{"id":764,"date":"2024-02-27T15:50:00","date_gmt":"2024-02-27T07:50:00","guid":{"rendered":"https:\/\/www.tchepai.com\/?p=764"},"modified":"2025-02-09T18:32:59","modified_gmt":"2025-02-09T10:32:59","slug":"tokenization","status":"publish","type":"post","link":"https:\/\/www.kdjingpai.com\/de\/tokenization\/","title":{"rendered":"Tokenization(\u5206\u8bcd\u6807\u8bb0\u5316)"},"content":{"rendered":"<p>\u5404\u4f4d\u597d\uff0c\u4eca\u5929\u6211\u4eec\u8981\u63a2\u8ba8\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08LLM\uff09\u4e2d\u7684\u5206\u8bcd\u6280\u672f\u3002\u9057\u61be\u7684\u662f\uff0c\u5206\u8bcd\u5728\u5f53\u524d\u9876\u5c16\u7684LLM\u4e2d\u662f\u4e00\u4e2a\u6bd4\u8f83\u590d\u6742\u4e14\u68d8\u624b\u7684\u90e8\u5206\uff0c\u4f46\u7406\u89e3\u5b83\u7684\u67d0\u4e9b\u7ec6\u8282\u975e\u5e38\u5fc5\u8981\uff0c\u56e0\u4e3a\u5f88\u591a\u4eba\u5c06LLM\u7684\u4e00\u4e9b\u7f3a\u70b9\u5f52\u548e\u4e8e\u795e\u7ecf\u7f51\u7edc\u6216\u5176\u4ed6\u663e\u5f97\u795e\u79d8\u7684\u539f\u56e0\uff0c\u5b9e\u9645\u4e0a\u8fd9\u4e9b\u95ee\u9898\u7684\u6e90\u5934\u5f80\u5f80\u662f\u5206\u8bcd\u6280\u672f\u3002<br \/>\n\u539f\u6587\uff1a<a href=\"https:\/\/github.com\/karpathy\/minbpe\">https:\/\/github.com\/karpathy\/minbpe<\/a><\/p>\n<h3>\u4e4b\u524d\uff1a\u57fa\u4e8e\u5b57\u7b26\u7684\u5206\u8bcd<\/h3>\n<p>\u90a3\u4e48\u4ec0\u4e48\u662f\u5206\u8bcd\u5462\uff1f\u5176\u5b9e\u5728\u6211\u4eec\u4e4b\u524d\u7684\u89c6\u9891[<a href=\"https:\/\/www.youtube.com\/watch?v=kCc8FmEb1nY\">\u4ece\u96f6\u5f00\u59cb\u6784\u5efaGPT<\/a>]\u4e2d\uff0c\u6211\u4eec\u5df2\u7ecf\u8bb2\u8fc7\u5206\u8bcd\u4e86\uff0c\u4f46\u90a3\u53ea\u662f\u4e00\u79cd\u975e\u5e38\u7b80\u5355\u4e14\u539f\u59cb\u7684\u5b57\u7b26\u7ea7\u522b\u5f62\u5f0f\u3002\u5982\u679c\u4f60\u53bb\u770b\u90a3\u4e2a\u89c6\u9891\u7684[<a href=\"https:\/\/colab.research.google.com\/drive\/1JMLa53HDuA-i7ZBmqV7ZnA3c_fvtXnx-?usp=sharing\">Google colab<\/a>]\u9875\u9762\uff0c\u4f60\u4f1a\u53d1\u73b0\u6211\u4eec\u4f7f\u7528\u7684\u662f\u6211\u4eec\u7684\u8bad\u7ec3\u6570\u636e\uff08[<a href=\"https:\/\/raw.githubusercontent.com\/karpathy\/char-rnn\/master\/data\/tinyshakespeare\/input.txt\">\u838e\u58eb\u6bd4\u4e9a\u4f5c\u54c1<\/a>]\uff0c\u5b83\u5728Python\u4e2d\u53ea\u662f\u4e00\u4e2a\u5f88\u957f\u7684\u5b57\u7b26\u4e32\uff1a<\/p>\n<blockquote><p>First Citizen: \u5728\u6211\u4eec\u7ee7\u7eed\u4e4b\u524d\uff0c\u8bf7\u542c\u6211\u8bf4\u3002<br \/>\nAll: \u8bf4\u5427\uff0c\u8bf4\u5427\u3002<br \/>\nFirst Citizen: \u4f60\u4eec\u90fd\u51b3\u5fc3\u5b81\u613f\u6b7b\u4e5f\u4e0d\u613f\u997f\u6b7b\u5417\uff1f<br \/>\nAll: \u51b3\u5fc3\u7684\u3002\u51b3\u5fc3\u7684\u3002<br \/>\nFirst Citizen: \u9996\u5148\uff0c\u4f60\u4eec\u77e5\u9053Gaius Marcius\u662f\u4eba\u6c11\u7684\u5934\u53f7\u654c\u4eba\u3002<br \/>\nAll: \u6211\u4eec\u77e5\u9053\uff0c\u6211\u4eec\u77e5\u9053\u3002<\/p><\/blockquote>\n<p>\u4f46\u662f\uff0c\u6211\u4eec\u5982\u4f55\u628a\u5b57\u7b26\u5e8f\u5217\u8f93\u5165\u5230\u8bed\u8a00\u6a21\u578b\u4e2d\u53bb\u5462\uff1f\u6211\u4eec\u8fd9\u6837\u505a\u7684\u7b2c\u4e00\u6b65\u662f\u6784\u9020\u4e00\u4e2a\u8bcd\u6c47\u8868\uff0c\u8fd9\u4e2a\u8868\u5305\u62ec\u4e86\u6211\u4eec\u5728\u6574\u4e2a\u8bad\u7ec3\u6570\u636e\u96c6\u4e2d\u627e\u5230\u7684\u6240\u6709\u4e0d\u540c\u5b57\u7b26\uff1a<\/p>\n<blockquote><p>&#8220;`python# \u8fd9\u662f\u6587\u672c\u4e2d\u51fa\u73b0\u7684\u552f\u4e00\u5b57\u7b26chars = sorted(list(set(text)))vocab_size = len(chars)print(&#8221;.join(chars))print(vocab_size)<br \/>\n# !$&amp;&#8217;,-.3:;?ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz# 65&#8220;`<\/p><\/blockquote>\n<p>\u7d27\u63a5\u7740\uff0c\u6211\u4eec\u4f1a\u521b\u5efa\u4e00\u4e2a\u67e5\u627e\u8868\uff0c\u7528\u4e8e\u5c06\u5355\u4e2a\u5b57\u7b26\u548c\u6574\u6570\u5b57\u4e4b\u95f4\u8fdb\u884c\u8f6c\u6362\uff0c\u8fd9\u4e2a\u67e5\u627e\u8868\u57fa\u4e8e\u4e0a\u9762\u5217\u51fa\u7684\u8bcd\u6c47\u3002\u8fd9\u4e2a\u67e5\u627e\u8868\u5b9e\u9645\u4e0a\u662f\u4e00\u4e2aPython\u5b57\u5178\uff1a<\/p>\n<blockquote><p>&#8220;`pythonstoi = { ch:i for i,ch in enumerate(chars) }itos = { i:ch for i,ch in enumerate(chars) }# \u7f16\u7801\u5668\uff1a\u63a5\u6536\u4e00\u4e2a\u5b57\u7b26\u4e32\uff0c\u5e76\u8f93\u51fa\u5bf9\u5e94\u7684\u6574\u6570\u5217\u8868encode = lambda s: [stoi[c] for c in s]# \u89e3\u7801\u5668\uff1a\u63a5\u6536\u4e00\u4e2a\u6574\u6570\u5217\u8868\uff0c\u5e76\u8fd8\u539f\u56de\u5b57\u7b26\u4e32decode = lambda l: &#8221;.join([itos[i] for i in l])<br \/>\nprint(encode(&#8220;hii there&#8221;))print(decode(encode(&#8220;hii there&#8221;)))<br \/>\n# [46, 47, 47, 1, 58, 46, 43, 56, 43]# hii