{"id":53756,"date":"2025-02-16T09:00:26","date_gmt":"2025-02-16T01:00:26","guid":{"rendered":"https:\/\/fwq.ai\/blog\/53756\/"},"modified":"2025-02-16T09:00:26","modified_gmt":"2025-02-16T01:00:26","slug":"%e7%a8%8b%e5%ba%8f%e5%91%98ai%e6%b7%98%e9%87%91%e7%bb%bc%e5%90%88%e6%8c%87%e5%8d%97","status":"publish","type":"post","link":"https:\/\/fwq.ai\/blog\/53756\/","title":{"rendered":"\u7a0b\u5e8f\u5458AI\u6dd8\u91d1\u7efc\u5408\u6307\u5357"},"content":{"rendered":"

\u65b0\u4e00\u8f6e\u7f16\u7a0b\u6dd8\u91d1\u70ed\u6b63\u5728\u5982\u706b\u5982\u837c\u5730\u5c55\u5f00\u3002\u53ea\u662f\u8fd9\u4e00\u6b21\uff0c\u4f60\u4e0d\u9700\u8981\u6602\u8d35\u7684\u94f2\u5b50\uff0c\u4f60\u4f1a\u5f97\u5230\u514d\u8d39\u7684\u6316\u6398\u673a\u3002\u6211\u5728 90 \u5e74\u4ee3\u7ecf\u5386\u8fc7\u4e92\u8054\u7f51\u6dd8\u91d1\u70ed\uff0c\u5728 21 \u4e16\u7eaa\u521d\u7ecf\u5386\u8fc7\u79fb\u52a8\u6dd8\u91d1\u70ed\uff0c\u73b0\u5728\uff0c\u6211\u4eec\u7ec8\u4e8e\u8fce\u6765\u4e86 AI \u6dd8\u91d1\u70ed\u3002\u542c\u8d77\u6765\u6709\u4e9b\u5938\u5f20\uff0c\u4f46\u4e8b\u5b9e\u5e76\u975e\u5982\u6b64\u3002<\/p>\n

OpenAI\u3001Anthropic\u3001DeepSeek \u548c X \u7b49\u516c\u53f8\u90fd\u9700\u8981\u5927\u91cf\u7684 Nvidia GPU\u3002\u50cf\u6211\u8fd9\u6837\u7684\u72ec\u7acb\u5f00\u53d1\u8005\u6839\u672c\u8d1f\u62c5\u4e0d\u8d77\u6216\u6295\u8d44\u3002\u6b63\u5982\u4f60\u4ece\u6211\u7684\u6587\u7ae0\u4e2d\u4e86\u89e3\u5230\u7684\u90a3\u6837\uff0c\u53ea\u6709 10% \u7684\u7528\u6237\u53ef\u80fd\u4f1a\u5728\u514d\u8d39\u589e\u503c\u6a21\u5f0f\u4e0b\u4ed8\u8d39\u3002\u8fd9\u610f\u5473\u7740\u4f60\u5fc5\u987b\u4e3a 90% \u7684\u514d\u8d39\u7528\u6237\u4fdd\u6301 AI \u63a8\u7406\u80fd\u529b\u3002\u5bf9\u4e8e\u72ec\u7acb\u5f00\u53d1\u8005\u6765\u8bf4\u8fd9\u662f\u4e0d\u53ef\u80fd\u7684\u3002<\/p>\n

