{"id":4421,"date":"2024-02-13T04:59:00","date_gmt":"2024-02-12T19:59:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4421"},"modified":"2024-02-13T04:59:00","modified_gmt":"2024-02-12T19:59:00","slug":"gssms-vs-transformer%e3%81%a8black-mamba","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4421","title":{"rendered":"GSSMs vs transformer\u3068Black Mamba"},"content":{"rendered":"\n<p>GSSM\uff08Generalized State Space Models\uff09\u3068transformer\u306e\u6bd4\u8f03\u3068MoE\u306a\u30a2\u30d7\u30ed\u30fc\u30c1\u3002\u6628\u65e5\u306e<a href=\"https:\/\/devneko.jp\/wordpress\/?p=4399\">Mamba\u306eICL\uff08In Context Learning\uff09\u6027\u80fd \u2013 arXiv\u6700\u65b0\u8ad6\u6587\u306e\u7d39\u4ecb (devneko.jp)<\/a>\u306e\u901a\u308a\u3001\u7279\u6027\u306f\u304b\u306a\u308a\u9055\u3046\u306e\u3067MoE\u3063\u307d\u304f\u4f7f\u3046\u306e\u306f\u3042\u308a\u306a\u306e\u304b\u3082\u3057\u308c\u306a\u3044\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Repeat After Me: Transformers are Better than State Space Models at Copying\u00a0<\/strong>[57.4]<br>\u4e00\u822c\u5316\u3055\u308c\u305f\u72b6\u614b\u7a7a\u9593\u30e2\u30c7\u30eb\u306f\u3001\u63a8\u8ad6\u6642\u9593\u52b9\u7387\u306e\u89b3\u70b9\u304b\u3089\u306f\u6709\u671b\u3067\u3042\u308b\u304c\u3001\u5165\u529b\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u304b\u3089\u306e\u30b3\u30d4\u30fc\u3092\u5fc5\u8981\u3068\u3059\u308b\u30bf\u30b9\u30af\u306e\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u30e2\u30c7\u30eb\u3068\u6bd4\u8f03\u3057\u3066\u9650\u5b9a\u7684\u3067\u3042\u308b\u3053\u3068\u3092\u793a\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2402.01032v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2402.01032v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 1 Feb 2024 21:44:11 GMT)<\/li>\n\n\n\n<li>\u30b7\u30f3\u30d7\u30eb\u306a\u4e8b\u4f8b\u3067\u306eGSSM\u3068transformer\u306e\u6bd4\u8f03\u3002\u5f53\u7136\u306a\u306e\u304b\u3082\u3060\u304c\u300ctransformer models dramatically outperform state space models at copying and retrieving information from context.\u300d<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>BlackMamba: Mixture of Experts for State-Space Models\u00a0<\/strong>[10.2]<br>\u72b6\u614b\u7a7a\u9593\u30e2\u30c7\u30eb(SSM)\u306f\u3001\u6700\u8fd1\u3001\u5927\u898f\u6a21\u306a\u8a00\u8a9e\u30e2\u30c7\u30ea\u30f3\u30b0\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u3068\u7af6\u5408\u3059\u308b\u6027\u80fd\u3092\u793a\u3057\u305f\u3002 MoE\u30e2\u30c7\u30eb\u306f\u3001\u8a08\u7b97\u30b3\u30b9\u30c8\u3068\u9045\u5ef6\u30b3\u30b9\u30c8\u3092\u5927\u5e45\u306b\u524a\u6e1b\u3057\u306a\u304c\u3089\u3001\u9855\u8457\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u793a\u3057\u3066\u3044\u308b\u3002 \u6211\u3005\u306f,Mamba SSM\u3068MoE\u3092\u7d44\u307f\u5408\u308f\u305b\u305f\u65b0\u3057\u3044\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3067\u3042\u308bBlackMamba\u3092\u7d39\u4ecb\u3057\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2402.01771v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2402.01771v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 1 Feb 2024 07:15:58 GMT)<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/Zyphra\/BlackMamba\">Zyphra\/BlackMamba: Code repository for Black Mamba (github.com)<\/a>\u3001\u30e2\u30c7\u30eb\u3082\u516c\u958b\u3055\u308c\u3066\u3044\u308b\u3000<a href=\"https:\/\/huggingface.co\/Zyphra\/BlackMamba-2.8B\">Zyphra\/BlackMamba-2.8B \u00b7 Hugging Face<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>GSSM\uff08Generalized State Space Models\uff09\u3068transformer\u306e\u6bd4\u8f03\u3068MoE\u306a\u30a2\u30d7\u30ed\u30fc\u30c1\u3002\u6628\u65e5\u306eMamba\u306eICL\uff08In Context Learning\uff09\u6027\u80fd \u2013 arXiv\u6700\u65b0\u8ad6\u6587\u306e &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=4421\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;GSSMs vs transformer\u3068Black Mamba&#8221; \u306e<\/span>\u7d9a\u304d\u3092\u8aad\u3080<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[235,415],"class_list":["post-4421","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-mamba","tag-transformer"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4421","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4421"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4421\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4421"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4421"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4421"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}