{"id":597,"date":"2024-08-12T16:51:49","date_gmt":"2024-08-12T08:51:49","guid":{"rendered":"http:\/\/mapleleaf666.vip\/?p=597"},"modified":"2024-08-12T16:54:02","modified_gmt":"2024-08-12T08:54:02","slug":"%e4%b9%a6%e7%94%9f%e5%a4%a7%e6%a8%a1%e5%9e%8b%e5%ae%9e%e6%88%98%e8%90%a5-llamaindexinternlm2-rag%e5%ae%9e%e8%b7%b5","status":"publish","type":"post","link":"http:\/\/mapleleaf666.vip\/?p=597","title":{"rendered":"\u4e66\u751f\u5927\u6a21\u578b\u5b9e\u6218\u8425\u2014\u2014llamaindex+Internlm2 RAG\u5b9e\u8df5"},"content":{"rendered":"\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u672c\u6587\u53c2\u8003\u5b9e\u6218\u8425\u63d0\u4f9b\u7684\u6559\u7a0b\uff1a<a href=\"https:\/\/github.com\/InternLM\/Tutorial\/blob\/camp3\/docs\/L1\/LlamaIndex\/readme.md\">camp3 \u00b7 InternLM\/Tutorial (github.com)<\/a><\/p>\n<\/blockquote>\n\n\n\n<p>\u672c\u6587\u5c06\u8bb0\u5f55RAG\u7684\u5177\u4f53\u5b9e\u8df5\u8fc7\u7a0b\uff0c\u5177\u4f53\u5185\u5bb9\u5305\u542b\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>RAG\u539f\u7406\u4ecb\u7ecd<\/li>\n\n\n\n<li>\u73af\u5883\u914d\u7f6e<\/li>\n\n\n\n<li>LlamaIndex HuggingFaceLLM<\/li>\n\n\n\n<li>LlamaIndex RAG<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">1 RAG\u539f\u7406\u4ecb\u7ecd<\/h2>\n\n\n\n<p>\u7ed9\u6a21\u578b\u6ce8\u5165\u65b0\u77e5\u8bc6\u7684\u65b9\u5f0f\uff0c\u53ef\u4ee5\u7b80\u5355\u5206\u4e3a\u4e24\u79cd\u65b9\u5f0f\uff0c\u4e00\u79cd\u662f\u5185\u90e8\u7684\uff0c\u5373\u66f4\u65b0\u6a21\u578b\u7684\u6743\u91cd\uff0c\u53e6\u4e00\u4e2a\u5c31\u662f\u5916\u90e8\u7684\u65b9\u5f0f\uff0c\u7ed9\u6a21\u578b\u6ce8\u5165\u683c\u5916\u7684\u4e0a\u4e0b\u6587\u6216\u8005\u8bf4\u5916\u90e8\u4fe1\u606f\uff0c\u4e0d\u6539\u53d8\u5b83\u7684\u7684\u6743\u91cd\u3002 \u7b2c\u4e00\u79cd\u65b9\u5f0f\uff0c\u6539\u53d8\u4e86\u6a21\u578b\u7684\u6743\u91cd\u5373\u8fdb\u884c\u6a21\u578b\u8bad\u7ec3\uff0c\u8fd9\u662f\u4e00\u4ef6\u4ee3\u4ef7\u6bd4\u8f83\u5927\u7684\u4e8b\u60c5\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u5177\u4f53\u7684\u8bad\u7ec3\u8fc7\u7a0b\uff0c\u53ef\u4ee5\u53c2\u8003<a href=\"https:\/\/arxiv.org\/abs\/2403.17297\">InternLM2<\/a>\u6280\u672f\u62a5\u544a\u3002\u7b2c\u4e8c\u79cd\u65b9\u5f0f\uff0c\u5e76\u4e0d\u6539\u53d8\u6a21\u578b\u7684\u6743\u91cd\uff0c\u53ea\u662f\u7ed9\u6a21\u578b\u5f15\u5165\u683c\u5916\u7684\u4fe1\u606f\u3002<\/p>\n\n\n\n<p>RAG\u6b63\u662f\u5c5e\u4e8e\u7b2c\u4e8c\u79cd\u65b9\u5f0f\u3002RAG\uff08Retrieval-Augmented Generation\uff09\u662f\u4e00\u79cd\u5c06\u4fe1\u606f\u68c0\u7d22\u548c\u751f\u6210\u5f0f\u6a21\u578b\u7ed3\u5408\u7684\u65b9\u6cd5\u3002\u5b83\u5229\u7528\u5916\u90e8\u77e5\u8bc6\u5e93\u6216\u6587\u6863\u5e93\u8fdb\u884c\u4fe1\u606f\u68c0\u7d22\uff0c\u5e76\u5c06\u68c0\u7d22\u5230\u7684\u4fe1\u606f\u4e0e\u751f\u6210\u6a21\u578b\u7ed3\u5408\u8d77\u6765\uff0c\u4ee5\u751f\u6210\u66f4\u51c6\u786e\u548c\u4fe1\u606f\u4e30\u5bcc\u7684\u56de\u7b54\u3002\u901a\u5e38\u5176\u5b8c\u6574\u6d41\u7a0b\u5305\u62ec\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u6570\u636e\u51c6\u5907\u9636\u6bb5\uff1a\n<ul