there&#8220;`<\/p><\/blockquote>\n<p>\u6211\u4eec\u5c06\u5e8f\u5217\u8f6c\u6362\u4e3a\u6574\u6570\u4e4b\u540e\uff0c\u4f1a\u770b\u5230\u6bcf\u4e2a\u6574\u6570\u90fd\u4f5c\u4e3a\u7d22\u5f15\u5728\u4e00\u4e2a\u4e8c\u7ef4\u7684\u53ef\u8bad\u7ec3\u5d4c\u5165\u77e9\u9635\u4e2d\u4f7f\u7528\u3002\u7531\u4e8e\u8bcd\u6c47\u8868\u7684\u5927\u5c0f\u4e3a`vocab_size=65`\uff0c\u56e0\u6b64\u8fd9\u4e2a\u5d4c\u5165\u77e9\u9635\u4e5f\u670965\u884c\uff1a<\/p>\n<blockquote><p>&#8220;`pythonclass BigramLanguageModel(nn.Module):<br \/>\ndef __init__(self, vocab_size):super().__init__()self.token_embedding_table = nn.Embedding(vocab_size, n_embd)<br \/>\ndef forward(self, idx, targets=None):tok_emb = self.token_embedding_table(idx) # (B,T,C)&#8220;`<\/p><\/blockquote>\n<p>\u5728\u8fd9\u91cc\uff0c\u201cinteger\u201d\u63d0\u53d6\u4e86\u5d4c\u5165\u8868\u4e2d\u7684\u4e00\u884c\uff0c\u800c\u8fd9\u884c\u5c31\u662f\u4ee3\u8868\u8be5\u6807\u8bb0(token)\u7684\u5411\u91cf\u3002\u63a5\u7740\uff0c\u8fd9\u4e2a\u5411\u91cf\u4f5c\u4e3a\u76f8\u5e94\u65f6\u95f4\u6b65\u7684\u8f93\u5165\u8fdb\u5165\u53d8\u6362\u5668(<a href=\"https:\/\/www.kdjingpai.com\/transformer\/\">Transformer<\/a>)\u3002<\/p>\n<h3>\u4f7f\u7528\u5b57\u8282\u5bf9\u7f16\u7801(BPE)\u7b97\u6cd5\u7684&#8221;\u5b57\u7b26\u5757&#8221;\u8fdb\u884c\u6807\u8bb0\u5316<\/h3>\n<p>\u8fd9\u5bf9\u4e8e\u4e00\u4e2a\u521d\u7ea7\u7684\u5b57\u7b26\u7ea7\u522b\u8bed\u8a00\u6a21\u578b\u6765\u8bf4\u5df2\u7ecf\u8db3\u591f\u597d\u4e86\u3002\u4f46\u5728\u5b9e\u8df5\u4e2d\uff0c\u6700\u5148\u8fdb\u7684\u8bed\u8a00\u6a21\u578b\u4f1a\u91c7\u7528\u66f4\u590d\u6742\u7684\u65b9\u6848\u6784\u5efa\u6807\u8bb0\u8bcd\u6c47\u8868\u3002\u8fd9\u4e9b\u65b9\u6848\u5e76\u4e0d\u662f\u5728\u5b57\u7b26\u7ea7\u522b\u4e0a\u5de5\u4f5c\uff0c\u800c\u662f\u5728\u5b57\u7b26\u5757\u7ea7\u522b\u4e0a\u3002\u8fd9\u4e9b\u5b57\u7b26\u5757\u8bcd\u6c47\u8868\u662f\u901a\u8fc7\u4f7f\u7528\u50cf\u5b57\u8282\u5bf9\u7f16\u7801\uff08BPE\uff09\u8fd9\u7c7b\u7b97\u6cd5\u6784\u5efa\u7684\uff0c\u6211\u4eec\u5c06\u5728\u4e0b\u6587\u8be6\u7ec6\u89e3\u91ca\u8fd9\u79cd\u65b9\u6cd5\u3002<br \/>\n\u8f6c\u5411\u8fd9\u79cd\u65b9\u6cd5\u7684\u5386\u53f2\u53d1\u5c55\uff0cOpenAI\u57282019\u5e74\u53d1\u8868\u7684GPT-2\u8bba\u6587\uff0c\u6807\u9898\u4e3a\u300a\u8bed\u8a00\u6a21\u578b\u662f\u65e0\u76d1\u7763\u591a\u4efb\u52a1\u5b66\u4e60\u8005\u300b\uff0c\u4f7f\u5f97\u5b57\u8282\u7ea7BPE\u7b97\u6cd5\u5728\u8bed\u8a00\u6a21\u578b\u7684\u6807\u8bb0\u5316\u4e2d\u6d41\u884c\u8d77\u6765\u3002\u9605\u8bfb\u672c\u65872.2\u8282\u5173\u4e8e\u201c\u8f93\u5165\u8868\u5f81\u201d\uff0c\u4e86\u89e3\u4ed6\u4eec\u5