{"id":25902,"date":"2025-05-19T09:34:49","date_gmt":"2025-05-19T01:34:49","guid":{"rendered":"https:\/\/aif.amtbbs.org\/?p=25902"},"modified":"2025-05-19T09:34:49","modified_gmt":"2025-05-19T01:34:49","slug":"%e5%a4%a7%e6%a8%a1%e5%9e%8b%e5%86%8d%e7%8e%b0%e9%bb%91%e9%a9%ac%ef%bc%81%e8%8b%b1%e4%bc%9f%e8%be%be%e5%bc%80%e6%ba%90llama-nemotron%e7%b3%bb%e5%88%97%e6%a8%a1%e5%9e%8b%ef%bc%8c%e6%95%88%e6%9e%9c","status":"publish","type":"post","link":"https:\/\/aif.amtbbs.org\/index.php\/2025\/05\/19\/25902\/","title":{"rendered":"\u5927\u6a21\u578b\u518d\u73b0\u9ed1\u9a6c\uff01\u82f1\u4f1f\u8fbe\u5f00\u6e90Llama-Nemotron\u7cfb\u5217\u6a21\u578b\uff0c\u6548\u679c\u4f18\u4e8eDeepSeek-R1"},"content":{"rendered":"
\"\"<\/div>\n
<\/div>\n
NVIDIA NeMo \u6846\u67b6\u662f\u4e00\u4e2a\u53ef\u6269\u5c55\u7684\u4e91\u539f\u751f\u751f\u6210\u5f0f AI \u6846\u67b6\uff0c\u4e13\u4e3a\u4ece\u4e8b\u5927\u578b\u8bed\u8a00\u6a21\u578b (LLM)\u3001\u591a\u6a21\u6001\u6a21\u578b (MM)\u3001\u81ea\u52a8\u8bed\u97f3\u8bc6\u522b (ASR)\u3001\u6587\u672c\u8f6c\u8bed\u97f3 (TTS ) \u548c\u8ba1\u7b97\u673a\u89c6\u89c9 (CV) \u9886\u57df\u7684\u7814\u7a76\u4eba\u5458\u548c PyTorch \u5f00\u53d1\u8005\u6253\u9020\u3002<\/div>\n
\n
\n

\u8fd1\u65e5\uff0c\u82f1\u4f1f\u8fbe\u63a8\u51fa\u4e86 Llama-Nemotron \u7cfb\u5217\u6a21\u578b\uff08\u57fa\u4e8e Meta AI \u7684 Llama \u6a21\u578b\u6784\u5efa\uff09\u2014\u2014 \u4e00\u4e2a\u9762\u5411\u9ad8\u6548\u63a8\u7406\u7684\u5927\u6a21\u578b\u5f00\u653e\u5bb6\u65cf\uff0c\u5177\u5907\u5353\u8d8a\u7684\u63a8\u7406\u80fd\u529b\u3001\u63a8\u7406\u6548\u7387\uff0c\u5e76\u91c7\u7528\u5bf9\u4f01\u4e1a\u53cb\u597d\u7684\u5f00\u653e\u8bb8\u53ef\u65b9\u5f0f\u3002<\/p>\n

\"\u56fe\u7247\"<\/p>\n

\u8be5\u7cfb\u5217\u5305\u62ec\u4e09\u4e2a\u6a21\u578b\u89c4\u6a21\uff1aNano\uff088B\uff09\u3001Super\uff0849B\uff09\u4e0e Ultra\uff08253B\uff09\uff0c\u53e6\u6709\u72ec\u7acb\u53d8\u4f53 UltraLong\uff088B\uff0c\u652f\u6301\u8d85\u957f\u4e0a\u4e0b\u6587\uff09\u3002\u8fd9\u4e9b\u6a21\u578b\u4e0d\u4ec5\u5177\u5907\u8d85\u5f3a\u7684\u63a8\u7406\u80fd\u529b\uff0c\u8fd8\u4e3a\u4f01\u4e1a\u4f7f\u7528\u63d0\u4f9b\u5f00\u653e\u8bb8\u53ef\u3002\u6a21\u578b\u6743\u91cd\u548c\u90e8\u5206\u8bad\u7ec3\u6570\u636e\u5728 Hugging Face \u4e0a\u516c\u5f00\uff0c\u9075\u5faa NVIDIA Open Model License \u548c Llama \u793e\u533a\u8bb8\u53ef\uff0c\u53ef\u5546\u4e1a\u4f7f\u7528\u3002<\/p>\n