class=\"wp-block-list\">\n<li>\u6570\u636e\u63d0\u53d6\uff1a\u4ece\u5916\u90e8\u77e5\u8bc6\u5e93\u3001\u6587\u6863\u5e93\u6216\u4e92\u8054\u7f51\u7b49\u6765\u6e90\u4e2d\u63d0\u53d6\u76f8\u5173\u6570\u636e\u3002\u8fd9\u4e9b\u6570\u636e\u53ef\u4ee5\u5305\u62ec\u6587\u7ae0\u3001\u8bba\u6587\u3001\u7f51\u9875\u3001\u4e66\u7c4d\u7b49\u591a\u79cd\u5f62\u5f0f\u3002<\/li>\n\n\n\n<li>\u5206\u5757\uff08Chunking\uff09\uff1a\u5c06\u63d0\u53d6\u5230\u7684\u5927\u5757\u6570\u636e\u5206\u5272\u6210\u8f83\u5c0f\u7684\u6bb5\u843d\u6216\u7247\u6bb5\uff08chunk\uff09\uff0c\u4ee5\u4fbf\u66f4\u9ad8\u6548\u5730\u8fdb\u884c\u540e\u7eed\u5904\u7406\u548c\u68c0\u7d22\u3002<\/li>\n\n\n\n<li>\u5411\u91cf\u5316\uff08embedding\uff09\uff1a\u4f7f\u7528\u9884\u8bad\u7ec3\u7684\u5d4c\u5165\u6a21\u578b\uff08\u5982BERT\u3001Sentence-BERT\u7b49\uff09\u5c06\u5206\u5757\u540e\u7684\u6570\u636e\u8f6c\u6362\u4e3a\u5411\u91cf\u8868\u793a\u3002\u8fd9\u4e9b\u5411\u91cf\u8868\u793a\u53ef\u4ee5\u6355\u6349\u5230\u6570\u636e\u7684\u8bed\u4e49\u4fe1\u606f\uff0c\u4fbf\u4e8e\u540e\u7eed\u7684\u76f8\u4f3c\u5ea6\u8ba1\u7b97\u3002<\/li>\n\n\n\n<li>\u6570\u636e\u5165\u5e93\uff1a\u5c06\u5411\u91cf\u5316\u540e\u7684\u6570\u636e\u5b58\u50a8\u5230\u4e00\u4e2a\u9ad8\u6548\u7684\u68c0\u7d22\u7cfb\u7edf\u4e2d\uff08\u5982FAISS\u3001Elasticsearch\u7b49\uff09\uff0c\u4ee5\u4fbf\u5feb\u901f\u68c0\u7d22\u548c\u5339\u914d\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>\u68c0\u7d22\u751f\u6210\u9636\u6bb5\uff1a\n<ul class=\"wp-block-list\">\n<li>\u95ee\u9898\u5411\u91cf\u5316\uff1a\u7528\u6237\u63d0\u51fa\u95ee\u9898\u540e\uff0c\u4f7f\u7528\u4e0e\u6570\u636e\u5411\u91cf\u5316\u76f8\u540c\u7684\u5d4c\u5165\u6a21\u578b\u5c06\u95ee\u9898\u8f6c\u6362\u4e3a\u5411\u91cf\u8868\u793a\u3002<\/li>\n\n\n\n<li>\u6839\u636e\u95ee\u9898\u5339\u914d\u6570\u636e\uff1a\u4f7f\u7528\u95ee\u9898\u7684\u5411\u91cf\u8868\u793a\u5728\u6570\u636e\u5411\u91cf\u5e93\u4e2d\u8fdb\u884c\u76f8\u4f3c\u5ea6\u8ba1\u7b97\uff0c\u627e\u5230\u4e0e\u95ee\u9898\u6700\u76f8\u5173\u7684\u82e5\u5e72\u4e2a\u6570\u636e\u7247\u6bb5\uff08chunk\uff09\u3002<\/li>\n\n\n\n<li>\u83b7\u53d6\u7d22\u5f15\u6570\u636e\uff1a\u4ece\u68c0\u7d22\u7cfb\u7edf\u4e2d\u83b7\u53d6\u4e0e\u95ee\u9898\u5339\u914d\u7684\u9ad8\u76f8\u4f3c\u5ea6\u6570\u636e\u7247\u6bb5\u3002\u8fd9\u4e9b\u7247\u6bb5\u5305\u542b\u4e86\u53ef\u80fd\u56de\u7b54\u7528\u6237\u95ee\u9898\u7684\u4fe1\u606f\u3002<\/li>\n\n\n\n<li>\u5c06\u6570\u636e\u6ce8\u5165prompt\uff1a\u5c06\u68c0\u7d22\u5230\u7684\u6570\u636e\u7247\u6bb5\u4e0e\u7528\u6237\u7684\u95ee\u9898\u7ed3\u5408\uff0c\u5f62\u6210\u4e00\u4e2a\u5b8c\u6574\u7684Prompt\u3002\u8fd9\u901a\u5e38\u5305\u62ec\u5c06\u95ee\u9898\u548c\u76f8\u5173\u6570\u636e\u7247\u6bb5\u62fc\u63a5\u5728\u4e00\u8d77\uff0c\u63d0\u4f9b\u7ed9\u751f\u6210\u6a21\u578b\u3002<\/li>\n\n\n\n<li>LLM\u751f\u6210\u7b54\u6848\uff1a\u751f\u6210\u6a21\u578b\uff08\u5982GPT-4\uff09\u63a5\u6536\u5305\u542b\u7528\u6237\u95ee\u9898\u548c\u68c0\u7d22\u6570\u636e\u7684Prompt\uff0c\u751f\u6210\u6700\u7ec8\u7684\u7b54\u6848\u3002\u751f\u6210\u6a21\u578b\u4f1a\u57fa\u4e8e\u63d0\u4f9b\u7684\u4e0a\u4e0b\u6587\u4fe1\u606f\uff0c\u751f\u6210\u4e00\u4e2a\u5c3d\u53ef\u80fd\u51c6\u786e\u548c\u4fe1\u606f\u4e30\u5bcc\u7684\u56de\u7b54\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n<figure class=\"wp-block-image\"><div class='fancybox-wrapper