\u73b0\u5728\u60c5\u51b5\u5df2\u7ecf\u6539\u53d8\u3002\u4f7f\u7528 llama.cpp \u5e93\u7684 Ollama \u7b49\u8f6f\u4ef6\u4e0a\u7684\u672c\u5730 AI \u548c\u672c\u5730 LLM \u6b63\u5728\u5174\u8d77\u3002\u6587\u672c\u8f6c\u8bed\u97f3\u3001\u6587\u672c\u8f6c\u6587\u672c\u3001\u6587\u672c\u8f6c\u56fe\u50cf\u3001\u56fe\u50cf\u8f6c\u6587\u672c\u548c\u5927\u91cf\u5176\u4ed6\u6a21\u578b\u73b0\u5728\u53ef\u4ee5\u5728\u6700\u7ec8\u7528\u6237\u8bbe\u5907\u4e0a\u8fd0\u884c\uff0c\u751a\u81f3\u53ef\u4ee5\u5728 iPhone \u548c iPad \u4e0a\u8fd0\u884c\u3002\u6211\u6700\u8fd1\u521a\u521a\u63a8\u51fa\u4e86\u6211\u7684\uff0c\u8fd9\u662f\u4e00\u6b3e\u5373\u65f6 AI \u64ad\u5ba2\u751f\u6210\u5668\uff0c\u53ef\u5728 iOS \u4e0a 100% \u672c\u5730\u751f\u6210\u3002\u5b83\u4f7f\u7528 TinyLlama \u548c Kokoro \u6587\u672c\u8f6c\u8bed\u97f3\u6a21\u578b\u6765\u751f\u6210\u64ad\u5ba2\u3002\u6211\u8fd9\u8fb9\u4e0d\u9700\u8981\u57fa\u7840\u8bbe\u65bd\uff0c\u90fd\u662f BYOG\uff08\u201c\u81ea\u5e26 GPU\u201d\uff09\u3002BYOG \u662f\u72ec\u7acb\u5f00\u53d1\u8005\u8fdb\u5165 AI \u7684\u5927\u95e8\u3002<\/p>\n

1\u3001\u6dd8\u91d1\u6240\u9700\u7684\u5de5\u5177<\/h2>\n

\u8bf4\u6559\u5df2\u7ecf\u591f\u591a\u4e86\uff0c\u8ba9\u6211\u4eec\u6765\u770b\u770b\u6211\u4eec\u624b\u5934\u4e0a\u6709\u54ea\u4e9b\u5de5\u5177\u53ef\u4ee5\u52a0\u5165\u672c\u5730 AI \u6dd8\u91d1\u70ed\u3002\u6211\u4eec\u5c06\u91cd\u70b9\u4ecb\u7ecd\u5728 Windows\u3001Linux \u548c Mac \u4e0a\u63d0\u4f9b 100% \u672c\u5730\u79bb\u7ebf AI \u8f6f\u4ef6\u6240\u9700\u7684\u6240\u6709\u672c\u5730 AI \u5e93\u3002<\/p>\n

1.1 \u4f7f\u7528 llama.cpp \u8fdb\u884c LLM \u63a8\u7406<\/h3>\n

\u5982\u679c\u4f60\u53bb\u5e74\u6ca1\u6709\u4f4f\u5728\u5c71\u6d1e\u91cc\uff0c\u90a3\u4e48\u53ef\u80fd\u5df2\u7ecf\u5728\u4f60\u7684\u673a\u5668\u4e0a\u8fd0\u884c\u4e86\u51e0\u4e2a\u6708\u7684 Ollama\u3002Ollama \u662f\u4e00\u6b3e\u4f7f\u7528 llama.cpp \u7684\u8f6f\u4ef6\uff0cllama.cpp \u662f\u5f53\u4eca\u6bcf\u4e2a\u4eba\u90fd\u5728\u4f7f\u7528\u7684 C\/C++ \u7684 LLM \u63a8\u7406\u5e93\u3002Ollama \u5728 llama.cpp \u5468\u56f4\u505a\u4e86\u5f88\u591a\u7e41\u91cd\u7684\u5de5\u4f5c\uff0c\u5e76\u4e3a\u4efb\u4f55\u5f00\u59cb\u4f7f\u7528\u672c\u5730 LLM \u7684\u4eba\u63d0\u4f9b\u4e86\u76f8\u5f53\u53ef\u9760\u7684\u53ef\u7528\u6027\u3002<\/p>\n

\u5728 MacBook M1 Pro \u4e0a\u672c\u5730\u8fd0\u884c\u7684\u5fb7\u56fd\u201cTeuken\u201d\u6a21\u578b\u7684 Q6_K \u91cf\u5316 <\/p>\n

\u5f00\u53d1\u4eba\u5458\u5df2\u7ecf\u5728\u56f4\u7ed5 Ollama \u6784\u5efa\u5305\u88c5\u5668\u548c\u5ba2\u6237\u7aef\u8f6f\u4ef6\uff0c\u5e76\u8fde\u63a5\u5230\u5176\u670d\u52a1\u5668\u7684 API\u3002\u5b83\u672c\u8d28\u4e0a\u4e0e\u4f7f\u7528 OpenAI\u3001Anthropic\u3001Gemini \u6216 AWS Bedrock \u7684 API \u76f8\u540c\uff0c\u53ea\u662f\u4f60\u81ea\u5df1\u6258\u7ba1\u6a21\u578b\u5e76\u9700\u8981\u7ef4\u62a4 Ollama \u670d\u52a1\u5668\u3002<\/p>\n