bf9\u8fd9\u4e00\u7b97\u6cd5\u7684\u63cf\u8ff0\u548c\u52a8\u673a\u3002\u5728\u672c\u8282\u672b\u5c3e\uff0c\u4f60\u5c06\u770b\u5230\u5b83\u4eec\u63d0\u5230\uff1a<\/p>\n<blockquote><p>\u8bcd\u6c47\u91cf\u5df2\u7ecf\u6269\u5c55\u5230\u4e8650,257\u3002\u6211\u4eec\u8fd8\u5c06\u4e0a\u4e0b\u6587\u5927\u5c0f\u4ece512\u589e\u52a0\u5230\u4e861024\u4e2a\u6807\u8bb0\uff0c\u5e76\u4e14\u4f7f\u7528\u4e86\u66f4\u5927\u7684\u6279\u6b21\u5927\u5c0f512\u4e2a\u3002<\/p><\/blockquote>\n<p>\u56de\u60f3\u4e00\u4e0b\u5728\u53d8\u6362\u5668(Transformer)\u7684\u5173\u6ce8(attention)\u5c42\uff0c\u6bcf\u4e00\u4e2a\u6807\u8bb0(token)\u90fd\u662f\u5bf9\u4e4b\u524d\u5e8f\u5217\u4e2d\u7684\u6709\u9650\u6570\u91cf\u6807\u8bb0\u8fdb\u884c\u5173\u6ce8\u3002\u8bba\u6587\u63d0\u5230\uff0cGPT-2\u6a21\u578b\u7684\u4e0a\u4e0b\u6587\u957f\u5ea6\u662f1024\u4e2a\u6807\u8bb0\uff0c\u6bd4GPT-1\u7684512\u4e2a\u6709\u6240\u63d0\u9ad8\u3002\u6362\u53e5\u8bdd\u8bf4\uff0c\u6807\u8bb0\u662f\u4f5c\u4e3a\u8bed\u8a00\u957f\u671f\u8bb0\u5fc6\u6a21\u578b(LLM)\u8f93\u5165\u7684\u57fa\u672c\u5355\u5143\u3002\u6807\u8bb0\u5316\u662f\u5c06Python\u4e2d\u7684\u539f\u59cb\u5b57\u7b26\u4e32\u8f6c\u6362\u4e3a\u6807\u8bb0\u5217\u8868\u7684\u8fc7\u7a0b\uff0c\u53cd\u4e4b\u4ea6\u7136\u3002\u4e3e\u4e00\u4e2a\u5176\u4ed6\u7684\u6d41\u884c\u4f8b\u5b50\u6765\u5c55\u793a\u8fd9\u79cd\u62bd\u8c61\u6982\u5ff5\u7684\u65e0\u5904\u4e0d\u5728\uff0c\u4f60\u5982\u679c\u67e5\u770b[<a href=\"https:\/\/arxiv.org\/abs\/2307.09288\">Llama 2<\/a>]\u7684\u8bba\u6587\uff0c\u540c\u6837\u641c\u7d22\u201c\u6807\u8bb0\u201d\uff0c\u4f60\u4f1a\u5f97\u523063\u5904\u63d0\u53ca\u3002\u5c31\u6bd4\u5982\u8bf4\uff0c\u8bba\u6587\u58f0\u79f0\u4ed6\u4eec\u5728\u4e24\u4e07\u4ebf\u4e2a\u6807\u8bb0\u4e0a\u8bad\u7ec3\u7b49\u7b49\u3002<br \/>\n\u5728\u6211\u4eec\u6df1\u5165\u63a2\u8ba8\u5b9e\u73b0\u7ec6\u8282\u4e4b\u524d\uff0c\u5148\u7b80\u8981\u63a2\u8ba8\u4e00\u4e0b\u4e3a\u4ec0\u4e48\u6211\u4eec\u9700\u8981\u6df1\u5165\u4e86\u89e3\u4ee4\u724c\u5316\u8fc7\u7a0b\u3002\u4ee4\u724c\u5316\u662fLLMs\u4e2d\u8bb8\u591a\u5947\u5f02\u73b0\u8c61\u7684\u6838\u5fc3\uff0c\u6211\u5efa\u8bae\u5927\u5bb6\u4e0d\u8981\u8f7b\u89c6\u5b83\u3002\u5f88\u591a\u770b\u4f3c\u4e0e\u795e\u7ecf\u7f51\u7edc\u7ed3\u6784\u6709\u5173\u7684\u95ee\u9898\uff0c\u5b9e\u9645\u4e0a\u90fd\u6e90\u81ea\u4e8e\u4ee4\u724c\u5316\u3002\u8fd9\u91cc\u5217\u4e3e\u51e0\u4e2a\u4f8b\u5b50\uff1a<\/p>\n<ul>\n<li>\u4e