lazyload-container-unload' data-fancybox='post-images' href='https:\/\/s2.loli.net\/2024\/07\/30\/RCNsFI5PyhdOYWf.png'><img class=\"lazyload lazyload-style-1\" src=\"data:image\/svg+xml;base64,PCEtLUFyZ29uTG9hZGluZy0tPgo8c3ZnIHdpZHRoPSIxIiBoZWlnaHQ9IjEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgc3Ryb2tlPSIjZmZmZmZmMDAiPjxnPjwvZz4KPC9zdmc+\"  decoding=\"async\" data-original=\"https:\/\/s2.loli.net\/2024\/07\/30\/RCNsFI5PyhdOYWf.png\" src=\"data:image\/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAANSURBVBhXYzh8+PB\/AAffA0nNPuCLAAAAAElFTkSuQmCC\" alt=\"image-20240730111930978\"\/><\/div><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">2 \u76f8\u5173\u77e5\u8bc6\u4ecb\u7ecd<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">2.1 llamaindex<\/h3>\n\n\n\n<p>LlamaIndex \u662f\u4e00\u4e2a\u5f00\u6e90\u7684\u5e93\uff0c\u7528\u4e8e\u5c06\u5927\u89c4\u6a21\u8bed\u8a00\u6a21\u578b\uff08LLM\uff0c\u5982 GPT-3\uff09\u4e0e\u7ed3\u6784\u5316\u6570\u636e\uff08\u5982\u8868\u683c\u3001\u6570\u636e\u5e93\u3001\u6587\u6863\uff09\u76f8\u7ed3\u5408\u3002LlamaIndex \u5141\u8bb8\u7528\u6237\u66f4\u5bb9\u6613\u5730\u4f7f\u7528\u8bed\u8a00\u6a21\u578b\u6765\u67e5\u8be2\u548c\u5904\u7406\u8fd9\u4e9b\u6570\u636e\u6e90\uff0c\u4ee5\u4fbf\u751f\u6210\u6709\u7528\u7684\u7b54\u6848\u6216\u6267\u884c\u590d\u6742\u7684\u4efb\u52a1\u3002<\/p>\n\n\n\n<p>LlamaIndex \u7684\u4e3b\u8981\u529f\u80fd\u5305\u62ec\uff1a<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u7d22\u5f15\u521b\u5efa<\/strong>\uff1aLlamaIndex \u53ef\u4ee5\u521b\u5efa\u6570\u636e\u7684\u7d22\u5f15\uff0c\u4f7f\u8bed\u8a00\u6a21\u578b\u80fd\u591f\u66f4\u9ad8\u6548\u5730\u8bbf\u95ee\u548c\u67e5\u8be2\u8fd9\u4e9b\u6570\u636e\u3002<\/li>\n\n\n\n<li><strong>\u67e5\u8be2\u4f18\u5316<\/strong>\uff1a\u901a\u8fc7\u5bf9\u67e5\u8be2\u8fdb\u884c\u4f18\u5316\uff0cLlamaIndex \u63d0\u4f9b\u4e86\u66f4\u5feb\u548c\u66f4\u51c6\u786e\u7684\u67e5\u8be2\u7ed3\u679c\u3002<\/li>\n\n\n\n<li><strong>\u6570\u636e\u5904\u7406<\/strong>\uff1aLlamaIndex \u652f\u6301\u591a\u79cd\u6570\u636e\u683c\u5f0f\uff0c\u5e76\u4e14\u53ef\u4ee5\u5bf9\u8fd9\u4e9b\u6570\u636e\u8fdb\u884c\u9884\u5904\u7406\u548c\u8f6c\u6362\uff0c\u4ee5\u4fbf\u66f4\u597d\u5730\u9002\u5e94\u8bed\u8a00\u6a21\u578b\u7684\u9700\u6c42\u3002<\/li>\n\n\n\n<li><strong>\u96c6\u6210\u652f\u6301<\/strong>\uff1aLlamaIndex \u53ef\u4ee5\u4e0e\u591a\u79cd\u6570\u636e\u5e93\u548c\u6570\u636e\u6e90\u96c6\u6210\uff0c\u5305\u62ec SQL \u6570\u636e\u5e93\u3001NoSQL \u6570\u636e\u5e93\u3001\u6587\u4ef6\u7cfb\u7edf\u7b49\u3002<\/li>\n<\/ol>\n\n\n\n<p>LlamaIndex\u63d0\u4f9b\u4ee5\u4e0b\u5de5\u5177\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u6570\u636e\u8fde\u63a5\u5668\uff1a\u4ece\u539f\u59cb\u6765\u6e90\u548c\u683c\u5f0f\u83b7\u53d6\u4f7f\u7528\u8005\u73b0\u6709\u7684\u6570\u636e\u3002\u8fd9\u4e9b\u6765\u6e90\u53ef\u80fd\u662f API\u3001PDF\u3001SQL \u7b49\u7b49\u3002<\/li>\n\n\n\n<li>\u6570\u636e\u7d22\u5f15\uff1a\u5c06\u4f7f\u7528\u7684\u6570\u636e\u7ed3\u6784\u5316\u4e3a\u4e2d\u95f4\u8868\u793a\uff0c\u4f7f LLMs \u66f4\u5bb9\u6613\u548c\u9ad8\u6548\u5730\u4f7f\u7528\u3002<\/li>\n\n\n\n<li>\u5f15\u64ce\uff1a\u63d0\u4f9b\u81ea\u7136\u8bed\u8a00\u8bbf\u95ee\u4f60\u7684\u6570\u636e\u3002\u4f8b\u5982\uff1a\n<ul