{"id":117948,"date":"2026-10-26T00:17:20","date_gmt":"2026-10-25T16:17:20","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/117948.html"},"modified":"2026-10-26T00:17:20","modified_gmt":"2026-10-25T16:17:20","slug":"%e5%ae%9e%e6%88%98%e6%95%99%e7%a8%8b%ef%bc%9anvidia-jetson-nano-8gb-%e6%90%ad%e5%bb%ba%e6%9c%ac%e5%9c%b0-rag-%e5%a4%a7%e8%af%ad%e8%a8%80%e6%a8%a1%e5%9e%8b%e7%9f%a5%e8%af%86%e5%ba%93%e5%ae%8c%e6%95%b4","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/117948.html","title":{"rendered":"\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357"},"content":{"rendered":"

\"\"<\/p>\n

\u4e00\u3001\u6280\u672f\u80cc\u666f<\/h3>\n

1.1\u00a0Jetson Nano 8GB \u786c\u4ef6\u7279\u6027<\/h4>\n

\u6839\u636e NVIDIA \u5b98\u65b9 datasheet，Jetson Nano 8GB \u642d\u8f7d\u56db\u6838 Cortex-A57 \u5904\u7406\u5668，\u4e3b\u9891 1.43GHz，\u5185\u7f6e 128 \u6838 Maxwell \u67b6\u6784 GPU，\u914d\u5907 8GB LPDDR4 \u5185\u5b58，\u529f\u8017\u4ec5 5-10W，\u662f\u8fb9\u7f18 AI \u90e8\u7f72\u7684\u9ad8\u6027\u4ef7\u6bd4\u5165\u95e8\u5e73\u53f0\u3002\u5c3d\u7ba1 FP16 \u7b97\u529b\u53ea\u6709 472 GFLOPS，\u4f46\u901a\u8fc7\u6a21\u578b\u91cf\u5316\u548c\u4f18\u5316，\u5b8c\u5168\u53ef\u4ee5\u8fd0\u884c 7B \u53c2\u6570\u7ea7\u522b\u7684\u8f7b\u91cf\u5316\u5927\u8bed\u8a00\u6a21\u578b\u3002<\/p>\n<\/p>\n

1.2\u00a0\u8fb9\u7f18\u7aef RAG \u90e8\u7f72\u7684\u4ef7\u503c<\/h4>\n

RAG（\u68c0\u7d22\u589e\u5f3a\u751f\u6210）\u6280\u672f\u53ef\u4ee5\u8ba9\u5927\u8bed\u8a00\u6a21\u578b\u63a5\u5165\u79c1\u6709\u77e5\u8bc6\u5e93，\u907f\u514d\u5e7b\u89c9\u95ee\u9898\u3002\u5728\u8fb9\u7f18\u7aef\u90e8\u7f72 RAG \u7cfb\u7edf\u5177\u5907\u4e09\u5927\u4f18\u52bf：<\/p>\n<\/p>\n

  • \u6570\u636e\u5b8c\u5168\u672c\u5730\u5316，\u65e0\u9700\u4e0a\u4f20\u4e91\u7aef，\u6ee1\u8db3\u9690\u79c1\u5408\u89c4\u8981\u6c42<\/li>\n
  • \u4e0d\u4f9d\u8d56\u7f51\u7edc，\u53ef\u5728\u79bb\u7ebf\u573a\u666f\u4e0b\u4f7f\u7528<\/li>\n
  • \u54cd\u5e94\u5ef6\u8fdf\u7a33\u5b9a，\u9002\u5408\u5de5\u4e1a\u8d28\u68c0\u3001\u667a\u80fd\u5bb6\u5c45\u7b49\u5d4c\u5165\u5f0f\u573a\u666f<\/li>\n
    \n

