{"id":14206,"date":"2024-06-28T09:17:33","date_gmt":"2024-06-28T01:17:33","guid":{"rendered":"https:\/\/www.1ai.net\/?p=14206"},"modified":"2024-06-28T09:17:33","modified_gmt":"2024-06-28T01:17:33","slug":"90-270-%e4%ba%bf%e5%8f%82%e6%95%b0%ef%bc%8c%e8%b0%b7%e6%ad%8c%e5%8f%91%e5%b8%83-gemma-2-%e5%bc%80%e6%ba%90-ai-%e6%a8%a1%e5%9e%8b%ef%bc%9a%e6%80%a7%e8%83%bd%e5%8a%9b%e5%8e%8b%e5%90%8c%e7%ba%a7%e3%80%81","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/14206.html","title":{"rendered":"Google releases Gemma 2 open source AI model with 9\/27 billion parameters: performance is better than its peers, and can be run on a single A100\/H100 GPU"},"content":{"rendered":"<p data-vmark=\"a0d3\"><a href=\"https:\/\/www.1ai.net\/en\/tag\/%e8%b0%b7%e6%ad%8c\" title=\"\u3010\u67e5\u770b\u542b\u6709[\u8c37\u6b4c]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">\u8c37\u6b4c<\/a>\u516c\u53f8\u6628\u65e5\u53d1\u5e03<a href=\"https:\/\/blog.google\/technology\/developers\/google-gemma-2\/\" target=\"_blank\" rel=\"noopener\">\u65b0\u95fb\u7a3f<\/a>\uff0c\u9762\u5411\u5168\u7403\u7814\u7a76\u4eba\u5458\u548c\u5f00\u53d1\u4eba\u5458\u53d1\u5e03 <a href=\"https:\/\/www.1ai.net\/en\/tag\/gemma-2\" title=\"\u3010\u67e5\u770b\u542b\u6709[Gemma 2]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">Gemma 2<\/a> <a href=\"https:\/\/www.1ai.net\/en\/tag\/%e5%a4%a7%e8%af%ad%e8%a8%80%e6%a8%a1%e5%9e%8b\" title=\"\u3010\u67e5\u770b\u542b\u6709[\u5927\u8bed\u8a00\u6a21\u578b]\u6807\u7b7e\u7684\u6587\u7ae0\u3011\" target=\"_blank\" rel=\"noopener\">\u5927\u8bed\u8a00\u6a21\u578b<\/a>\uff0c<strong>\u5171\u6709 90 \u4ebf\u53c2\u6570\uff089B\uff09\u548c 270 \u4ebf\u53c2\u6570\uff0827B\uff09\u4e24\u79cd\u5927\u5c0f\u3002<\/strong><\/p>\n<p data-vmark=\"6b86\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-14207\" title=\"9b1a398e-c05c-4882-90a6-185496c40029\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2024\/06\/9b1a398e-c05c-4882-90a6-185496c40029.jpg\" alt=\"9b1a398e-c05c-4882-90a6-185496c40029\" width=\"1024\" height=\"427\" \/><\/p>\n<p data-vmark=\"1ad9\">Gemma 2 \u5927\u8bed\u8a00\u6a21\u578b\u76f8\u6bd4\u8f83\u7b2c\u4e00\u4ee3\uff0c\u63a8\u7406\u6027\u80fd\u66f4\u9ad8\u3001\u6548\u7387\u66f4\u9ad8\uff0c\u5e76\u5728\u5b89\u5168\u6027\u65b9\u9762\u53d6\u5f97\u4e86\u91cd\u5927\u8fdb\u6b65\u3002<\/p>\n<p data-vmark=\"6e51\">\u8c37\u6b4c\u5728\u65b0\u95fb\u7a3f\u4e2d\u8868\u793a\uff0cGemma 2-27B \u6a21\u578b\u7684\u6027\u80fd\u5ab2\u7f8e\u4e24\u500d\u89c4\u6a21\u7684\u4e3b\u6d41\u6a21\u578b\uff0c\u800c\u4e14\u53ea\u9700\u8981\u4e00\u7247\u82f1\u4f1f\u8fbe H100 ensor Core GPU \u6216 TPU \u4e3b\u673a\u5c31\u80fd\u5b9e\u73b0\u8fd9\u79cd\u6027\u80fd\uff0c\u4ece\u800c\u5927\u5927\u964d\u4f4e\u4e86\u90e8\u7f72\u6210\u672c\u3002<\/p>\n<p data-vmark=\"7258\">Gemma 2-9B \u6a21\u578b\u4f18\u4e8e Llama 3 8B \u548c\u5176\u4ed6\u7c7b\u4f3c\u89c4\u6a21\u7684\u5f00\u6e90\u6a21\u578b\u3002\u8c37\u6b4c\u8fd8\u8ba1\u5212\u5728\u672a\u6765\u51e0\u4e2a\u6708\u53d1\u5e03\u53c2\u6570\u4e3a 26 \u4ebf\u7684 Gemma 2 \u6a21\u578b\uff0c\u66f4\u9002\u5408\u667a\u80fd\u624b\u673a\u7684\u4eba\u5de5\u667a\u80fd\u5e94\u7528\u573a\u666f\u3002<\/p>\n<p data-vmark=\"4ce9\">\u8c37\u6b4c\u8868\u793a\u4e3a Gemma 2 \u91cd\u65b0\u8bbe\u8ba1\u4e86\u6574\u4f53\u67b6\u6784\uff0c\u5b9e\u73b0\u5353\u8d8a\u7684\u6027\u80fd\u548c\u63a8\u7406\u6548\u7387\u3002<\/p>\n<p data-vmark=\"4ce9\">Gemma 2 \u4e3b\u8981\u7279\u70b9\u5982\u4e0b\uff1a<\/p>\n<h3 data-vmark=\"f150\">\u6027\u80fd\u4f18\u5f02\uff1a<\/h3>\n<p data-vmark=\"e42a\">27B \u7248\u672c\u5728\u540c\u89c4\u6a21\u7ea7\u522b\u4e2d\u6027\u80fd\u6700\u4f73\uff0c\u751a\u81f3\u6bd4\u4e24\u500d\u4e8e\u5176\u5c3a\u5bf8\u7684\u673a\u578b\u66f4\u5177\u7ade\u4e89\u529b\u30029B \u7248\u672c\u7684\u6027\u80fd\u5728\u540c\u7c7b\u4ea7\u54c1\u4e2d\u4e5f\u5904\u4e8e\u9886\u5148\u5730\u4f4d\uff0c\u8d85\u8fc7\u4e86 Llama 3 8B \u548c\u5176\u4ed6\u540c\u89c4\u6a21\u7684\u5f00\u653e\u6a21\u578b\u3002<\/p>\n<h3 data-vmark=\"5a4b\">\u6548\u7387\u548c\u6210\u672c\uff1a<\/h3>\n<p data-vmark=\"831c\">27B Gemma 2 \u6a21\u578b\u53ef\u5728\u5355\u4e2a\u8c37\u6b4c\u4e91 TPU \u4e3b\u673a\u3001\u82f1\u4f1f\u8fbe A100 80GB Tensor Core GPU \u6216\u82f1\u4f1f\u8fbe H100 Tensor Core GPU \u4e0a\u4ee5\u5168\u7cbe\u5ea6\u9ad8\u6548\u8fd0\u884c\u63a8\u7406\uff0c\u5728\u4fdd\u6301\u9ad8\u6027\u80fd\u7684\u540c\u65f6\u5927\u5e45\u964d\u4f4e\u6210\u672c\u3002\u8fd9\u4f7f\u5f97\u4eba\u5de5\u667a\u80fd\u90e8\u7f72\u66f4\u5bb9\u6613\u5b9e\u73b0\uff0c\u9884\u7b97\u4e5f\u66f4\u5408\u7406\u3002<\/p>\n<h3 data-vmark=\"b7c1\">\u8de8\u786c\u4ef6\u5feb\u901f\u63a8\u7406<\/h3>\n<p data-vmark=\"946b\">Gemma 2 \u7ecf\u8fc7\u4f18\u5316\uff0c\u53ef\u5728\u5404\u79cd\u786c\u4ef6\uff08\u4ece\u529f\u80fd\u5f3a\u5927\u7684\u6e38\u620f\u7b14\u8bb0\u672c\u7535\u8111\u548c\u9ad8\u7aef\u53f0\u5f0f\u673a\u5230\u57fa\u4e8e\u4e91\u7684\u8bbe\u7f6e\uff09\u4e0a\u4ee5\u60ca\u4eba\u7684\u901f\u5ea6\u8fd0\u884c\u3002<\/p>\n<p data-vmark=\"514a\">\u5728 Google AI Studio \u4e2d\u5c1d\u8bd5\u5168\u7cbe\u5ea6\u7684 Gemma 2\uff0c\u5728 CPU \u4e0a\u4f7f\u7528\u00a0<span class=\"link-text-start-with-http\">Gemma.cpp<\/span>\u00a0\u7684\u91cf\u5316\u7248\u672c\u89e3\u9501\u672c\u5730\u6027\u80fd\uff0c\u6216\u901a\u8fc7 Hugging Face Transformers \u5728\u914d\u5907 NVIDIA RTX \u6216 GeForce RTX \u7684\u5bb6\u7528\u7535\u8111\u4e0a\u8fdb\u884c\u5c1d\u8bd5\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Google issued a press release yesterday, issuing Gemma 2 large-language models for researchers and developers around the world, totalling 9 billion parameters (9B) and 27 billion parameters (27B). Gemma 2 Large-language models are more deduced and more efficient than the first generation and have made significant progress in safety. Google stated in its press release that the Gemma 2-27B model performed twice as well as the mainstream model, and that it could only be achieved by an English-led H100 endor Core GPU or TPU host, thus significantly reducing deployment costs. Gemma 2-9B model better than the open source of Llama 3 8B and other similar scales<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[146],"tags":[3252,706,713,281],"collection":[],"class_list":["post-14206","post","type-post","status-publish","format-standard","hentry","category-news","tag-gemma-2","tag-706","tag-ai","tag-281"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/14206","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=14206"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/14206\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=14206"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=14206"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=14206"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=14206"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}