{"id":8285,"date":"2026-05-17T18:56:40","date_gmt":"2026-05-17T10:56:40","guid":{"rendered":"https:\/\/infernews.com\/blog\/fast-lossless-llm-inference-via-dual-view-diffusion-decoding\/"},"modified":"2026-05-17T18:56:40","modified_gmt":"2026-05-17T10:56:40","slug":"fast-lossless-llm-inference-via-dual-view-diffusion-decoding","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/fast-lossless-llm-inference-via-dual-view-diffusion-decoding\/","title":{"rendered":"Orthrus\u5982\u4f55\u4ee4Qwen3\u751f\u6210\u66f4\u5feb"},"content":{"rendered":"<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/05\/orthrus_logo-b20acfbdd2f2.jpg\" alt=\"Orthrus logo\"><\/figure>\n<p>Orthrus \u662f\u4e00\u500b\u570d\u7e5e Qwen3 \u6a21\u578b\u5efa\u7acb\u7684\u751f\u6210\u6846\u67b6\uff0c\u91cd\u9ede\u4e0d\u662f\u505a\u5168\u65b0\u804a\u5929\u6a21\u578b\uff0c\u800c\u662f\u60f3\u8fa6\u6cd5\u4ee4\u6587\u5b57\u751f\u6210\u66f4\u5feb\uff0c\u540c\u6642\u4fdd\u6301\u8207\u539f\u672c\u57fa\u790e\u6a21\u578b\u4e00\u81f4\u7684\u8f38\u51fa\u5206\u4f48\u3002\u5c0d\u4e00\u822c\u8b80\u8005\u4f86\u8aaa\uff0c\u53ef\u4ee5\u7406\u89e3\u6210\u5b83\u60f3\u4fdd\u7559\u50b3\u7d71\u9010\u5b57\u751f\u6210\u7684\u6e96\u78ba\u611f\uff0c\u53c8\u501f\u7528\u64f4\u6563\u5f0f\u4e26\u884c\u751f\u6210\u7684\u901f\u5ea6\u512a\u52e2\u3002<\/p>\n<p>\u9019\u985e\u5de5\u5177\u4e3b\u8981\u91dd\u5c0d\u5927\u578b\u6a21\u578b\u751f\u6210\u6642\u300c\u8981\u9010\u500b\u5b57\u7b49\u300d\u7684\u6a3d\u9838\u3002Orthrus \u63d0\u51fa\u96d9\u91cd\u67b6\u69cb\u505a\u6cd5\uff0c\u8b93\u540c\u4e00\u500b\u6a21\u578b\u540c\u6642\u5177\u5099\u5169\u7a2e\u89c0\u770b\u65b9\u5f0f\uff0c\u4e26\u5f37\u8abf\u7d50\u679c\u662f<strong>\u7121\u640d<\/strong>\u7684\uff0c\u4e5f\u5c31\u662f\u76ee\u6a19\u4e26\u975e\u7528\u8fd1\u4f3c\u7b54\u6848\u63db\u901f\u5ea6\uff1b\u6839\u64da\u5c08\u6848\u8cc7\u6599\uff0c\u751f\u6210\u53ef\u6709\u6700\u9ad8\u7d04 7.8 \u500d\u52a0\u901f\u3002<\/p>\n<p>\u5982\u679c\u4f60\u60f3\u4e0a\u624b\uff0c\u6700\u76f4\u63a5\u65b9\u6cd5\u4e0d\u662f\u81ea\u884c\u8a13\u7df4\uff0c\u800c\u662f\u5148\u8a66\u7528\u4f5c\u8005\u63d0\u4f9b\u7684\u6a21\u578b\u6aa2\u67e5\u9ede\uff0c\u518d\u7528 Hugging Face \u7684\u5e38\u898b\u8f09\u5165\u6d41\u7a0b\u505a\u63a8\u7406\u3002\u73fe\u6642\u516c\u958b\u578b\u865f\u5305\u62ec <strong>Orthrus-Qwen3-1.7B<\/strong>\u3001<strong>Orthrus-Qwen3-4B<\/strong> \u548c <strong>Orthrus-Qwen3-8B<\/strong>\uff0c\u5206\u5225\u5c0d\u61c9 Qwen3 \u7684 1.7B\u30014B \u8207 8B \u57fa\u790e\u6a21\u578b\u3002<\/p>\n<p>\u503c\u5f97\u7559\u610f\u7684\u662f\uff0c\u5b83\u4e0d\u662f\u9760\u628a\u6574\u500b\u6a21\u578b\u91cd\u8a13\u4f86\u63db\u901f\u5ea6\uff0c\u800c\u662f\u53ea\u5fae\u8abf\u90e8\u5206\u53c3\u6578\uff0c\u57fa\u790e LLM \u4fdd\u6301\u51cd\u7d50\uff0c\u540c\u6642\u5169\u7a2e\u751f\u6210\u8996\u89d2\u53ef\u5171\u7528\u540c\u4e00\u5957\u9ad8\u4fdd\u771f KV