    \"\"<\/p>\n

    \u4e8c\u3001\u73af\u5883\u51c6\u5907<\/h3>\n

    2.1 \u7cfb\u7edf\u955c\u50cf\u70e7\u5f55\u4e0e\u57fa\u7840\u914d\u7f6e<\/h4>\n

    \u9996\u5148\u9700\u8981\u70e7\u5f55\u5b98\u65b9\u9002\u914d\u7684 JetPack 4.6.1 \u7cfb\u7edf\u955c\u50cf，\u8fd9\u662f Jetson Nano \u5b98\u65b9\u6700\u7ec8\u652f\u6301\u7684\u7248\u672c，\u517c\u5bb9\u6027\u6700\u597d：<\/p>\n<\/p>\n

  • \u4ece NVIDIA \u5b98\u65b9\u7f51\u7ad9\u4e0b\u8f7d Jetson Nano 8GB \u5bf9\u5e94\u7684 JetPack 4.6.1 \u955c\u50cf<\/li>\n
  • \u4f7f\u7528 balenaEtcher \u5de5\u5177\u5c06\u955c\u50cf\u70e7\u5f55\u5230 64GB \u4ee5\u4e0a\u7684\u9ad8\u901f SD \u5361<\/li>\n
  • \u9996\u6b21\u5f00\u673a\u5b8c\u6210\u7cfb\u7edf\u914d\u7f6e，\u5efa\u8bae\u542f\u7528 SSH \u670d\u52a1\u65b9\u4fbf\u8fdc\u7a0b\u64cd\u4f5c<\/li>\n

    \u57fa\u7840\u914d\u7f6e\u547d\u4ee4：<\/p>\n<\/p>\n

    # \u5347\u7ea7\u7cfb\u7edf\u5305
    \nsudo apt update && sudo apt upgrade -y<\/p>\n

    # \u589e\u52a0\u4ea4\u6362\u7a7a\u95f4（\u89e3\u51b3\u5185\u5b58\u4e0d\u8db3\u95ee\u9898，\u5efa\u8bae\u8bbe\u7f6e4GB，\u8fc7\u5927\u4f1a\u5bfc\u81f4\u7cfb\u7edf\u5361\u987f）
    \nsudo fallocate -l 4G \/swapfile
    \nsudo chmod 600 \/swapfile
    \nsudo mkswap \/swapfile
    \nsudo swapon \/swapfile
    \n# \u6c38\u4e45\u751f\u6548
    \necho '\/swapfile none swap defaults 0 0' | sudo tee -a \/etc\/fstab<\/p>\n

    # \u914d\u7f6ejetson_clocks\u6700\u5927\u5316\u6027\u80fd
    \nsudo jetson_clocks<\/p>\n

    2.2 \u4f9d\u8d56\u5e93\u5b89\u88c5<\/h4>\n

    JetPack 4.6.1 \u81ea\u5e26 CUDA 10.2\u3001cuDNN 8.2.1，\u6211\u4eec\u9700\u8981\u5b89\u88c5 Python \u76f8\u5173\u4f9d\u8d56：<\/p>\n<\/p>\n

    # \u5b89\u88c5pip
    \nsudo apt install python3-pip -y
    \npip3 install –upgrade pip<\/p>\n

    # \u5b89\u88c5PyTorch（\u5b98\u65b9\u9884\u7f16\u8bd1\u7248\u672c，\u9002\u914dCUDA 10.2）
    \nwget https:\/\/nvidia.box.com\/shared\/static\/fjtbno0vpo676a25cgvuqc1wty0fkkg6.whl -O torch-1.10.0-cp36-cp36m-linux_aarch64.whl
    \npip3 install torch-1.10.0-cp36-cp36m-linux_aarch64.whl<\/p>\n

    # \u5b89\u88c5TensorRT\u548c\u76f8\u5173\u4f9d\u8d56
    \nsudo apt install python3-libnvinfer-dev -y
    \npip3 install transformers==4.18.0 sentence-transformers==2.2.0 faiss-cpu==1.7.2 langchain==0.0.142 accelerate==0.12.0<\/p>\n

    # \u5b89\u88c5llama.cpp\u7528\u4e8e\u6a21\u578b\u91cf\u5316\u63a8\u7406
    \ngit clone https:\/\/github.com\/ggerganov\/llama.cpp.git
    \ncd llama.cpp
    \nmake LLAMA_CUDA=1 # \u542f\u7528CUDA\u52a0\u901f
    \npip3 install .<\/p>\n