\"\u56fe\u7247\"\"\u622a\u81f3<\/p>\n

\u76f8\u5173\u94fe\u63a5<\/h2>\n
    \n
  • \u8bba\u6587\uff1ahttps:\/\/arxiv.org\/pdf\/2505.00949<\/li>\n
  • \u4ee3\u7801\uff1ahttps:\/\/github.com\/NVIDIA\/NeMo<\/li>\n
  • \u6570\u636e\u96c6\uff1ahttps:\/\/huggingface.co\/datasets\/nvidia\/Llama-Nemotron-Post-Training-Dataset<\/li>\n
  • \u6a21\u578b\uff1ahttps:\/\/huggingface.co\/models?library=nemo&sort=downloads&search=nvidia<\/li>\n<\/ul>\n

    NeMo\u4ecb\u7ecd<\/h2>\n

    NVIDIA NeMo \u6846\u67b6\u662f\u4e00\u4e2a\u53ef\u6269\u5c55\u7684\u4e91\u539f\u751f\u751f\u6210\u5f0f AI \u6846\u67b6\uff0c\u4e13\u4e3a\u4ece\u4e8b\u5927\u578b\u8bed\u8a00\u6a21\u578b (LLM)\u3001\u591a\u6a21\u6001\u6a21\u578b (MM)\u3001\u81ea\u52a8\u8bed\u97f3\u8bc6\u522b (ASR)\u3001\u6587\u672c\u8f6c\u8bed\u97f3 (TTS) \u548c\u8ba1\u7b97\u673a\u89c6\u89c9 (CV) \u9886\u57df\u7684\u7814\u7a76\u4eba\u5458\u548c PyTorch \u5f00\u53d1\u8005\u6253\u9020\u3002\u5b83\u65e8\u5728\u5e2e\u52a9\u60a8\u5229\u7528\u73b0\u6709\u4ee3\u7801\u548c\u9884\u8bad\u7ec3\u7684\u6a21\u578b\u68c0\u67e5\u70b9\uff0c\u9ad8\u6548\u5730\u521b\u5efa\u3001\u5b9a\u5236\u548c\u90e8\u7f72\u65b0\u7684\u751f\u6210\u5f0f AI \u6a21\u578b\u3002<\/p>\n

