{"id":736,"date":"2021-09-27T05:30:00","date_gmt":"2021-09-26T20:30:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=736"},"modified":"2021-09-27T05:30:00","modified_gmt":"2021-09-26T20:30:00","slug":"%e3%83%9e%e3%83%ab%e3%83%81%e3%82%bf%e3%82%b9%e3%82%af%e3%83%bb%e3%83%9e%e3%83%ab%e3%83%81%e3%83%aa%e3%83%b3%e3%82%ac%e3%83%ab%e3%81%ae%e3%81%9f%e3%82%81%e3%81%ae%e5%a4%a7%e8%a6%8f%e6%a8%a1%e3%83%bb","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=736","title":{"rendered":"\u30de\u30eb\u30c1\u30bf\u30b9\u30af\u30fb\u30de\u30eb\u30c1\u30ea\u30f3\u30ac\u30eb\u306e\u305f\u3081\u306e\u5927\u898f\u6a21\u30fb\u52b9\u7387\u7684\u306aMoE(Mixture of Experts)\u5b66\u7fd2"},"content":{"rendered":"\n<ul class=\"wp-block-list\"><li><strong>Scalable and Efficient MoE Training for Multitask Multilingual Models\u00a0<\/strong>[56.0]<br>\u6211\u3005\u306f,MoE\u30e2\u30c7\u30eb\u3092\u6570\u5146\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u306b\u52b9\u7387\u7684\u306b\u30b9\u30b1\u30fc\u30eb\u3067\u304d\u308b\u30b7\u30b9\u30c6\u30e0\u3092\u958b\u767a\u3057\u305f\u3002 \u307e\u305f,MoE\u30b5\u30f3\u30d7\u30eb\u306e\u52b9\u7387\u3092\u5411\u4e0a\u3055\u305b\u308b\u305f\u3081\u306e\u65b0\u305f\u306a\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u624b\u6cd5\u3092\u63d0\u6848\u3057,\u6642\u9593\u52b9\u7387\u3092\u5411\u4e0a\u3055\u305b\u308b\u305f\u3081\u306b\u5c02\u9580\u5bb6\u306e\u5208\u308a\u53d6\u308a\u6226\u7565\u3092\u6d3b\u7528\u3059\u308b\u3002 50\u8a00\u8a9e\u3067100\u5104\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u3067\u8a13\u7df4\u3055\u308c\u305f\u30e2\u30c7\u30eb\u306f\u3001\u6a5f\u68b0\u7ffb\u8a33(MT)\u304a\u3088\u3073\u591a\u8a00\u8a9e\u81ea\u7136\u8a00\u8a9e\u751f\u6210\u30bf\u30b9\u30af\u306b\u304a\u3051\u308b\u6700\u5148\u7aef\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u9054\u6210\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2109.10465v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2109.10465v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 22 Sep 2021 00:57:46 GMT)<ul><li>\u6975\u3081\u3066\u5927\u898f\u6a21\u306a\u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u53ef\u80fd\u306a\u624b\u6cd5\u306b\u95a2\u3059\u308b\u8ad6\u6587\u3002\u65e2\u5b58\u624b\u6cd5\u306b\u6bd4\u3079\u3066\u540c\u3058\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u30678\u500d\u306e\u30e2\u30c7\u30eb\u30b5\u30a4\u30ba\u306e\u5b66\u7fd2\u3092\u5b9f\u73fe\u3002<\/li><li>\u69cb\u7bc9\u3055\u308c\u305f\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2\u52b9\u7387\u3068\u6027\u80fd\u304c\u51c4\u3044\u3002dense\u306a\u30e2\u30c7\u30eb\u306b\u6bd4\u3079\u306610\u500d\u53ce\u675f\u304c\u901f\u3044\u3002\u5358\u7d14\u306a\u500b\u5225\u306e\u30d0\u30a4\u30ea\u30f3\u30ac\u30eb\u30e2\u30c7\u30eb\u306b\u304f\u3089\u3079\u3066\u30de\u30eb\u30c1\u30ea\u30f3\u30ac\u30eb\u8a2d\u5b9a\u3067BLEU +4pt\u3001M2M-100\u306b\u6bd4\u3079\u3066BLEU + 3pt\u3002\u306a\u3069\u3059\u3054\u3044\u6570\u5024\u304c\u4e26\u3093\u3067\u3044\u308b\u3002<\/li><li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/microsoft\/DeepSpeed\">https:\/\/github.com\/microsoft\/DeepSpeed<\/a>\u3001\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u3078\u306e\u30ea\u30f3\u30af\u3082\u3042\u3063\u3066\u975e\u5e38\u306b\u53c2\u8003\u306b\u306a\u308b\u3002<\/li><\/ul><\/li><\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Scalable and Efficient MoE Training for Multitask Multilingual Models\u00a0[56.0]\u6211\u3005\u306f,MoE\u30e2\u30c7\u30eb\u3092\u6570\u5146\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u306b\u52b9\u7387\u7684\u306b\u30b9\u30b1\u30fc\u30eb\u3067\u304d\u308b\u30b7\u30b9\u30c6 &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=736\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;\u30de\u30eb\u30c1\u30bf\u30b9\u30af\u30fb\u30de\u30eb\u30c1\u30ea\u30f3\u30ac\u30eb\u306e\u305f\u3081\u306e\u5927\u898f\u6a21\u30fb\u52b9\u7387\u7684\u306aMoE(Mixture of Experts)\u5b66\u7fd2&#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":[250,567,602,634],"class_list":["post-736","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-mixture-of-experts","tag-567","tag-602","tag-634"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/736","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=736"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/736\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=736"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=736"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=736"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}