    \n

    \u4e09\u3001\u6838\u5fc3\u7ec4\u4ef6\u9009\u578b<\/h3>\n

    3.1 \u8f7b\u91cf\u5316\u5927\u8bed\u8a00\u6a21\u578b\u9009\u62e9<\/h4>\n

    \u8003\u8651\u5230 Jetson Nano 8GB \u7684\u5185\u5b58\u9650\u5236，\u6211\u4eec\u9009\u62e9\u7ecf\u8fc7 4-bit \u91cf\u5316\u7684 Llama-2-7B-chat \u6a21\u578b，\u91cf\u5316\u540e\u5927\u5c0f\u4ec5 3.8GB，\u5b8c\u5168\u53ef\u4ee5\u5728\u8bbe\u5907\u4e0a\u8fd0\u884c：<\/p>\n<\/p>\n

      \n
    • \u57fa\u7840\u6a21\u578b：Meta Llama-2-7B-chat<\/li>\n
    • \u91cf\u5316\u65b9\u5f0f：4-bit GGUF \u683c\u5f0f<\/li>\n
    • \u63a8\u7406\u5f15\u64ce：llama.cpp + CUDA \u52a0\u901f<\/li>\n
    • \u5b9e\u6d4b\u5185\u5b58\u5360\u7528：~6GB（\u542b\u7cfb\u7edf\u5f00\u9500）<\/li>\n<\/ul>\n

      \u6a21\u578b\u53ef\u4ece Hugging Face \u7684 TheBloke\/Llama-2-7B-Chat-GGUF \u4ed3\u5e93\u4e0b\u8f7d，\u9009\u62e9llama-2-7b-chat.Q4_K_M.gguf\u7248\u672c\u3002<\/p>\n

      3.2 \u5411\u91cf\u6570\u636e\u5e93\u9009\u578b<\/h4>\n

      \u9488\u5bf9\u8fb9\u7f18\u8bbe\u5907\u7684\u8d44\u6e90\u9650\u5236，\u9009\u62e9 FAISS \u4f5c\u4e3a\u5411\u91cf\u6570\u636e\u5e93，\u5177\u5907\u4ee5\u4e0b\u4f18\u52bf：<\/p>\n<\/p>\n

        \n
      • \u5185\u5b58\u5360\u7528\u4f4e，\u65e0\u9700\u989d\u5916\u670d\u52a1\u8fdb\u7a0b<\/li>\n
      • \u652f\u6301 CPU\/GPU \u52a0\u901f，\u68c0\u7d22\u901f\u5ea6\u5feb<\/li>\n
      • \u63a5\u53e3\u7b80\u5355，\u9002\u5408\u5d4c\u5165\u5f0f\u573a\u666f\u90e8\u7f72<\/li>\n
      • 768 \u7ef4\u5411\u91cf\u300110 \u4e07\u6761\u6570\u636e\u5185\u5b58\u5360\u7528\u4ec5～300MB<\/li>\n<\/ul>\n
        \n

        \u56db\u3001RAG \u7cfb\u7edf\u5b9e\u73b0<\/h3>\n

        \"\"<\/p>\n

        4.1 \u6587\u6863\u9884\u5904\u7406\u6a21\u5757<\/h4>\n

        \u652f\u6301 TXT\u3001MD\u3001PDF \u683c\u5f0f\u6587\u6863\u7684\u89e3\u6790\u548c\u5207\u5206：<\/p>\n

        from langchain.document_loaders import TextLoader, PyPDFLoader
        \nfrom langchain.text_splitter import RecursiveCharacterTextSplitter
        \nimport os<\/p>\n

        def load_documents(doc_dir="docs"):
        \n """\u52a0\u8f7d\u6307\u5b9a\u76ee\u5f55\u4e0b\u7684\u6240\u6709\u6587\u6863"""
        \n documents = []
        \n for filename in os.listdir(doc_dir):
        \n file_path = os.path.join(doc_dir, filename)
        \n if filename.endswith(".txt") or filename.endswith(".md"):
        \n loader = TextLoader(file_path, encoding="utf-8")
        \n documents.extend(loader.load())
        \n elif filename.endswith(".pdf"):
        \n loader = PyPDFLoader(file_path)
        \n documents.extend(loader.load())<\/p>\n

