{"id":9926,"date":"2026-07-07T09:41:53","date_gmt":"2026-07-07T01:41:53","guid":{"rendered":"https:\/\/infernews.com\/blog\/supra-router-51m-ai\/"},"modified":"2026-07-07T09:41:53","modified_gmt":"2026-07-07T01:41:53","slug":"supra-router-51m-ai","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/supra-router-51m-ai\/","title":{"rendered":"Supra-Router-51M\uff1aAI \u5605\u667a\u80fd\u5206\u6d41"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/lClDSdp9BkKyv_VGWSQbt.png\" alt=\"Og image\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Supra-Router-51M \u4fc2\u4e00\u500b\u57fa\u65bc SupraLabs\/Supra-1.5-50M-Base-exp \u57fa\u790e\u6a21\u578b\u5fae\u8abf\u800c\u6210\u5605\u5c0f\u578b\u8a9e\u8a00\u6a21\u578b\uff08SLM\uff09\uff0c\u5b9a\u4f4d\u4fc2\u591a\u6a21\u578b\u751f\u614b\u7cfb\u7d71\u5605\u908a\u7de3\u8def\u7531\u9598\u9053\u3002\u4f62\u5605\u4e3b\u8981\u7528\u9014\u4fc2\u5206\u6790\u7528\u6236\u8f38\u5165\u5605\u63d0\u793a\u8a5e\uff0c\u5224\u65b7\u8a72\u8acb\u6c42\u9069\u5408\u7531\u672c\u5730\u5c0f\u578b\u6a21\u578b\u8655\u7406\uff0c\u5b9a\u4fc2\u9700\u8981\u8f49\u4ea4\u96f2\u7aef\u5927\u578b\u6a21\u578b\uff0c\u5f9e\u800c\u512a\u5316\u6574\u9ad4\u904b\u7b97\u8cc7\u6e90\u5206\u914d\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6a21\u578b\u63a1\u7528 <strong>Multi-Task Sequence Generation<\/strong>\uff08\u591a\u4efb\u52d9\u5e8f\u5217\u751f\u6210\uff09\u5605\u65b9\u6cd5\uff0c\u5514\u4fc2\u55ae\u7d14\u505a\u4e8c\u5143\u5206\u985e\u3002\u4f62\u6703\u5148\u8f38\u51fa\u4e00\u9023\u4e32\u7d50\u69cb\u5316\u8cc7\u8a0a\uff0c\u5305\u62ec\u8a9e\u7fa9\u9818\u57df\uff08Domain\uff09\u3001\u8907\u96dc\u5ea6\u8a55\u5206\uff08Complexity 1-5\uff09\u3001\u6578\u5b78\u9700\u6c42\uff08Math\uff09\u3001\u7a0b\u5f0f\u78bc\u9700\u6c42\uff08Code\uff09\uff0c\u6700\u5f8c\u5148\u8f38\u51fa\u8def\u7531\u6c7a\u5b9a\uff08Route\uff09\u3002\u5462\u7a2e\u8a2d\u8a08\u4ee4\u6a21\u578b\u55ba\u505a\u6700\u7d42\u5224\u65b7\u524d\uff0c\u5148\u5efa\u7acb\u5167\u90e8\u7279\u5fb5\u6fc0\u6d3b\u5716\uff0c\u907f\u514d\u7d30\u6a21\u578b\u5bb9\u6613\u51fa\u73fe\u5605\u6b0a\u91cd\u5d29\u584c\u554f\u984c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u8a13\u7df4\u65b9\u9762\u4f7f\u7528\u5497 SupraLabs\/Prompt-Routing-Dataset \u6578\u64da\u96c6\uff08992 \u7b46\u6a23\u672c\uff09\uff0c\u8a13\u7df4 5 \u500b Epoch\uff0c\u4f46\u55ba Epoch 3\uff08eval_loss \u70ba 0.1342\uff09\u5df2\u9054\u5230\u5cf0\u503c\u5f8c\u5c31\u56de\u9000\u4fdd\u5b58\uff0c\u907f\u514d\u5f8c\u6bb5\u904e\u5ea6\u64ec\u5408\u3002\u652f\u63f4\u5e8f\u5217\u9577\u5ea6\u9054 3840 tokens\uff0c\u4f7f\u7528 bfloat16 \u7cbe\u5ea6\uff0c\u55ba CPU\/GPU \u90fd\u80fd\u505a\u5230\u6b21\u6beb\u79d2\u7d1a\u63a8\u8ad6\u901f\u5ea6\u3002\u63a8\u8ad6\u6642\u5fc5\u9808\u7528 greedy