{"id":22622,"date":"2024-11-06T01:04:06","date_gmt":"2024-11-05T17:04:06","guid":{"rendered":"https:\/\/www.1ai.net\/?p=22622"},"modified":"2024-11-05T21:08:25","modified_gmt":"2024-11-05T13:08:25","slug":"%e8%85%be%e8%ae%af%e6%8e%a8%e5%87%ba-hunyuan-large-%e5%a4%a7%e6%a8%a1%e5%9e%8b%ef%bc%9a389b-%e6%80%bb%e5%8f%82%e6%95%b0%ef%bc%8c%e4%b8%9a%e7%95%8c%e5%b7%b2%e5%bc%80%e6%ba%90%e5%9f%ba%e4%ba%8e-transfor","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/22622.html","title":{"rendered":"Tencent Launches Hunyuan-Large Large Model: 389B Total Parameters, Industry's Largest Transformer-Based MoE Model Open-Sourced"},"content":{"rendered":"<p><a href=\"https:\/\/www.1ai.net\/en\/tag\/%e8%85%be%e8%ae%af\" title=\"\u3010\u67e5\u770b\u542b\u6709[\u817e\u8baf]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">\u817e\u8baf<\/a>\u5ba3\u5e03\u63a8\u51fa\u00a0Hunyuan-Large <a href=\"https:\/\/www.1ai.net\/en\/tag\/%e5%a4%a7%e6%a8%a1%e5%9e%8b\" title=\"\u3010\u67e5\u770b\u542b\u6709[\u5927\u6a21\u578b]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">\u5927\u6a21\u578b<\/a>\uff0c\u5b98\u65b9\u8868\u793a\u8fd9\u662f<strong>\u76ee\u524d\u4e1a\u754c\u5df2\u7ecf<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e5%bc%80%e6%ba%90\" title=\"\u3010\u67e5\u770b\u542b\u6709[\u5f00\u6e90]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">\u5f00\u6e90<\/a>\u7684\u57fa\u4e8e Transformer \u7684\u6700\u5927 MoE \u6a21\u578b<\/strong>\uff0c\u62e5\u6709 3890 \u4ebf\u603b\u53c2\u6570\uff08389B\uff09\u548c 520 \u4ebf\u6fc0\u6d3b\u53c2\u6570\uff0852B\uff09\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-22623\" title=\"0473510dj00smhb4g006md000v900jjp\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2024\/11\/0473510dj00smhb4g006md000v900jjp.jpg\" alt=\"0473510dj00smhb4g006md000v900jjp\" width=\"1125\" height=\"703\" \/><\/p>\n<p>\u817e\u8baf\u5728 Hugging Face \u5f00\u6e90\u4e86 Hunyuan-A52B-Pretrain\u00a0\u3001\u00a0Hunyuan-A52B-Instruct\u00a0\u548c Hunyuan-A52B-Instruct-FP8\u3002\u5e76\u53d1\u5e03\u4e86\u6280\u672f\u62a5\u544a\u548c\u8bad\u7ec3\u63a8\u7406\u64cd\u4f5c\u624b\u518c\uff0c\u8be6\u7ec6\u4ecb\u7ecd\u4e86\u6a21\u578b\u80fd\u529b\u548c\u8bad\u7ec3\u4e0e\u63a8\u7406\u7684\u64cd\u4f5c\u3002<\/p>\n<p>\u5176\u4e2d\u6a21\u578b\u6280\u672f\u4f18\u52bf\u5982\u4e0b\uff1a<\/p>\n<ul>\n<li><strong>\u9ad8\u8d28\u91cf\u5408\u6210\u6570\u636e<\/strong>\uff1a\u901a\u8fc7\u5408\u6210\u6570\u636e\u589e\u5f3a\u8bad\u7ec3\uff0cHunyuan-Large \u80fd\u591f\u5b66\u4e60\u5230\u66f4\u4e30\u5bcc\u7684\u8868\u793a\uff0c\u5904\u7406\u957f\u4e0a\u4e0b\u6587\u8f93\u5165\uff0c\u5e76\u66f4\u597d\u5730\u6cdb\u5316\u5230\u672a\u89c1\u6570\u636e<\/li>\n<li><strong>KV \u7f13\u5b58\u538b\u7f29<\/strong>\uff1a\u91c7\u7528\u5206\u7ec4\u67e5\u8be2\u6ce8\u610f\u529b\uff08GQA\uff09\u548c\u8de8\u5c42\u6ce8\u610f\u529b\uff08CLA\uff09\u7b56\u7565\uff0c\u663e\u8457\u51cf\u5c11\u4e86 KV \u7f13\u5b58\u7684\u5185\u5b58\u5360\u7528\u548c\u8ba1\u7b97\u5f00\u9500\uff0c\u63d0\u9ad8\u4e86\u63a8\u7406\u541e\u5410<\/li>\n<li><strong>\u4e13\u5bb6\u7279\u5b9a\u5b66\u4e60\u7387\u7f29\u653e<\/strong>\uff1a\u4e3a\u4e0d\u540c\u4e13\u5bb6\u8bbe\u7f6e\u4e0d\u540c\u7684\u5b66\u4e60\u7387\uff0c\u786e\u4fdd\u6bcf\u4e2a\u5b50\u6a21\u578b\u90fd\u80fd\u6709\u6548\u5730\u4ece\u6570\u636e\u4e2d\u5b66\u4e60\uff0c\u5e76\u4e3a\u6574\u4f53\u6027\u80fd\u505a\u51fa\u8d21\u732e<\/li>\n<li><strong>\u957f\u4e0a\u4e0b\u6587\u5904\u7406\u80fd\u529b<\/strong>\uff1a\u9884\u8bad\u7ec3\u6a21\u578b\u652f\u6301\u9ad8\u8fbe 