        # \u6587\u672c\u5207\u5206，\u6bcf\u5757512\u5b57\u7b26，\u91cd\u53e064\u5b57\u7b26
        \n text_splitter = RecursiveCharacterTextSplitter(
        \n chunk_size=512,
        \n chunk_overlap=64,
        \n length_function=len
        \n )
        \n split_docs = text_splitter.split_documents(documents)
        \n print(f"\u52a0\u8f7d\u5b8c\u6210，\u5171\u5207\u5206\u4e3a{len(split_docs)}\u4e2a\u6587\u6863\u5757")
        \n return split_docs<\/p>\n

        4.2 \u5411\u91cf\u5165\u5e93\u6a21\u5757<\/h4>\n

        \u4f7f\u7528 all-MiniLM-L6-v2 \u4f5c\u4e3a\u5d4c\u5165\u6a21\u578b，\u751f\u6210 768 \u7ef4\u5411\u91cf：<\/p>\n<\/p>\n

        from langchain.embeddings import HuggingFaceEmbeddings
        \nfrom langchain.vectorstores import FAISS
        \nimport torch<\/p>\n

        def build_vector_db(documents, save_path="faiss_index"):
        \n """\u6784\u5efaFAISS\u5411\u91cf\u6570\u636e\u5e93"""
        \n # \u52a0\u8f7d\u8f7b\u91cf\u7ea7\u5d4c\u5165\u6a21\u578b，\u4f7f\u7528CPU\u8fd0\u884c（GPU\u5185\u5b58\u7559\u7ed9\u5927\u6a21\u578b）
        \n embeddings = HuggingFaceEmbeddings(
        \n model_name="all-MiniLM-L6-v2",
        \n model_kwargs={"device": "cpu"}
        \n )<\/p>\n

        # \u6784\u5efa\u5411\u91cf\u6570\u636e\u5e93
        \n db = FAISS.from_documents(documents, embeddings)<\/p>\n

        # \u4fdd\u5b58\u5230\u672c\u5730
        \n db.save_local(save_path)
        \n print(f"\u5411\u91cf\u6570\u636e\u5e93\u5df2\u4fdd\u5b58\u5230{save_path}")
        \n return db<\/p>\n

        if __name__ == "__main__":
        \n docs = load_documents()
        \n db = build_vector_db(docs)<\/p>\n

        4.3 \u68c0\u7d22\u95ee\u7b54\u6a21\u5757<\/h4>\n

        \u5b9e\u73b0\u5b8c\u6574\u7684 RAG \u68c0\u7d22 + \u751f\u6210\u6d41\u7a0b：<\/p>\n<\/p>\n

        from langchain.llms import LlamaCpp
        \nfrom langchain.chains import RetrievalQA
        \nfrom langchain.prompts import PromptTemplate<\/p>\n

        def load_rag_system(model_path, index_path="faiss_index"):
        \n """\u52a0\u8f7dRAG\u95ee\u7b54\u7cfb\u7edf"""
        \n # \u52a0\u8f7d\u5d4c\u5165\u6a21\u578b\u548c\u5411\u91cf\u6570\u636e\u5e93
        \n embeddings = HuggingFaceEmbeddings(
        \n model_name="all-MiniLM-L6-v2",
        \n model_kwargs={"device": "cpu"}
        \n )
        \n db = FAISS.load_local(index_path, embeddings)<\/p>\n