class=\"wp-block-list\">\n<li>\u67e5\u8be2\u5f15\u64ce\uff1a\u7528\u4e8e\u95ee\u7b54\u7684\u5f3a\u5927\u63a5\u53e3\uff08\u4f8b\u5982\uff0cRAG \u7ba1\u9053\uff09\u3002<\/li>\n\n\n\n<li>\u804a\u5929\u5f15\u64ce\uff1a\u7528\u4e8e\u591a\u6d88\u606f\u201c\u6765\u56de\u201d\u4e92\u52a8\u7684\u5bf9\u8bdd\u63a5\u53e3\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li>\u4ee3\u7406\uff1a\u7531 LLM \u9a71\u52a8\u7684\u77e5\u8bc6\u5de5\u4f5c\u8005\uff0c\u589e\u5f3a\u4e86\u5de5\u5177\uff0c\u4ece\u7b80\u5355\u7684\u8f85\u52a9\u51fd\u6570\u5230 API \u96c6\u6210\u7b49\u3002<\/li>\n\n\n\n<li>\u53ef\u89c2\u5bdf\u6027\/\u8bc4\u4f30\u96c6\u6210\uff1a\u4f7f\u4f7f\u7528\u8005\u80fd\u591f\u4e25\u683c\u5730\u5b9e\u9a8c\u3001\u8bc4\u4f30\u548c\u76d1\u63a7\u4f60\u7684\u5e94\u7528\u7a0b\u5e8f\uff0c\u5f62\u6210\u826f\u6027\u5faa\u73af\u3002<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">2.2 huggingface<\/h3>\n\n\n\n<p>Hugging Face \u662f\u4e00\u5bb6\u81f4\u529b\u4e8e\u63a8\u8fdb\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff08NLP\uff09\u548c\u673a\u5668\u5b66\u4e60\u7684\u516c\u53f8\uff0c\u63d0\u4f9b\u4e86\u5e7f\u6cdb\u7684\u5de5\u5177\u548c\u5e73\u53f0\uff0c\u5e2e\u52a9\u5f00\u53d1\u8005\u548c\u7814\u7a76\u4eba\u5458\u5728\u8fd9\u4e9b\u9886\u57df\u4e2d\u66f4\u9ad8\u6548\u5730\u5de5\u4f5c\u3002\u5176\u63d0\u4f9b\u4e86\u4ee5\u4e0b\u5b9e\u7528\u5de5\u5177\uff1a<\/p>\n\n\n\n<p><strong>Transformers \u5e93<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4e00\u4e2a\u6d41\u884c\u7684\u5f00\u6e90\u5e93\uff0c\u63d0\u4f9b\u4e86\u5927\u91cf\u9884\u8bad\u7ec3\u7684\u8bed\u8a00\u6a21\u578b\uff0c\u5982 BERT\u3001GPT\u3001RoBERTa \u7b49\u3002<\/li>\n\n\n\n<li>\u7528\u4e8e\u5404\u79cd NLP \u4efb\u52a1\uff0c\u5982\u6587\u672c\u5206\u7c7b\u3001\u95ee\u7b54\u3001\u7ffb\u8bd1\u548c\u6458\u8981\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>Datasets \u5e93<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u63d0\u4f9b\u4e86\u8f7b\u91cf\u7ea7\u4e14\u6613\u7528\u7684\u6570\u636e\u96c6\u52a0\u8f7d\u548c\u5904\u7406\u5de5\u5177\uff0c\u652f\u6301\u6570\u767e\u4e2a\u516c\u5f00\u6570\u636e\u96c6\u3002<\/li>\n\n\n\n<li>\u7b80\u5316\u4e86\u6570\u636e\u52a0\u8f7d\u3001\u9884\u5904\u7406\u548c\u4f7f\u7528\u7684\u8fc7\u7a0b\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>Tokenizers \u5e93<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4e00\u4e2a\u5feb\u901f\u4e14\u7075\u6d3b\u7684\u6587\u672c\u5206\u8bcd\u5e93\uff0c\u652f\u6301\u591a\u79cd\u5206\u8bcd\u7b97\u6cd5\uff0c\u5982 BERT\u3001GPT-2 \u7b49\u3002<\/li>\n\n\n\n<li>\u9ad8\u6548\u5904\u7406\u5927\u89c4\u6a21\u6570\u636e\uff0c\u4f7f\u7528\u7b80\u5355\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>Hugging Face Hub<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4e00\u4e2a\u6258\u7ba1\u548c\u5171\u4eab\u673a\u5668\u5b66\u4e60\u6a21\u578b\u548c\u6570\u636e\u96c6\u7684\u5e73\u53f0\u3002<\/li>\n\n\n\n<li>\u7528\u6237\u53ef\u4ee5\u4e0a\u4f20\u548c\u4e0b\u8f7d\u6a21\u578b\u4e0e\u6570\u636e\u96c6\uff0c\u5e76\u4e0e\u793e\u533a\u5171\u4eab\u8d44\u6e90\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>Inference