{"id":1163,"date":"2021-12-11T15:09:33","date_gmt":"2021-12-11T06:09:33","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=1163"},"modified":"2021-12-11T15:09:33","modified_gmt":"2021-12-11T06:09:33","slug":"t-nlrv5-ss-moe-gopher-glam-microsoft-google-deepmind%e3%81%ae%e5%b7%a8%e5%a4%a7%e8%a8%80%e8%aa%9e%e3%83%a2%e3%83%87%e3%83%ab","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=1163","title":{"rendered":"T-NLRv5, SS-MoE, Gopher, GLaM: Microsoft, Google, DeepMind\u306e\u5de8\u5927\u8a00\u8a9e\u30e2\u30c7\u30eb"},"content":{"rendered":"\n<p>\u3053\u3053\u6700\u8fd1\u306f\u5de8\u5927\u8a00\u8a9e\u30e2\u30c7\u30eb\u306e\u8a71\u984c\u304c\u591a\u304b\u3063\u305f\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>12\/3 Microsoft: <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/blog\/efficiently-and-effectively-scaling-up-language-model-pretraining-for-best-language-representation-model-on-glue-and-superglue\/\">Efficiently and effectively scaling up language model pretraining for best language representation model on GLUE and SuperGLUE &#8211; Microsoft Research<\/a><ul><li>Microsoft\u30c1\u30e5\u30fc\u30ea\u30f3\u30b0\u30e2\u30c7\u30eb\uff08T-NLRv5\uff09\u304c\u3001SuperGLUE\u3068GLUE\u3067SoTA\u3092\u5831\u544a\u3002<\/li><li>MNLI\u3001RTE\u3067\u521d\u3081\u3066\u4eba\u9593\u306e\u30b9\u30b3\u30a2\u3092\u8d85\u3048\u305f\u3068\u306e\u3053\u3068\u3002<\/li><\/ul><\/li><\/ul>\n\n\n\n<ul class=\"wp-block-list\"><li>12\/8 Google: SS-MoE\uff08<a href=\"https:\/\/arxiv.org\/abs\/2101.03961\">[2101.03961] Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity (arxiv.org)<\/a>\u30d9\u30fc\u30b9\u3001Mixture-of-Expert\u3067269B \u30d1\u30e9\u30e1\u30fc\u30bf\uff09\u304cSuperGLUE\u306eSoTA\u3092\u66f4\u65b0<\/li><\/ul>\n\n\n\n<ul class=\"wp-block-list\"><li>12\/8 DeepMind: <a href=\"https:\/\/storage.googleapis.com\/deepmind-media\/research\/language-research\/Training%20Gopher.pdf\">Scaling Language Models: Methods, Analysis &amp; Insights from Training Gopher<\/a><ul><li>DeepMind\u304c280B\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u5de8\u5927\u8a00\u8a9e\u30e2\u30c7\u30ebGopher\u3092\u767a\u8868<\/li><li>\u4f7f\u7528\u3057\u305f\u30c7\u30fc\u30bf\u306fMassiveText\u3068\u547c\u3070\u308c\u308b235\u5104\u6587\u66f8\u300110.5TB\u306e\u82f1\u8a9e\u306e\u30c6\u30ad\u30b9\u30c8\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3002WEB\u30da\u30fc\u30b8\u3001\u30cb\u30e5\u30fc\u30b9\u8a18\u4e8b\u3001github\u306e\u30b3\u30fc\u30c9\u306a\u3069\u69d8\u3005\u306a\u30c9\u30e1\u30a4\u30f3\u3092\u542b\u3080\u3002<\/li><li>\u4ed6\u306e\u30e2\u30c7\u30eb\u3068\u6bd4\u8f03\u53ef\u80fd\u306a124\u30bf\u30b9\u30af\u4e2d100\u30bf\u30b9\u30af\u3067SoTA<\/li><\/ul><\/li><\/ul>\n\n\n\n<ul class=\"wp-block-list\"><li>12\/9 Google: <a href=\"https:\/\/ai.googleblog.com\/2021\/12\/more-efficient-in-context-learning-with.html\">Google AI Blog: More Efficient In-Context Learning with GLaM (googleblog.com)<\/a><ul><li>Google\u304c1.2T\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u30b9\u30d1\u30fc\u30b9\u306a\u8a00\u8a9e\u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3002GLaM\u306fGeneralist Language Model\u306e\u7565\u3002<\/li><li>\u4f7f\u7528\u30c7\u30fc\u30bf\u306f1.6\u5146\u30c8\u30fc\u30af\u30f3\u3001WEB\u30da\u30fc\u30b8\u306e\u30c7\u30fc\u30bf\u3092\u66f8\u7c4d\u30fbWikipedia\uff08\u9ad8\u54c1\u8cea\u30c7\u30fc\u30bf\uff09\u3067\u4f5c\u6210\u3057\u305f\u30d5\u30a3\u30eb\u30bf\u3092\u7528\u3044\u3066\u9ad8\u54c1\u8cea\u5316\u3002<\/li><li>8\u3064\u306eNLG\u30bf\u30b9\u30af\u300121\u306eNLU\u30bf\u30b9\u30af\u3067\u8a55\u4fa1\u3001dense\u306a\u30e2\u30c7\u30eb\u3067\u3042\u308bGPT-3\u3068\u6bd4\u3079\u3066\u30bc\u30ed\u30b7\u30e7\u30c3\u30c8\u3067\u306f80%\u3001\u30ef\u30f3\u30b7\u30e7\u30c3\u30c8\u3067\u306f90%\u3067\u540c\u7b49\u4ee5\u4e0a\u7d50\u679c\u3068\u5831\u544a\u3002<ul><li>NLG: Natural Language Generation \/ \u30c6\u30ad\u30b9\u30c8\u751f\u6210<\/li><li>NLU: Natural Language Understanding&nbsp; \/ \u30c6\u30ad\u30b9\u30c8\u7406\u89e3<\/li><li>\u30bc\u30ed\u30b7\u30e7\u30c3\u30c8\u3001\u30ef\u30f3\u30b7\u30e7\u30c3\u30c8\u3068\u3082\u306b24\/29\u3067\u540c\u7b49\u4ee5\u4e0a\u306b\u898b\u3048\u308b\u304c\u7b97\u51fa\u65b9\u6cd5\u304c\u5206\u304b\u3089\u306a\u3044\u30fb\u30fb\u30fb<\/li><\/ul><\/li><li>\u758e\u306a\u30e2\u30c7\u30eb\u3068\u5bc6\u306a\u30e2\u30c7\u30eb\u3092\u6bd4\u8f03\u3059\u308b\u3068\u3001\u758e\u306a\u30e2\u30c7\u30eb\u306e\u65b9\u304c\u5c11\u306a\u3044\u30c7\u30fc\u30bf\u3067\u512a\u308c\u305f\u6027\u80fd\u3092\u767a\u63ee\u3067\u304d\u308b\u3068\u306e\u3053\u3068\u3002\u6700\u5f8c\u306bGlaM\u306f\u96fb\u529b\u52b9\u7387\u3082\u512a\u308c\u3066\u3044\u308b\u3068\u4e3b\u5f35\u3002<\/li><\/ul><\/li><\/ul>\n\n\n\n<p>\u3053\u306e\u5148\u306bAGI\u304c\u3042\u3063\u3066\u3082\u4e0d\u601d\u8b70\u3067\u306f\u306a\u3044\u611f\u899a\u304c\u3042\u308b\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u3053\u3053\u6700\u8fd1\u306f\u5de8\u5927\u8a00\u8a9e\u30e2\u30c7\u30eb\u306e\u8a71\u984c\u304c\u591a\u304b\u3063\u305f\u3002 12\/3 Microsoft: Efficiently and effectively scaling up language model pretraining for be &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=1163\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;T-NLRv5, SS-MoE, Gopher, GLaM: Microsoft, Google, DeepMind\u306e\u5de8\u5927\u8a00\u8a9e\u30e2\u30c7\u30eb&#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":[170,567,634],"class_list":["post-1163","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-gigantic-language-model","tag-567","tag-634"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/1163","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=1163"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/1163\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1163"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1163"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1163"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}