        # \u52a0\u8f7d\u91cf\u5316\u5927\u8bed\u8a00\u6a21\u578b
        \n llm = LlamaCpp(
        \n model_path=model_path,
        \n n_ctx=1024, # \u4e0a\u4e0b\u6587\u7a97\u53e3\u5927\u5c0f，Jetson Nano\u5185\u5b58\u6709\u9650\u4e0d\u5efa\u8bae\u8d85\u8fc71024
        \n n_threads=4, # \u4f7f\u75284\u4e2aCPU\u6838\u5fc3
        \n n_gpu_layers=20, # \u5c0620\u5c42Transformer\u5378\u8f7d\u5230GPU\u8fd0\u884c
        \n temperature=0.3, # \u751f\u6210\u6e29\u5ea6，\u8d8a\u4f4e\u8d8a\u51c6\u786e
        \n max_tokens=100, # \u6700\u5927\u751f\u6210\u957f\u5ea6，\u63a7\u5236\u54cd\u5e94\u65f6\u95f4
        \n verbose=False
        \n )<\/p>\n

        # \u81ea\u5b9a\u4e49Prompt\u6a21\u677f，\u51cf\u5c11\u5e7b\u89c9
        \n prompt_template = """
        \n \u57fa\u4e8e\u4ee5\u4e0b\u4e0a\u4e0b\u6587\u56de\u7b54\u7528\u6237\u95ee\u9898，\u5982\u679c\u4e0a\u4e0b\u6587\u6ca1\u6709\u76f8\u5173\u4fe1\u606f，\u8bf7\u56de\u7b54"\u6211\u65e0\u6cd5\u56de\u7b54\u8fd9\u4e2a\u95ee\u9898"，\u4e0d\u8981\u7f16\u9020\u5185\u5bb9\u3002
        \n \u4e0a\u4e0b\u6587：{context}
        \n \u95ee\u9898：{question}
        \n \u56de\u7b54：
        \n """
        \n PROMPT = PromptTemplate(
        \n template=prompt_template,
        \n input_variables=["context", "question"]
        \n )<\/p>\n

        # \u6784\u5efaRAG\u95ee\u7b54\u94fe
        \n qa_chain = RetrievalQA.from_chain_type(
        \n llm=llm,
        \n chain_type="stuff",
        \n retriever=db.as_retriever(search_kwargs={"k": 3}), # \u68c0\u7d22\u6700\u76f8\u5173\u76843\u4e2a\u6587\u6863\u5757
        \n return_source_documents=True, # \u8fd4\u56de\u6e90\u6587\u6863\u7528\u4e8e\u6eaf\u6e90
        \n chain_type_kwargs={"prompt": PROMPT}
        \n )
        \n return qa_chain<\/p>\n

        if __name__ == "__main__":
        \n qa_chain = load_rag_system(model_path="llama-2-7b-chat.Q4_K_M.gguf")<\/p>\n

        # \u4ea4\u4e92\u6d4b\u8bd5
        \n while True:
        \n user_query = input("\u8bf7\u8f93\u5165\u95ee\u9898：")
        \n if user_query in ["\u9000\u51fa", "exit", "quit"]:
        \n break
        \n result = qa_chain({"query": user_query})
        \n print(f"\u56de\u7b54：{result['result']}")
        \n print("\u53c2\u8003\u6765\u6e90：")
        \n for doc in result["source_documents"]:
        \n print(f"- {doc.metadata['source']}")<\/p>\n

        \n

        \u4e94\u3001\u6027\u80fd\u4f18\u5316\u4e0e\u8e29\u5751\u6307\u5357<\/h3>\n

        5.1 \u5185\u5b58\u4f18\u5316\u65b9\u6848<\/h4>\n
      • \u6a21\u578b\u5206\u5c42\u90e8\u7f72：\u5c06\u5d4c\u5165\u6a21\u578b\u653e\u5728 CPU \u8fd0\u884c，\u5927\u6a21\u578b\u7684 20 \u5c42 Transformer \u653e\u5728 GPU，\u5176\u4f59\u653e\u5728 CPU，\u5e73\u8861\u5185\u5b58\u5360\u7528<\/li>\n
      • \u4ea4\u6362\u7a7a\u95f4\u914d\u7f6e：\u5efa\u8bae\u914d\u7f6e 4GB \u4ea4\u6362\u7a7a\u95f4，\u907f\u514d\u6a21\u578b\u52a0\u8f7d\u65f6 OOM，\u8fc7\u5927\u4f1a\u5bfc\u81f4\u5b58\u50a8 I\/O \u74f6\u9888<\/li>\n
      • \u5173\u95ed\u4e0d\u5fc5\u8981\u670d\u52a1：\u505c\u7528\u684c\u9762\u73af\u5883\u3001\u84dd\u7259\u7b49\u4e0d\u5fc5\u8981\u670d\u52a1，\u53ef\u8282\u7701\u7ea6 1GB \u5185\u5b58：\n

        sudo systemctl set-default multi-user.target
        \nsudo systemctl stop bluetooth<\/p>\n<\/li>\n