decoding\uff08do_sample=False\uff09\u4ee5\u78ba\u4fdd\u6c7a\u7b56\u7a69\u5b9a\u6027\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u4f7f\u7528\u6642\u9700\u8981\u5c07\u7528\u6236\u67e5\u8a62\u5305\u88dd\u55ba\u7279\u5b9a\u683c\u5f0f\uff08<code>Task: [Prompt]\nAnalysis:<\/code>\uff09\u5165\u9762\uff0c\u6a21\u578b\u6703\u8f38\u51fa pipe-separated \u5605\u7d50\u69cb\u5316\u5b57\u4e32\uff0c\u5305\u542b\u5b8c\u6574\u5605\u63d0\u793a\u8a5e\u9059\u6e2c\u8cc7\u8a0a\u3002\u914d\u5408 FastAPI \u7b49\u9598\u9053\u6846\u67b6\u5c31\u53ef\u4ee5\u90e8\u7f72\u6210\u751f\u7522\u74b0\u5883\u5605\u81ea\u52d5\u5206\u6d41\u7cfb\u7d71\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u7531\u65bc\u6a21\u578b\u53ea\u6709 51.7M \u53c3\u6578\uff0c\u904b\u884c\u9580\u6abb\u6975\u4f4e\uff0c\u9069\u5408\u908a\u7de3\u88dd\u7f6e\u540c\u9700\u8981\u5373\u6642\u6c7a\u7b56\u5605\u5834\u666f\uff0c\u4f46\u4f62\u5605\u8def\u7531\u6e96\u7e69\u5ea6\u4ecd\u7136\u53d7\u5236\u65bc\u8a13\u7df4\u6578\u64da\u96c6\u898f\u6a21\u8f03\u7d30\u5462\u500b\u9650\u5236\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u91cd\u9ede\u6458\u8981\uff1a<\/strong><br \/>\n&#8211; \u57fa\u65bc SupraLabs\/Supra-1.5-50M-Base-exp \u5fae\u8abf\uff0c\u50c5 51.7M \u53c3\u6578<br \/>\n&#8211; \u63a1\u7528\u591a\u4efb\u52d9\u5e8f\u5217\u751f\u6210\u67b6\u69cb\uff0c\u5148\u5206\u6790\u63d0\u793a\u8a5e\u7279\u5fb5\u518d\u6c7a\u5b9a\u8def\u7531<br \/>\n&#8211; \u8a13\u7df4\u6578\u64da\u96c6 992 \u7b46\uff0c3840 tokens \u5e8f\u5217\u9577\u5ea6\uff0c\u6b21\u6beb\u79d2\u7d1a\u63a8\u8ad6<br \/>\n&#8211; \u652f\u63f4 FastAPI \u9598\u9053\u90e8\u7f72\uff0c\u9700\u4f7f\u7528 greedy decoding \u78ba\u4fdd\u7a69\u5b9a<br \/>\n&#8211; \u9069\u5408\u908a\u7de3\u88dd\u7f6e\u8cc7\u6e90\u8abf\u5ea6\uff0c\u9650\u5236\u4fc2\u8a13\u7df4\u6578\u64da\u898f\u6a21\u8f03\u5c0f<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/huggingface.co\/SupraLabs\/Supra-Router-51M\" rel=\"noopener noreferrer\" target=\"_blank\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>51.7M \u53c3\u6578\u5605\u5fae\u578b\u8def\u7531\u6a21\u578b\uff0c\u81ea\u52d5\u5224\u65b7\u8acb\u6c42\u8a72\u4ea4\u7531\u672c\u5730\u5c0f\u578b\u6a21\u578b\u5b9a\u96f2\u7aef\u5927\u578b\u6a21\u578b\u8655\u7406\u3002<\/p>\n","protected":false},"author":8,"featured_media":9925,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,40,76,197,199],"tags":[],"class_list":["post-9926","post","type-post","status-publish","format-standard","hentry","category-133","category-llama","category-76","category-framework","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9926","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=9926"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9926\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9925"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9926"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9926"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9926"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}