256K \u7684\u6587\u672c\u5e8f\u5217\uff0cInstruct \u6a21\u578b\u652f\u6301 128K \u7684\u6587\u672c\u5e8f\u5217\uff0c\u663e\u8457\u63d0\u5347\u4e86\u957f\u4e0a\u4e0b\u6587\u4efb\u52a1\u7684\u5904\u7406\u80fd\u529b<\/li>\n<li><strong>\u5e7f\u6cdb\u7684\u57fa\u51c6\u6d4b\u8bd5<\/strong>\uff1a\u5728\u591a\u79cd\u8bed\u8a00\u548c\u4efb\u52a1\u4e0a\u8fdb\u884c\u5e7f\u6cdb\u5b9e\u9a8c\uff0c\u9a8c\u8bc1\u4e86 Hunyuan-Large \u7684\u5b9e\u9645\u5e94\u7528\u6548\u679c\u548c\u5b89\u5168\u6027<\/li>\n<\/ul>\n<p data-vmark=\"461d\">\u9644\u76f8\u5173\u94fe\u63a5\u5982\u4e0b\uff1a<\/p>\n<ul class=\"medium-size list-paddingleft-2\">\n<li>\n<p data-vmark=\"c8a2\"><strong>\u8bba\u6587<\/strong>\uff1a<a href=\"https:\/\/arxiv.org\/pdf\/2411.02265\" target=\"_blank\" rel=\"noopener\"><span class=\"link-text-start-with-http\">https:\/\/arxiv.org\/pdf\/2411.02265<\/span><\/a><\/p>\n<\/li>\n<li>\n<p data-vmark=\"2925\"><strong>Github<\/strong>\uff1a<a href=\"https:\/\/github.com\/Tencent\/Tencent-Hunyuan-Large\" target=\"_blank\" rel=\"noopener\"><span class=\"link-text-start-with-http\">https:\/\/github.com\/Tencent\/Tencent-Hunyuan-Large<\/span><\/a><\/p>\n<\/li>\n<li>\n<p data-vmark=\"9218\"><strong>Huggingface<\/strong>\uff1a<a href=\"https:\/\/huggingface.co\/tencent\/Tencent-Hunyuan-Large\" target=\"_blank\" rel=\"noopener\"><span class=\"link-text-start-with-http\">https:\/\/huggingface.co\/tencent\/Tencent-Hunyuan-Large<\/span><\/a><\/p>\n<\/li>\n<li>\n<p data-vmark=\"6859\"><strong>\u817e\u8baf\u4e91<\/strong>\uff1a<a href=\"https:\/\/cloud.tencent.com\/product\/hunyuan\" target=\"_blank\" rel=\"noopener\"><span class=\"link-text-start-with-http\">https:\/\/cloud.tencent.com\/product\/hunyuan<\/span><\/a><\/p>\n<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Tencent announced the launch of Hunyuan-Large, the largest Transformer-based MoE model open-sourced in the industry, with 389 billion total parameters (389B) and 52 billion activation parameters (52B). Tencent open-sourced Hunyuan-A52B-Pretrain, Hunyuan-A52B-Instruct, and Hunyuan-A52B-Instruct-FP8 at Hugging Face, and released a technical report and a training and inference operation manual, which introduces in detail the model's capabilities and the operation of training and inference. Among the model technology advantages are as follows: High-quality synthetic data: by synthesizing the number of<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[146],"tags":[216,219,323],"collection":[],"class_list":["post-22622","post","type-post","status-publish","format-standard","hentry","category-news","tag-216","tag-219","tag-323"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/22622","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=22622"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/22622\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=22622"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=22622"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=22622"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=22622"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}