3a\u4ec0\u4e48LLM\u65e0\u6cd5\u6b63\u786e\u62fc\u5199\u5355\u8bcd\uff1f<\/li>\n<li>\u4e3a\u4ec0\u4e48LLM\u65e0\u6cd5\u5b8c\u6210\u5982\u5b57\u7b26\u4e32\u53cd\u8f6c\u8fd9\u6837\u7684\u7b80\u5355\u5b57\u7b26\u4e32\u5904\u7406\u4efb\u52a1\uff1f<\/li>\n<li>\u4e3a\u4ec0\u4e48LLM\u5728\u5904\u7406\u975e\u82f1\u8bed\u8bed\u8a00\uff08\u6bd4\u5982\u65e5\u8bed\uff09\u65f6\u8868\u73b0\u4e0d\u4f73\uff1f<\/li>\n<li>\u4e3a\u4ec0\u4e48LLM\u5728\u8fdb\u884c\u7b80\u5355\u7b97\u672f\u8fd0\u7b97\u65f6\u8868\u73b0\u4e0d\u4f73\uff1f<\/li>\n<li>\u4e3a\u4ec0\u4e48GPT-2\u5728Python\u7f16\u7a0b\u65f6\u9047\u5230\u4e86\u989d\u5916\u7684\u56f0\u96be\uff1f<\/li>\n<li>\u4e3a\u4ec0\u4e48\u6211\u7684LLM\u5728\u9047\u5230&#8221;&lt;|endoftext|&gt;&#8221;\u5b57\u7b26\u4e32\u65f6\u7a81\u7136\u505c\u6b62\u5de5\u4f5c\uff1f<\/li>\n<li>\u6211\u6536\u5230\u7684\u5173\u4e8e\u201c\u5c3e\u968f\u7a7a\u683c\u201d\u7684\u5947\u602a\u8b66\u544a\u662f\u4ec0\u4e48\u610f\u601d\uff1f<\/li>\n<li>\u5f53\u6211\u8be2\u95ee\u5173\u4e8e&#8221;SolidGoldMagikarp&#8221;\u65f6\uff0c\u4e3a\u4ec0\u4e48LLM\u4f1a\u5d29\u6e83\uff1f<\/li>\n<li>\u4e3a\u4ec0\u4e48\u6211\u5e94\u8be5\u66f4\u503e\u5411\u4e8e\u4f7f\u7528YAML\u800c\u4e0d\u662fJSON\u6765\u4e0eLLMs\u4ea4\u4e92\uff1f<\/li>\n<li>\u4e3a\u4ec0\u4e48LLM\u5b9e\u9645\u4e0a\u5e76\u4e0d\u662f\u771f\u6b63\u7684\u7aef\u5230\u7aef\u8bed\u8a00\u6a21\u578b\uff1f<\/li>\n<li>\u75db\u82e6\u7684\u6839\u6e90\u7a76\u7adf\u662f\u4ec0\u4e48\uff1f<\/li>\n<\/ul>\n<p>\u6211\u4eec\u5c06\u5728\u89c6\u9891\u7684\u6700\u540e\u56de\u5230\u8fd9\u4e9b\u8bdd\u9898\u3002<br \/>\n\u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u6765\u52a0\u8f7d\u8fd9\u4e2a[<a href=\"https:\/\/tiktokenizer.vercel.app\/\">\u4ee4\u724c\u5316\u5728\u7ebf\u5de5\u5177<\/a>]\u3002\u8fd9\u4e2a\u5728\u7ebf\u5de5\u5177\u7684\u4f18\u52bf\u5728\u4e8e\uff0c\u4ee4\u724c\u5316\u8fc7\u7a0b\u662f\u5728\u60a8\u7684\u7f51\u7edc\u6d4f\u89c8\u5668\u4e2d\u5b9e\u65f6\u8fdb\u884c\u7684\uff0c\u60a8\u53ef\u4ee5\u8f7b\u677e\u5730\u5728\u8f93\u5165\u6846\u8f93\u5165\u6587\u672c\u5b57\u7b26\u4e32\uff0c\u5e76\u7acb\u5373\u5728\u53f3\u4fa7\u770b\u5230\u4ee4\u724c\u5316\u7684\u7ed3\u679c\u3002\u9875\u9762\u9876\u90e8\u663e\u793a\u6211\u4eec\u76ee\u524d\u4f7f\u7528\u7684\u662f<code>gpt2<\/code>\u4ee4\u724c\u5316\u5668\uff0c\u6211\u4eec\u53ef\u4ee5\u770b\u5230\u793a\u4f8b\u4e2d\u7684\u5b57\u7b26\u4e32\u6b63\u5728\u88ab\u5206\u89e3\u6210300\u4e2a\u4ee4\u724c\u3002\u8fd9\u4e9b\u4ee4\u724c\u4ee5\u4e0d\u540c\u7684\u989c\u8272\u6e05\u6670\u5c55\u793a\uff1a<\/p>\n<p><img