cache\u3002\u5c0d\u90e8\u7f72\u8005\u4f86\u8aaa\uff0c\u9019\u4ee3\u8868\u5b83\u9664\u4e86\u8b1b\u6c42\u5feb\uff0c\u4ea6\u6709\u610f\u63a7\u5236\u984d\u5916\u8a18\u61b6\u9ad4\u6210\u672c\uff0c\u9019\u9ede\u5c0d\u9577\u8f38\u51fa\u6216\u9ad8\u983b\u63a8\u7406\u5834\u666f\u7279\u5225\u5be6\u969b\u3002<\/p>\n<ul>\n<li>\u4ee5 Qwen3 \u70ba\u9aa8\u5e79\uff0c\u73fe\u6709 1.7B\u30014B\u30018B \u5e7e\u500b\u7248\u672c<\/li>\n<li>\u91cd\u9ede\u5728\u63d0\u5347\u751f\u6210\u541e\u5410\uff0c\u800c\u975e\u6539\u8b8a\u6a21\u578b\u7528\u9014<\/li>\n<li>\u5f37\u8abf\u7d50\u679c\u8207\u539f\u57fa\u790e\u6a21\u578b\u4fdd\u6301\u4e00\u81f4\uff0c\u800c\u975e\u8fd1\u4f3c\u52a0\u901f<\/li>\n<li>\u984d\u5916\u8a18\u61b6\u9ad4\u958b\u92b7\u8f03\u4f4e\uff0c\u8f03\u9069\u5408\u63a8\u7406\u90e8\u7f72\u8a55\u4f30<\/li>\n<li>\u5c0d\u7814\u7a76\u8005\u3001\u6a21\u578b\u5de5\u7a0b\u5e2b\u53ca\u9700\u8981\u5927\u91cf\u6587\u5b57\u751f\u6210\u7684\u5718\u968a\u8f03\u6709\u53c3\u8003\u50f9\u503c<\/li>\n<\/ul>\n<p>\u6574\u9ad4\u4f86\u770b\uff0cOrthrus \u6700\u5438\u5f15\u4e4b\u8655\u5728\u65bc\u5b83\u628a\u300c\u5feb\u300d\u8207\u300c\u4e0d\u8d70\u6a23\u300d\u653e\u5728\u540c\u4e00\u500b\u65b9\u6848\u5167\u8655\u7406\u3002\u82e5\u4f60\u6b63\u95dc\u6ce8\u672c\u5730\u6216\u4f3a\u670d\u5668\u7aef LLM \u63a8\u7406\u6548\u80fd\uff0c\u5c24\u5176\u5df2\u7d93\u5728\u4f7f\u7528 Qwen3 \u751f\u614b\uff0c\u9019\u500b\u5c08\u6848\u5f88\u9069\u5408\u4f5c\u70ba\u5be6\u9a57\u8207\u6bd4\u8f03\u57fa\u6e96\uff1b\u81f3\u65bc\u8207 vLLM \u6216 SGLang \u7684\u66f4\u539f\u751f\u6574\u5408\uff0c\u5247\u4f3c\u4e4e\u4ecd\u5728\u5f8c\u7e8c\u898f\u5283\u4e2d\u3002<\/p>\n<p><strong>\u7db2\u5740\uff1a<\/strong> <a href=\"https:\/\/github.com\/chiennv2000\/orthrus\" rel=\"noopener noreferrer\">https:\/\/github.com\/chiennv2000\/orthrus<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u9019\u500b\u5c08\u6848\u5617\u8a66\u5728\u4e0d\u6539\u8b8a\u539f\u672c\u8f38\u51fa\u54c1\u8cea\u4e0b\uff0c\u52a0\u5feb\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u751f\u6210\u901f\u5ea6\u3002\u5c0d\u60f3\u7528Qwen3\u505a\u63a8\u7406\u7684\u4eba\u5c24\u5176\u503c\u5f97\u7559\u610f\u3002<\/p>\n","protected":false},"author":8,"featured_media":8284,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[133,185,76,127,189],"tags":[],"class_list":["post-8285","post","type-post","status-publish","format-standard","hentry","category-133","category-qwen","category-76","category-127","category-189"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8285","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=8285"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8285\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/8284"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=8285"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=8285"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=8285"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}