{"id":4173,"date":"2024-02-22T09:14:48","date_gmt":"2024-02-22T01:14:48","guid":{"rendered":"https:\/\/www.1ai.net\/?p=4173"},"modified":"2024-02-22T09:14:48","modified_gmt":"2024-02-22T01:14:48","slug":"%e8%b0%b7%e6%ad%8c-deepmind-%e6%8e%a8%e5%87%ba-gemma-ai-%e6%a8%a1%e5%9e%8b%ef%bc%9a2b-7b-%e5%8f%8c%e7%89%88%e6%9c%ac%e3%80%81%e5%9f%ba%e5%87%86%e6%b5%8b%e8%af%95%e8%b6%85%e8%b6%8a-meta-llama-2","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/4173.html","title":{"rendered":"Google DeepMind launches Gemma AI model: 2B\/7B dual versions, benchmarks surpass Meta Llama-2"},"content":{"rendered":"<p data-vmark=\"2b7c\">according to<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e8%b0%b7%e6%ad%8c\" title=\"[View articles tagged with [Google]]\" target=\"_blank\" >Google<\/a>Official press release, Google <a href=\"https:\/\/www.1ai.net\/en\/tag\/deepmind\" title=\"_Other Organiser\" target=\"_blank\" >DeepMind<\/a> The team launched a\u00a0<a href=\"https:\/\/www.1ai.net\/en\/tag\/gemma\" title=\"_Other Organiser\" target=\"_blank\" >Gemma<\/a>\u00a0<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e5%bc%80%e6%94%be%e6%a8%a1%e5%9e%8b\" title=\"_Other Organiser\" target=\"_blank\" >Open Model<\/a>, the AI model claims to be inspired by Gemini,<span class=\"accentTextColor\">Available in two sizes: Gemma 2B and Gemma 7B<\/span>.<\/p>\n<p data-vmark=\"aea2\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-4174\" title=\"8abf21d8-db15-48a0-b97b-7fcb5e339b40\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2024\/02\/8abf21d8-db15-48a0-b97b-7fcb5e339b40.jpg\" alt=\"8abf21d8-db15-48a0-b97b-7fcb5e339b40\" width=\"1600\" height=\"900\" \/><\/p>\n<p>\u25b2 Image source: Google official press release (the same below)<\/p>\n<p data-vmark=\"6b73\">IT Home noticed that this Gemma model can be fine-tuned using tools such as Keras 3.0 and PyTorch, and can run on low-performance devices such as laptops\/IoT, and is said to have stronger performance than competing products of the same size.<span class=\"accentTextColor\">Said to &quot;significantly outperform the Meta Llama-2 model&quot; on key benchmarks<\/span>.<\/p>\n<p data-vmark=\"4bed\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-4175\" title=\"b54945be-3d33-4115-b4ae-7f3fc19ffa7f\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2024\/02\/b54945be-3d33-4115-b4ae-7f3fc19ffa7f.jpg\" alt=\"b54945be-3d33-4115-b4ae-7f3fc19ffa7f\" width=\"1000\" height=\"749\" \/><\/p>\n<p data-vmark=\"2bdf\">The Google research team mentioned that this Gemma model is said to be built based on &quot;artificial intelligence principles&quot;. In order to make the Gemma model &quot;safe and reliable&quot;, the researchers used automated technology to filter out personal information and other sensitive data from the training set, and made a lot of fine-tuning to the model, and combined it with &quot;reinforcement learning (RLHF)&quot; to make the model behaviorally consistent. At the same time, the model was also subjected to &quot;red team testing&quot;, &quot;automatic adversarial testing&quot; and model capability evaluation, so that the model has &quot;responsible characteristics&quot;.<\/p>","protected":false},"excerpt":{"rendered":"<p>According to Google's official press release, the Google DeepMind team today launched a Gemma open model, which is an AI model that is said to be inspired by Gemini, and offers Gemma 2B and Gemma 7B scale versions to choose from. The Gemma model, which can be fine-tuned using tools such as Keras 3.0 and PyTorch, can be run on lower-performance devices such as laptops and IoT devices, and is said to have stronger performance compared to competitors of the same size, and is said to have \"significantly outperformed\" the Meta Llama-2 model in key benchmark tests. Meta Llama-2 model\" in key benchmarks. Google's research team mentions that the Gemma model<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[593,1289,1290,281],"collection":[],"class_list":["post-4173","post","type-post","status-publish","format-standard","hentry","category-news","tag-deepmind","tag-gemma","tag-1290","tag-281"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/4173","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=4173"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/4173\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=4173"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=4173"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=4173"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=4173"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}