decoding=\"async\" title=\"Tokenization(\u5206\u8bcd\u6807\u8bb0\u5316)-1\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2024\/02\/e13ba1f7f6db156.png\" alt=\"Tokenization(\u5206\u8bcd\u6807\u8bb0\u5316)-1\" \/><\/p>\n<p>\u4f8b\u5982\uff0c\u5b57\u7b26\u4e32&#8221;Tokenization&#8221;\u88ab\u7f16\u7801\u4e3a\u4ee4\u724c30642\uff0c\u7d27\u63a5\u7740\u662f\u4ee4\u724c1634\u3002\u4ee4\u724c&#8221; is&#8221;\uff08\u6ce8\u610f\u8fd9\u91cc\u5305\u542b\u4e09\u4e2a\u5b57\u7b26\uff0c\u5305\u62ec\u524d\u9762\u7684\u7a7a\u683c\uff0c\u8fd9\u4e00\u70b9\u5f88\u91cd\u8981\uff01\uff09\u7684\u7d22\u5f15\u662f318\u3002\u8981\u6ce8\u610f\u7684\u662f\u7a7a\u683c\uff0c\u56e0\u4e3a\u5b83\u786e\u5b9e\u5b58\u5728\u4e8e\u5b57\u7b26\u4e32\u4e2d\uff0c\u5fc5\u987b\u4e0e\u6240\u6709\u5176\u4ed6\u5b57\u7b26\u4e00\u8d77\u88ab\u4ee4\u724c\u5316\uff0c\u4f46\u4e3a\u4e86\u53ef\u89c6\u5316\u7684\u6e05\u6670\u901a\u5e38\u88ab\u7701\u7565\u3002\u60a8\u53ef\u4ee5\u5728\u5de5\u5177\u7684\u5e95\u90e8\u5207\u6362\u7a7a\u683c\u7684\u53ef\u89c6\u5316\u663e\u793a\u3002\u540c\u6837\uff0c\u4ee4\u724c&#8221; at&#8221;\u662f379\uff0c&#8221; the&#8221;\u662f262\uff0c\u7b49\u7b49\u3002<br \/>\n\u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u6765\u770b\u4e00\u4e2a\u7b80\u5355\u7684\u7b97\u672f\u4f8b\u5b50\u3002\u8fd9\u91cc\u6211\u4eec\u53d1\u73b0\uff0c\u6570\u5b57\u53ef\u80fd\u4f1a\u88ab\u4ee4\u724c\u5316\u5668\u4ee5\u4e0d\u4e00\u81f4\u7684\u65b9\u5f0f\u5206\u89e3\u3002\u6bd4\u5982\uff0c\u6570\u5b57127\u88ab\u5206\u89e3\u4e3a\u4e00\u4e2a\u7531\u4e09\u4e2a\u5b57\u7b26\u7ec4\u6210\u7684\u5355\u4e00\u4ee4\u724c\uff0c\u800c\u6570\u5b57677\u5219\u88ab\u5206\u89e3\u4e3a\u4e24\u4e2a\u4ee4\u724c\uff1a\u4ee4\u724c&#8221; 6&#8243;\uff08\u518d\u6b21\u6ce8\u610f\u524d\u9762\u7684\u7a7a\u683c\uff01\uff09\u548c\u4ee4\u724c&#8221;77&#8243;\u3002\u6211\u4eec\u4f9d\u8d56\u4e8e\u5927\u578b\u8bed\u8a00\u6a21\u578b\u6765\u7406\u89e3\u8fd9\u79cd\u968f\u610f\u6027\u3002\u6a21\u578b\u5fc5\u987b\u5728\u5176\u53c2\u6570\u548c\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u5b66\u4e60\uff0c\u8fd9\u4e24\u4e2a\u4ee4\u724c\uff08&#8221; 