        5.2 \u63a8\u7406\u52a0\u901f\u914d\u7f6e<\/h4>\n
      • \u542f\u7528 CUDA \u52a0\u901f：\u7f16\u8bd1 llama.cpp \u65f6\u5fc5\u987b\u6dfb\u52a0LLAMA_CUDA=1\u53c2\u6570，\u5b9e\u6d4b\u63a8\u7406\u901f\u5ea6\u6bd4\u7eaf CPU \u5feb 3-4 \u500d（\u6d4b\u8bd5\u73af\u5883：JetPack 4.6.1，Llama 2 7B INT4）<\/li>\n
      • n_gpu_layers \u53c2\u6570\u8c03\u4f18：Jetson Nano 8GB \u5efa\u8bae\u8bbe\u7f6e\u4e3a 20，\u8d85\u8fc7 20 \u4f1a\u5bfc\u81f4 GPU \u5185\u5b58\u4e0d\u8db3<\/li>\n
      • \u5b9e\u6d4b\u6027\u80fd：\u5355 token \u751f\u6210\u901f\u5ea6\u7ea6 3-5token\/s，\u68c0\u7d22\u8017\u65f6 < 50ms，100token \u4ee5\u5185\u8f93\u51fa\u7684\u6574\u4f53\u95ee\u7b54\u54cd\u5e94\u65f6\u95f4\u7ea6 20-30 \u79d2<\/li>\n

        5.3 \u5e38\u89c1\u95ee\u9898\u89e3\u51b3<\/h4>\n
      • \u6a21\u578b\u52a0\u8f7d\u5931\u8d25：\u68c0\u67e5\u662f\u5426\u4e0b\u8f7d\u4e86\u6b63\u786e\u7684 GGUF \u683c\u5f0f\u6a21\u578b，\u786e\u8ba4\u6a21\u578b\u8def\u5f84\u65e0\u8bef<\/li>\n
      • \u63a8\u7406\u65f6\u5d29\u6e83：\u5927\u6982\u7387\u662f\u5185\u5b58\u4e0d\u8db3，\u68c0\u67e5\u4ea4\u6362\u7a7a\u95f4\u662f\u5426\u6b63\u786e\u914d\u7f6e，\u9002\u5f53\u51cf\u5c11n_ctx\u548cmax_tokens\u53c2\u6570<\/li>\n
      • \u56de\u7b54\u8d28\u91cf\u5dee：\u8c03\u6574chunk_size\u548cchunk_overlap\u53c2\u6570，\u589e\u52a0\u68c0\u7d22\u8fd4\u56de\u7684\u6587\u6863\u6570\u91cfk，\u964d\u4f4etemperature\u503c<\/li>\n
        \n

        \u516d\u3001\u5b9e\u6d4b\u6548\u679c\u5c55\u793a<\/h3>\n

        \u6211\u4eec\u4ee5\u5d4c\u5165\u5f0f\u5f00\u53d1\u6587\u6863\u4f5c\u4e3a\u77e5\u8bc6\u5e93\u8fdb\u884c\u6d4b\u8bd5：<\/p>\n<\/p>\n

        \u95ee\u9898：Jetson Nano \u7684 GPU \u6709\u591a\u5c11\u4e2a CUDA \u6838\u5fc3？
        \n\u56de\u7b54：\u6839\u636e\u6587\u6863\u4fe1\u606f，Jetson Nano \u5185\u7f6e 128 \u4e2a Maxwell \u67b6\u6784\u7684 CUDA \u6838\u5fc3\u3002
        \n\u53c2\u8003\u6765\u6e90：<\/p>\n