API<\/strong>\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u5141\u8bb8\u7528\u6237\u8f7b\u677e\u5730\u5c06\u673a\u5668\u5b66\u4e60\u6a21\u578b\u90e8\u7f72\u5230\u751f\u4ea7\u73af\u5883\u4e2d\uff0c\u901a\u8fc7\u7b80\u5355\u7684 API \u8c03\u7528\u5728\u4e91\u7aef\u8fd0\u884c\u63a8\u7406\u4efb\u52a1\u3002<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">3 \u73af\u5883\u914d\u7f6e<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">3.1 \u57fa\u7840\u914d\u7f6e<\/h3>\n\n\n\n<p>\u9996\u5148\u6211\u4eec\u4f9d\u7136\u4f7f\u7528InterLM\u63d0\u4f9b\u7684\u5f00\u53d1\u673a\uff0c\u91c7\u7528\u7684\u914d\u7f6e\u5982\u4e0b\u6240\u793a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><div class='fancybox-wrapper lazyload-container-unload' data-fancybox='post-images' href='https:\/\/s2.loli.net\/2024\/07\/30\/d2tbcJMsXNKh9QC.png'><img class=\"lazyload lazyload-style-1\" src=\"data:image\/svg+xml;base64,PCEtLUFyZ29uTG9hZGluZy0tPgo8c3ZnIHdpZHRoPSIxIiBoZWlnaHQ9IjEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgc3Ryb2tlPSIjZmZmZmZmMDAiPjxnPjwvZz4KPC9zdmc+\"  decoding=\"async\" data-original=\"https:\/\/s2.loli.net\/2024\/07\/30\/d2tbcJMsXNKh9QC.png\" src=\"data:image\/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAANSURBVBhXYzh8+PB\/AAffA0nNPuCLAAAAAElFTkSuQmCC\" alt=\"image-20240730155029943\"\/><\/div><\/figure>\n\n\n\n<p>\u7136\u540e\u6211\u4eec\u5728\u7ec8\u7aef\u901a\u8fc7\u547d\u4ee4\u521b\u5efa\u65b0\u7684\u73af\u5883\uff1a<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>conda create -n llamaindex python=3.10<\/code><\/pre>\n\n\n\n<p>\u901a\u8fc7\u5982\u4e0b\u547d\u4ee4\u8fdb\u5165\u73af\u5883\uff0c\u5e76\u5b89\u88c5\u5fc5\u8981\u7684\u5305<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>conda activate llamaindex\nconda install pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.7 -c pytorch -c nvidia<\/code><\/pre>\n\n\n\n<p>\u518d\u5b89\u88c5python\u4f9d\u8d56\u5305<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>pip install einops\npip install  protobuf<\/code><\/pre>\n\n\n\n<h3 class=\"wp-block-heading\">3.2 \u5b89\u88c5llamaindex\u76f8\u5173\u7684\u5305<\/h3>\n\n\n\n<pre class=\"wp-block-code\"><code>conda activate llamaindex\npip install llama-index==0.10.38 llama-index-llms-huggingface==0.2.0 \"transformers&#91;torch]==4.41.1\" \"huggingface_hub&#91;inference]==0.23.1\" huggingface_hub==0.23.1 sentence-transformers==2.7.0 sentencepiece==0.2.0<\/code><\/pre>\n\n\n\n<h3 class=\"wp-block-heading\">3.3 \u4e0b\u8f7dSentence Transformer \u6a21\u578b<\/h3>\n\n\n\n<p>\u8fd9\u4e2a\u6a21\u578b\u662f\u76f8\u5bf9\u8f7b\u91cf\u3001\u652f\u6301\u4e2d\u6587\u4e14\u6548\u679c\u8f83\u597d\u7684\u6e90\u8bcd\u5411\u91cf\u6a21\u578b\uff0c\u5148\u8981\u65b0\u5efa\u4e00\u4e2apython\u6587\u4ef6<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>cd ~\nmkdir llamaindex_demo\nmkdir model\ncd ~\/llamaindex_demo\ntouch download_hf.py<\/code><\/pre>\n\n\n\n<p>\u5e76\u5c06\u5982\u4e0b\u4ee3\u7801\u7c98\u8d34\u5230\u6587\u4ef6\u4e2d<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>import os\n\n# \u8bbe\u7f6e\u73af\u5883\u53d8\u91cf\nos.environ&#91;'HF_ENDPOINT'] = 'https:\/\/hf-mirror.com'\n\n# \u4e0b\u8f7d\u6a21\u578b\nos.system('huggingface-cli download --resume-download sentence-transformers\/paraphrase-multilingual-MiniLM-L12-v2 --local-dir \/root\/model\/sentence-transformer')<\/code><\/pre>\n\n\n\n<p>\u5728llamaindex\u73af\u5883\u4e2d\u8fd0\u884c\u8be5\u811a\u672c<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">3.4 \u4e0b\u8f7d NLTK \u76f8\u5173\u8d44\u6e90<\/h3>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>NLTK\uff0c\u5168\u79f0\u662fNatural Language Toolkit\uff0c\u662f\u4e00\u4e2a\u7528\u4e8e\u5904\u7406\u548c\u5206\u6790\u4eba\u7c7b\u8bed\u8a00\u6570\u636e\u7684Python\u5e93\u3002\u5b83\u63d0\u4f9b\u4e86\u4e00\u7cfb\u5217\u5de5\u5177\u548c\u5e93\uff0c\u7528\u4e8e\u6587\u672c\u5904\u7406\u3001\u8bed\u6cd5\u5206\u6790\u3001\u8bed\u8a00\u5efa\u6a21\u3001\u4fe1\u606f\u63d0\u53d6\u3001\u673a\u5668\u5b66\u4e60\u7b49\u4efb\u52a1<\/p>\n<\/blockquote>\n\n\n\n<p>\u6211\u4eec\u5728\u4f7f\u7528\u5f00\u6e90\u8bcd\u5411\u91cf\u6a21\u578b\u6784\u5efa\u5f00\u6e90\u8bcd\u5411\u91cf\u7684\u65f6\u5019\uff0c\u9700\u8981\u7528\u5230\u7b2c\u4e09\u65b9\u5e93 <code>nltk<\/code> \u7684\u4e00\u4e9b\u8d44\u6e90\u3002\u6b63\u5e38\u60c5\u51b5\u4e0b\uff0c\u5176\u4f1a\u81ea\u52a8\u4ece\u4e92\u8054\u7f51\u4e0a\u4e0b\u8f7d\uff0c\u4f46\u53ef\u80fd\u7531\u4e8e\u7f51\u7edc\u539f\u56e0\u4f1a\u5bfc\u81f4\u4e0b\u8f7d\u4e2d\u65ad\uff0c\u6b64\u5904\u6211\u4eec\u53ef\u4ee5\u4ece\u56fd\u5185\u4ed3\u5e93\u955c\u50cf\u5730\u5740\u4e0b\u8f7d\u76f8\u5173\u8d44\u6e90\uff0c\u4fdd\u5b58\u5230\u670d\u52a1\u5668\u4e0a\u3002 \u6211\u4eec\u7528\u4ee5\u4e0b\u547d\u4ee4\u4e0b\u8f7d nltk \u8d44\u6e90\u5e76\u89e3\u538b\u5230\u670d\u52a1\u5668\u4e0a\uff1a<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">cd \/root<br>git clone https:\/\/gitee.com\/yzy0612\/nltk_data.git  --branch gh-pages<br>cd nltk_data<br>mv packages\/*  .\/<br>cd tokenizers<br>unzip punkt.zip<br>cd ..\/taggers<br>unzip averaged_perceptron_tagger.zip<\/pre>\n\n\n\n<h2 class=\"wp-block-heading\">4 LlamaIndex HuggingFaceLLM<\/h2>\n\n\n\n<p>\u8fd0\u884c\u4ee5\u4e0b\u6307\u4ee4\uff0c\u628a <code>InternLM2 1.8B<\/code> \u8f6f\u8fde\u63a5\u51fa\u6765<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u8fd9\u4e2a\u547d\u4ee4\u4f1a\u5728\u5f53\u524d\u76ee\u5f55\uff08<code>.\/<\/code>\uff09\u4e0b\u521b\u5efa\u4e00\u4e2a\u7b26\u53f7\u94fe\u63a5\uff0c\u8fd9\u4e2a\u94fe\u63a5\u4f1a\u6307\u5411\u6307\u5b9a\u76ee\u5f55\u3002\u8fd9\u6837\uff0c\u5728\u5f53\u524d\u76ee\u5f55\u4e0b\uff0c\u4f7f\u7528\u8fd9\u4e2a\u7b26\u53f7\u94fe\u63a5\u5c31\u53ef\u4ee5\u8bbf\u95ee\u5b9e\u9645\u5b58\u50a8\u5728\u6307\u5b9a\u76ee\u5f55\u7684\u5185\u5bb9<\/p>\n<\/blockquote>\n\n\n\n<pre class=\"wp-block-code\"><code>cd ~\/model<br>ln -s \/root\/share\/new_models\/Shanghai_AI_Laboratory\/internlm2-chat-1_8b\/ .\/<\/code><\/pre>\n\n\n\n<p>\u6211\u4eec\u521b\u5efa\u4e00\u4e2apython\u811a\u672c\uff0c\u6765\u8bd5\u9a8c\u76f4\u63a5\u5411\u5927\u6a21\u578b\u63d0\u4e00\u4e2a\u95ee\u9898<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>from llama_index.llms.huggingface import HuggingFaceLLM<br>from llama_index.core.llms import ChatMessage<br>llm = HuggingFaceLLM(<br>    model_name=\"\/root\/model\/internlm2-chat-1_8b\",<br>    tokenizer_name=\"\/root\/model\/internlm2-chat-1_8b\",<br>    model_kwargs={\"trust_remote_code\":True},<br>    