    NeMo 2.0 \u7684\u65b0\u529f\u80fd<\/h2>\n
      \n
    • NVIDIA NeMo 2.0 \u4e0e\u5176\u524d\u8eab NeMo 1.0 \u76f8\u6bd4\u5f15\u5165\u4e86\u51e0\u9879\u91cd\u5927\u6539\u8fdb\uff0c\u589e\u5f3a\u4e86\u7075\u6d3b\u6027\u3001\u6027\u80fd\u548c\u53ef\u6269\u5c55\u6027\u3002<\/li>\n
    • \u57fa\u4e8e Python \u7684\u914d\u7f6e- NeMo 2.0 \u4ece YAML \u6587\u4ef6\u8f6c\u6362\u4e3a\u57fa\u4e8e Python \u7684\u914d\u7f6e\uff0c\u4ece\u800c\u63d0\u4f9b\u66f4\u9ad8\u7684\u7075\u6d3b\u6027\u548c\u63a7\u5236\u529b\u3002\u8fd9\u79cd\u8f6c\u53d8\u4f7f\u5f97\u4ee5\u7f16\u7a0b\u65b9\u5f0f\u6269\u5c55\u548c\u81ea\u5b9a\u4e49\u914d\u7f6e\u53d8\u5f97\u66f4\u52a0\u5bb9\u6613\u3002<\/li>\n
    • \u6a21\u5757\u5316\u62bd\u8c61\u2014\u2014\u901a\u8fc7\u91c7\u7528 PyTorch Lightning \u7684\u6a21\u5757\u5316\u62bd\u8c61\uff0cNeMo 2.0 \u7b80\u5316\u4e86\u9002\u914d\u548c\u5b9e\u9a8c\u3002\u8fd9\u79cd\u6a21\u5757\u5316\u65b9\u6cd5\u4f7f\u5f00\u53d1\u4eba\u5458\u80fd\u591f\u66f4\u8f7b\u677e\u5730\u4fee\u6539\u548c\u5b9e\u9a8c\u5176\u6a21\u578b\u7684\u4e0d\u540c\u7ec4\u4ef6\u3002<\/li>\n
    • \u53ef\u6269\u5c55\u6027\u2014\u2014NeMo 2.0 \u4f7f\u7528NeMo-Run\u5728\u6570\u5343\u4e2a GPU \u4e0a\u65e0\u7f1d\u6269\u5c55\u5927\u89c4\u6a21\u5b9e\u9a8c\uff0cNeMo-Run \u662f\u4e00\u79cd\u5f3a\u5927\u7684\u5de5\u5177\uff0c\u65e8\u5728\u7b80\u5316\u8de8\u8ba1\u7b97\u73af\u5883\u7684\u673a\u5668\u5b66\u4e60\u5b9e\u9a8c\u7684\u914d\u7f6e\u3001\u6267\u884c\u548c\u7ba1\u7406\u3002<\/li>\n<\/ul>\n

      \u603b\u7684\u6765\u8bf4\uff0c\u8fd9\u4e9b\u589e\u5f3a\u529f\u80fd\u4f7f NeMo 2.0 \u6210\u4e3a\u4e00\u4e2a\u5f3a\u5927\u3001\u53ef\u6269\u5c55\u4e14\u7528\u6237\u53cb\u597d\u7684 AI \u6a21\u578b\u5f00\u53d1\u6846\u67b6\u3002NeMo 2.0 \u76ee\u524d\u7531 LLM\uff08\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff09\u548c VLM\uff08\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff09\u96c6\u5408\u652f\u6301\u3002<\/p>\n

      \u5b9a\u6027\u8bc4\u4f30<\/h2>\n

      \"LN-UltraLN-Ultra \u5728\u5e7f\u6cdb\u7684\u63a8\u7406\u548c\u975e\u63a8\u7406\u57fa\u51c6\u6d4b\u8bd5\u4e2d\uff0c\u5728\u5f00\u653e\u6a21\u578b\u4e2d\u8868\u73b0\u51fa\u8272\u3002\"\u56fe\u7247\"<\/p>\n<\/div>\n

      \u6587\u7ae0\u6765\u81ea\uff1a51CTO<\/p>\n<\/div>\n

      <\/div>\n

      <\/g><\/svg><\/i> \"Loading\"<\/p>\n

      <\/div>\n","protected":false},"excerpt":{"rendered":"

      NVIDIA NeMo \u6846\u67b6\u662f\u4e00\u4e2a\u53ef\u6269\u5c55\u7684\u4e91\u539f\u751f\u751f\u6210\u5f0f AI \u6846\u67b6\uff0c\u4e13\u4e3a\u4ece\u4e8b\u5927\u578b\u8bed\u8a00\u6a21\u578b (LLM)\u3001\u591a\u6a21\u6001\u6a21 […]<\/p>\n

      <\/div>\n

      <\/g><\/svg><\/i> \"Loading\"<\/p>\n

      <\/div>\n","protected":false},"author":56,"featured_media":25904,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3,23,20,80],"tags":[124,1167,820],"class_list":["post-25902","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","category-23","category-20","category-80","tag-ai","tag-1167","tag-820"],"_links":{"self":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/25902","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/users\/56"}],"replies":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/comments?post=25902"}],"version-history":[{"count":1,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/25902\/revisions"}],"predecessor-version":[{"id":25905,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/25902\/revisions\/25905"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media\/25904"}],"wp:attachment":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media?parent=25902"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/categories?post=25902"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/tags?post=25902"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}