        \u00a0<\/p>\n

          \n
        • docs\/Jetson_Nano_\u5b98\u65b9\u53c2\u6570\u8bf4\u660e.pdf
          \n\u54cd\u5e94\u8017\u65f6：22 \u79d2<\/li>\n<\/ul>\n

          \u95ee\u9898：\u5982\u4f55\u5f00\u542f Jetson Nano \u7684\u6027\u80fd\u6a21\u5f0f？
          \n\u56de\u7b54：\u53ef\u4ee5\u901a\u8fc7\u8fd0\u884csudo jetson_clocks\u547d\u4ee4\u5f00\u542f\u6027\u80fd\u6a21\u5f0f，\u6700\u5927\u5316 CPU \u548c GPU \u9891\u7387\u3002
          \n\u53c2\u8003\u6765\u6e90：<\/p>\n

          \u00a0<\/p>\n

            \n
          • docs\/Jetson_Nano_\u5f00\u53d1\u6307\u5357.md
            \n\u54cd\u5e94\u8017\u65f6：18 \u79d2<\/li>\n<\/ul>\n

            \"\"<\/p>\n

            \n

            \u516d\u3001\u603b\u7ed3\u4e0e\u6269\u5c55<\/h3>\n

            \u672c\u6587\u5b9e\u73b0\u7684 RAG \u7cfb\u7edf\u53ef\u4ee5\u5728 Jetson Nano 8GB \u4e0a\u5b8c\u5168\u79bb\u7ebf\u8fd0\u884c，\u9002\u5408\u5c0f\u4f53\u91cf\u79c1\u6709\u77e5\u8bc6\u5e93\u573a\u666f\u3002\u5982\u679c\u9700\u8981\u8fdb\u4e00\u6b65\u4f18\u5316，\u53ef\u4ee5\u8003\u8651：<\/p>\n<\/p>\n

          • \u4f7f\u7528\u66f4\u5c0f\u7684\u6a21\u578b\u5982 Qwen-1.8B，\u5c06\u63a8\u7406\u901f\u5ea6\u63d0\u5347\u5230 10token\/s \u4ee5\u4e0a，\u54cd\u5e94\u65f6\u95f4\u7f29\u77ed\u5230 10 \u79d2\u4ee5\u5185<\/li>\n
          • \u5bf9\u63a5\u8bed\u97f3\u8bc6\u522b \/ \u5408\u6210\u6a21\u5757，\u5b9e\u73b0\u8bed\u97f3\u4ea4\u4e92\u529f\u80fd<\/li>\n
          • \u6dfb\u52a0\u8f7b\u91cf WebUI \u754c\u9762，\u65b9\u4fbf\u5c40\u57df\u7f51\u5185\u591a\u7528\u6237\u8bbf\u95ee<\/li>\n
          • \u6269\u5c55\u652f\u6301\u56fe\u7247\u3001\u89c6\u9891\u7b49\u591a\u6a21\u6001\u77e5\u8bc6\u5e93\u68c0\u7d22<\/li>\n

            Jetson Nano \u867d\u7136 2019 \u5e74\u53d1\u5e03，\u7b97\u529b\u6709\u9650，\u4f46\u901a\u8fc7\u5408\u7406\u7684\u6a21\u578b\u9009\u578b\u548c\u4f18\u5316，\u5b8c\u5168\u53ef\u4ee5\u6ee1\u8db3\u8fb9\u7f18\u7aef\u8f7b\u91cf\u7ea7\u5927\u8bed\u8a00\u6a21\u578b\u5e94\u7528\u7684\u9700\u6c42，\u4e3a\u5d4c\u5165\u5f0f AI \u4ea7\u54c1\u7684\u667a\u80fd\u5316\u5347\u7ea7\u63d0\u4f9b\u4e86\u9ad8\u6027\u4ef7\u6bd4\u7684\u89e3\u51b3\u65b9\u6848\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