tokenizer_kwargs={\"trust_remote_code\":True}<br>)<br><br>rsp = llm.chat(messages=&#91;ChatMessage(content=\"LMDeploy\u662f\u4ec0\u4e48\uff1f\")])<br>print(rsp)<\/code><\/pre>\n\n\n\n<p>\u8fd0\u884c\u7ed3\u679c\u5982\u4e0b\u6240\u793a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><div class='fancybox-wrapper lazyload-container-unload' data-fancybox='post-images' href='https:\/\/s2.loli.net\/2024\/07\/30\/qbFo6Kp3v1XScVU.png'><img class=\"lazyload lazyload-style-1\" src=\"data:image\/svg+xml;base64,PCEtLUFyZ29uTG9hZGluZy0tPgo8c3ZnIHdpZHRoPSIxIiBoZWlnaHQ9IjEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgc3Ryb2tlPSIjZmZmZmZmMDAiPjxnPjwvZz4KPC9zdmc+\"  decoding=\"async\" data-original=\"https:\/\/s2.loli.net\/2024\/07\/30\/qbFo6Kp3v1XScVU.png\" src=\"data:image\/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAANSURBVBhXYzh8+PB\/AAffA0nNPuCLAAAAAElFTkSuQmCC\" alt=\"image-20240730204153373\"\/><\/div><\/figure>\n\n\n\n<p>\u53ef\u4ee5\u770b\u5230\u6a21\u578b\u7684\u6548\u679c\u5f88\u4e00\u822c\uff0c\u5e76\u4e0d\u80fd\u6709\u6548\u56de\u7b54\u95ee\u9898\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">5 LlamaIndex RAG<\/h2>\n\n\n\n<p>\u6211\u4eec\u5c06\u91c7\u7528RAG\u6280\u672f\u6765\u589e\u5f3a\u5927\u6a21\u578b\u7684\u80fd\u529b\uff0c\u9996\u5148\u9700\u8981\u5b89\u88c5LlamaIndex\u8bcd\u5d4c\u5165\u5411\u91cf\u4f9d\u8d56<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>pip install llama-index-embeddings-huggingface llama-index-embeddings-instructor<\/code><\/pre>\n\n\n\n<p>\u7136\u540e\u6211\u4eec\u9700\u8981\u8fd0\u884c\u4ee5\u4e0b\u547d\u4ee4\u6765\u83b7\u53d6\u65b0\u7684\u77e5\u8bc6\u5e93\uff1a<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>cd ~\/llamaindex_demo\nmkdir data\ncd data\ngit clone https:\/\/github.com\/InternLM\/lmdeploy.git\nmv lmdeploy\/README_zh-CN.md .\/<\/code><\/pre>\n\n\n\n<p>\u6211\u4eec\u518d\u521b\u5efa\u4e00\u4e2a\u65b0\u7684\u811a\u672cllamaindex_RAG.py\uff0c\u586b\u5165\u4ee5\u4e0b\u4ee3\u7801\uff1a<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings<br><br>from llama_index.embeddings.huggingface import HuggingFaceEmbedding<br>from llama_index.llms.huggingface import HuggingFaceLLM<br><br>#\u521d\u59cb\u5316\u4e00\u4e2aHuggingFaceEmbedding\u5bf9\u8c61\uff0c\u7528\u4e8e\u5c06\u6587\u672c\u8f6c\u6362\u4e3a\u5411\u91cf\u8868\u793a<br>embed_model = HuggingFaceEmbedding(<br>#\u6307\u5b9a\u4e86\u4e00\u4e2a\u9884\u8bad\u7ec3\u7684sentence-transformer\u6a21\u578b\u7684\u8def\u5f84<br>    model_name=\"\/root\/model\/sentence-transformer\"<br>)<br>#\u5c06\u521b\u5efa\u7684\u5d4c\u5165\u6a21\u578b\u8d4b\u503c\u7ed9\u5168\u5c40\u8bbe\u7f6e\u7684embed_model\u5c5e\u6027\uff0c<br>#\u8fd9\u6837\u5728\u540e\u7eed\u7684\u7d22\u5f15\u6784\u5efa\u8fc7\u7a0b\u4e2d\u5c31\u4f1a\u4f7f\u7528\u8fd9\u4e2a\u6a21\u578b\u3002<br>Settings.embed_model = embed_model<br><br>llm = HuggingFaceLLM(<br>    model_name=\"\/root\/model\/internlm2-chat-1_8b\",<br>    tokenizer_name=\"\/root\/model\/internlm2-chat-1_8b\",<br>    model_kwargs={\"trust_remote_code\":True},<br>    tokenizer_kwargs={\"trust_remote_code\":True}<br>)<br>#\u8bbe\u7f6e\u5168\u5c40\u7684llm\u5c5e\u6027\uff0c\u8fd9\u6837\u5728\u7d22\u5f15\u67e5\u8be2\u65f6\u4f1a\u4f7f\u7528\u8fd9\u4e2a\u6a21\u578b\u3002<br>Settings.llm = llm<br><br>#\u4ece\u6307\u5b9a\u76ee\u5f55\u8bfb\u53d6\u6240\u6709\u6587\u6863\uff0c\u5e76\u52a0\u8f7d\u6570\u636e\u5230\u5185\u5b58\u4e2d<br>documents = SimpleDirectoryReader(\"\/root\/llamaindex_demo\/data\").load_data()<br>#\u521b\u5efa\u4e00\u4e2aVectorStoreIndex\uff0c\u5e76\u4f7f\u7528\u4e4b\u524d\u52a0\u8f7d\u7684\u6587\u6863\u6765\u6784\u5efa\u7d22\u5f15\u3002<br># \u6b64\u7d22\u5f15\u5c06\u6587\u6863\u8f6c\u6362\u4e3a\u5411\u91cf\uff0c\u5e76\u5b58\u50a8\u8fd9\u4e9b\u5411\u91cf\u4ee5\u4fbf\u4e8e\u5feb\u901f\u68c0\u7d22\u3002<br>index = VectorStoreIndex.from_documents(documents)<br># \u521b\u5efa\u4e00\u4e2a\u67e5\u8be2\u5f15\u64ce\uff0c\u8fd9\u4e2a\u5f15\u64ce\u53ef\u4ee5\u63a5\u6536\u67e5\u8be2\u5e76\u8fd4\u56de\u76f8\u5173\u6587\u6863\u7684\u54cd\u5e94\u3002<br>query_engine = index.as_query_engine()<br>response = query_engine.query(\"LMDeploy\u662f\u4ec0\u4e48?\")<br><br>print(response)<\/code><\/pre>\n\n\n\n<p>\u4e4b\u540e\u8fd0\u884c\u8be5\u6587\u4ef6\uff0c\u8fd0\u884c\u7ed3\u679c\u5982\u4e0b\u6240\u793a\uff1a<\/p>\n\n\n\n<figure class=\"wp-block-image\"><div class='fancybox-wrapper lazyload-container-unload' data-fancybox='post-images' href='https:\/\/s2.loli.net\/2024\/07\/30\/V27rMnaFxAN9UTO.png'><img class=\"lazyload lazyload-style-1\" src=\"data:image\/svg+xml;base64,PCEtLUFyZ29uTG9hZGluZy0tPgo8c3ZnIHdpZHRoPSIxIiBoZWlnaHQ9IjEiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgc3Ryb2tlPSIjZmZmZmZmMDAiPjxnPjwvZz4KPC9zdmc+\"  decoding=\"async\" data-original=\"https:\/\/s2.loli.net\/2024\/07\/30\/V27rMnaFxAN9UTO.png\" src=\"data:image\/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsQAAA7EAZUrDhsAAAANSURBVBhXYzh8+PB\/AAffA0nNPuCLAAAAAElFTkSuQmCC\" alt=\"image-20240730210848551\"\/><\/div><\/figure>\n\n\n\n<p>\u53ef\u4ee5\u770b\u5230\u6a21\u578b\u80fd\u591f\u56de\u7b54\u6211\u4eec\u7684\u95ee\u9898\u4e86\uff0c\u4f46\u662f\u6548\u679c\u5e76\u6ca1\u6709\u975e\u5e38\u597d\uff0c\u8fd9\u662f\u56e0\u4e3a\u6211\u4eec\u63d0\u4f9b\u7684\u77e5\u8bc6\u5e93\u5bf9\u5e94\u7684\u8bad\u7ec3\u6570\u636e\u4e0d\u8db3\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u672c\u6587\u53c2\u8003\u5b9e\u6218\u8425\u63d0\u4f9b\u7684\u6559\u7a0b\uff1acamp3 \u00b7 InternLM\/Tutorial (github.com) \u672c\u6587\u5c06 [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4],"tags":[35],"class_list":["post-597","post","type-post","status-publish","format-standard","hentry","category-note","tag-llm"],"_links":{"self":[{"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=\/wp\/v2\/posts\/597","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=597"}],"version-history":[{"count":2,"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=\/wp\/v2\/posts\/597\/revisions"}],"predecessor-version":[{"id":599,"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=\/wp\/v2\/posts\/597\/revisions\/599"}],"wp:attachment":[{"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=597"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=597"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/mapleleaf666.vip\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=597"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}