a\u4e86\u5176\u5728\u4f18\u5316\u4ee3\u7801\u5927\u5c0f\u4ee5\u53ca\u4ece x86_64 \u548c ARM \u6c47\u7f16\u53cd\u6c47\u7f16\u56de LLVM-IR \u65b9\u9762\u7684\u589e\u5f3a\u529f\u80fd\u3002\u8fd9\u4e9b\u5b9e\u73b0\u4e86\u81ea\u52a8\u8c03\u6574\u641c\u7d22\u7684 77% \u7684\u4f18\u5316\u6f5c\u529b\uff0c\u4ee5\u53ca 45% \u7684\u53cd\u6c47\u7f16\u5f80\u8fd4\uff0814% \u7684\u5b8c\u5168\u5339\u914d\uff09\u3002\u6b64\u7248\u672c\u65e8\u5728\u4e3a\u5b66\u672f\u7814\u7a76\u4eba\u5458\u548c\u884c\u4e1a\u4ece\u4e1a\u8005\u5728\u7f16\u8bd1\u5668\u4f18\u5316\u65b9\u9762\u7684\u8fdb\u4e00\u6b65\u7814\u7a76\u548c\u5f00\u53d1\u63d0\u4f9b\u53ef\u6269\u5c55\u3001\u7ecf\u6d4e\u9ad8\u6548\u7684\u57fa\u7840\u3002<\/p>\n\n\n\n<p><a href=\"https:\/\/scontent-hkg4-1.xx.fbcdn.net\/v\/t39.2365-6\/448997590_1496256481254967_2304975057370160015_n.pdf?_nc_cat=106&amp;ccb=1-7&amp;_nc_sid=3c67a6&amp;_nc_ohc=4Yn8V9DFdbsQ7kNvgFZG7U0&amp;_nc_ht=scontent-hkg4-1.xx&amp;gid=AXhmez8O_-tSNZ6-aZuRRcc&amp;oh=00_AYAgoxtgRGqXFnllF0rqnoPH-95blS1fgtD2qvKu8cjq6w&amp;oe=66842ACD\">Download the Paper\u00a0\u4e0b\u8f7d\u8bba\u6587<\/a><\/p>\n\n\n\n<p>\u6a21\u578b\u5730\u5740\uff1a<a href=\"https:\/\/huggingface.co\/collections\/facebook\/llm-compiler-667c5b05557fe99a9edd25cb\">LLM Compiler &#8211; a facebook Collection (huggingface.co)<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Meta\u63a8\u51fa\u4e86Meta LLM \u7f16\u8bd1\u5668\uff08Meta Large Language Model Compiler\uff09 [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"set","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[444,445,443,442],"tags":[500,404,499,501],"class_list":["post-4209","post","type-post","status-publish","format-standard","hentry","category-ai","category-ainews","category-llm","category-llms","tag-compiler","tag-llm","tag-meta","tag-501"],"views":2186,"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=\/wp\/v2\/posts\/4209","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4209"}],"version-history":[{"count":4,"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=\/wp\/v2\/posts\/4209\/revisions"}],"predecessor-version":[{"id":4213,"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=\/wp\/v2\/posts\/4209\/revisions\/4213"}],"wp:attachment":[{"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4209"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4209"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.aqwu.net\/wp\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4209"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}