            \u4e00\u3001\u6280\u672f\u80cc\u666f1.1\u00a0Jetson Nano 8GB \u786c\u4ef6\u7279\u6027\u6839\u636e NVIDIA \u5b98\u65b9 datasheet，Jetson Nano 8GB \u642d\u8f7d\u56db\u6838 Cortex-A57 \u5904\u7406\u5668，\u4e3b\u9891 1.43GHz，\u5185\u7f6e 128 \u6838 Maxwell \u67b6\u6784 GPU，\u914d\u5907 8GB LPDDR4 \u5185\u5b58，\u529f\u8017\u4ec5 5-10W，\u662f\u8fb9\u7f18 AI \u90e8\u7f72\u7684\u9ad8\u6027\u4ef7\u6bd4\u5165\u95e8\u5e73\u53f0\u3002\u5c3d\u7ba1 FP16 \u7b97\u529b\u53ea\u6709 472 GFLOPS&#xff<\/p>\n","protected":false},"author":2,"featured_media":117944,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[13178,13389,868,66,4119,50],"topic":[],"class_list":["post-117948","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-1024---","tag-jetson-nano","tag-868","tag-ai","tag-rag","tag-50"],"yoast_head":"\n\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/117948.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u4e00\u3001\u6280\u672f\u80cc\u666f1.1\u00a0Jetson Nano 8GB \u786c\u4ef6\u7279\u6027\u6839\u636e NVIDIA \u5b98\u65b9 datasheet，Jetson Nano 8GB \u642d\u8f7d\u56db\u6838 Cortex-A57 \u5904\u7406\u5668，\u4e3b\u9891 1.43GHz，\u5185\u7f6e 128 \u6838 Maxwell \u67b6\u6784 GPU，\u914d\u5907 8GB LPDDR4 \u5185\u5b58，\u529f\u8017\u4ec5 5-10W，\u662f\u8fb9\u7f18 AI \u90e8\u7f72\u7684\u9ad8\u6027\u4ef7\u6bd4\u5165\u95e8\u5e73\u53f0\u3002\u5c3d\u7ba1 FP16 \u7b97\u529b\u53ea\u6709 472 GFLOPS&#xff\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/117948.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-25T16:17:20+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/10\/20261025161712-6ade2b88c0ce4.png\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/117948.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/117948.html\",\"name\":\"\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-10-25T16:17:20+00:00\",\"dateModified\":\"2026-10-25T16:17:20+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/117948.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/117948.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/117948.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/117948.html","og_locale":"zh_CN","og_type":"article","og_title":"\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u4e00\u3001\u6280\u672f\u80cc\u666f1.1\u00a0Jetson Nano 8GB \u786c\u4ef6\u7279\u6027\u6839\u636e NVIDIA \u5b98\u65b9 datasheet，Jetson Nano 8GB \u642d\u8f7d\u56db\u6838 Cortex-A57 \u5904\u7406\u5668，\u4e3b\u9891 1.43GHz，\u5185\u7f6e 128 \u6838 Maxwell \u67b6\u6784 GPU，\u914d\u5907 8GB LPDDR4 \u5185\u5b58，\u529f\u8017\u4ec5 5-10W，\u662f\u8fb9\u7f18 AI \u90e8\u7f72\u7684\u9ad8\u6027\u4ef7\u6bd4\u5165\u95e8\u5e73\u53f0\u3002\u5c3d\u7ba1 FP16 \u7b97\u529b\u53ea\u6709 472 GFLOPS&#xff","og_url":"https:\/\/www.wsisp.com\/helps\/117948.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-10-25T16:17:20+00:00","og_image":[{"url":"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/10\/20261025161712-6ade2b88c0ce4.png"}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"4 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/117948.html","url":"https:\/\/www.wsisp.com\/helps\/117948.html","name":"\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-10-25T16:17:20+00:00","dateModified":"2026-10-25T16:17:20+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/117948.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/117948.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/117948.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"\u5b9e\u6218\u6559\u7a0b\uff1aNVIDIA Jetson Nano 8GB \u642d\u5efa\u672c\u5730 RAG \u5927\u8bed\u8a00\u6a21\u578b\u77e5\u8bc6\u5e93\u5b8c\u6574\u6307\u5357"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/117948","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=117948"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/117948\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media\/117944"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=117948"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=117948"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=117948"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=117948"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}