6&#8243;\u548c&#8221;77&#8243;\uff09\u5b9e\u9645\u4e0a\u7ed3\u5408\u8d77\u6765\u8868\u793a\u6570\u5b57677\u3002\u540c\u6837\uff0c\u5982\u679cLLM\u60f3\u8981\u9884\u6d4b\u8fd9\u4e2a\u52a0\u6cd5\u7684\u7ed3\u679c\u6570\u5b57\u662f804\uff0c\u5b83\u5fc5\u987b\u5206\u4e24\u6b65\u8f93\u51fa\uff1a\u9996\u5148\u53d1\u51fa\u4ee4\u724c&#8221; 8&#8243;\uff0c\u7136\u540e\u662f\u4ee4\u724c&#8221;04&#8243;\u3002\u8bf7\u6ce8\u610f\uff0c\u6240\u6709\u8fd9\u4e9b\u5206\u89e3\u770b\u8d77\u6765\u90fd\u662f\u5b8c\u5168\u4efb\u610f\u7684\u3002\u5728\u4e0b\u9762\u7684\u793a\u4f8b\u4e2d\uff0c\u6211\u4eec\u770b\u52301275\u88ab\u5206\u89e3\u4e3a&#8221;12&#8243;\u548c&#8221;75&#8243;\uff0c6773\u5b9e\u9645\u4e0a\u662f\u4e24\u4e2a\u4ee4\u724c&#8221; 6&#8243;\u548c&#8221;773&#8243;\uff0c\u800c8041\u5219\u88ab\u5206\u89e3\u4e3a&#8221; 8&#8243;\u548c&#8221;041&#8243;\u3002<br \/>\n\uff08\u5f85\u7eed&#8230;\uff09\uff08TODO\uff1a\u9664\u975e\u6211\u4eec\u627e\u5230\u4e00\u79cd\u65b9\u6cd5\uff0c\u80fd\u591f\u81ea\u52a8\u4ece\u89c6\u9891\u4e2d\u751f\u6210\u8fd9\u4e9b\u5185\u5bb9\uff0c\u5426\u5219\u6211\u4eec\u53ef\u80fd\u4f1a\u7ee7\u7eed\u8fd9\u4e2a\u3002\uff09<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u5404\u4f4d\u597d\uff0c\u4eca\u5929\u6211\u4eec\u8981\u63a2\u8ba8\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08LLM\uff09\u4e2d\u7684\u5206\u8bcd\u6280\u672f\u3002\u9057\u61be\u7684\u662f\uff0c\u5206\u8bcd\u5728\u5f53\u524d\u9876\u5c16\u7684LLM\u4e2d\u662f\u4e00\u4e2a\u6bd4\u8f83\u590d\u6742\u4e14\u68d8\u624b\u7684\u90e8\u5206\uff0c\u4f46\u7406\u89e3\u5b83\u7684\u67d0\u4e9b\u7ec6\u8282\u975e\u5e38\u5fc5\u8981\uff0c\u56e0\u4e3a\u5f88\u591a\u4eba\u5c06LLM\u7684\u4e00\u4e9b\u7f3a\u70b9\u5f52\u548e\u4e8e\u795e\u7ecf\u7f51\u7edc\u6216\u5176\u4ed6\u663e\u5f97\u795e\u79d8\u7684\u539f\u56e0\uff0c\u5b9e\u9645\u4e0a\u8fd9\u4e9b\u95ee\u9898\u7684\u6e90\u5934\u5f80\u5f80\u662f\u5206&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[34],"tags":[],"class_list":["post-764","post","type-post","status-publish","format-standard","hentry","category-knowledge"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts\/764","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/comments?post=764"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/posts\/764\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/media?parent=764"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/categories?post=764"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/de\/wp-json\/wp\/v2\/tags?post=764"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}