\u6709\u8da3\u7684\u90e8\u5206\u6765\u81ea llama.cpp \u672c\u8eab\u3002\u5982\u679c\u4f60\u60f3\u53d1\u5e03\u672c\u5730 AI \u8f6f\u4ef6\uff0c\u5c31\u50cf\u6211\u5bf9 Botcast \u6240\u505a\u7684\u90a3\u6837\uff0c\u9700\u8981\u5c06 llama.cpp \u6253\u5305\u5230\u4f60\u7684\u5e94\u7528\u7a0b\u5e8f\u4e2d\u3002\u8fd9\u5e76\u4e0d\u662f\u592a\u5927\u7684\u6311\u6218\u3002 \u5df2\u7ecf\u88ab\u5927\u91cf\u5e94\u7528\u7a0b\u5e8f\u8bc1\u660e\uff0c\u5305\u62ec\u6211\u7684\u5e94\u7528\u7a0b\u5e8f\uff0c\u5e76\u4e14\u8fd8\u6709 Go\uff08\uff09\u3001Rust\uff08\uff09\u6216 C#\uff08\uff09\u7684\u7ed1\u5b9a\u3002\u901a\u8fc7\u8fd9\u4e9b\u7ed1\u5b9a\uff0c\u4f60\u53ef\u4ee5\u8f7b\u677e\u6784\u5efa PHP \u6a21\u5757\u3001iOS \u6846\u67b6\u6216 Node.js \u5305\u3002<\/p>\n

\/* example Llama chat session in C# executing a local model *\/\nusing LLama.Common;\nusing LLama;\n\nstring modelPath = @\"<Your Model Path>\"; \/\/ change it to your own model path.\n\nvar parameters = new ModelParams(modelPath)\n{\n    ContextSize = 1024, \/\/ The longest length of chat as memory.\n    GpuLayerCount = 5 \/\/ How many layers to offload to GPU. Please adjust it according to your GPU memory.\n};\nusing var model = LLamaWeights.LoadFromFile(parameters);\nusing var context = model.CreateContext(parameters);\nvar executor = new InteractiveExecutor(context);\n\n\/\/ Add chat histories as prompt to tell AI how to act.\nvar chatHistory = new ChatHistory();\nchatHistory.AddMessage(AuthorRole.System, \"Transcript of a dialog, where the User interacts with an Assistant named Bob. Bob is helpful, kind, honest, good at writing, and never fails to answer the User's requests immediately and with precision.\");\nchatHistory.AddMessage(AuthorRole.User, \"Hello, Bob.\");\nchatHistory.AddMessage(AuthorRole.Assistant, \"Hello. How may I help you today?\");\n\nChatSession session = new(executor, chatHistory);\n\nInferenceParams inferenceParams = new InferenceParams()\n{\n    MaxTokens = 256, \/\/ No more than 256 tokens should appear in answer. Remove it if antiprompt is enough for control.\n    AntiPrompts = new List<string> { \"User:\" }, \/\/ Stop generation once antiprompts appear.\n\n    SamplingPipeline = new DefaultSamplingPipeline(),\n};\n\nConsole.ForegroundColor = ConsoleColor.Yellow;\nConsole.Write(\"The chat session has started.\\nUser: \");\nConsole.ForegroundColor = ConsoleColor.Green;\nstring userInput = Console.ReadLine() ?? \"\";\n\nwhile (userInput != \"exit\")\n{\n    await foreach ( \/\/ Generate the response streamingly.\n        var text\n        in session.ChatAsync(\n            new ChatHistory.Message(AuthorRole.User, userInput),\n            inferenceParams))\n    {\n        Console.ForegroundColor = ConsoleColor.White;\n        Console.Write(text);\n    }\n    Console.ForegroundColor = ConsoleColor.Green;\n    userInput = Console.ReadLine() ?? \"\";\n}<\/code><\/pre>\n
\u6211\u5bf9\u53ef\u9884\u89c1\u7684\u672a\u6765\u7684\u9884\u6d4b\uff1f\u51e0\u4e4e\u6bcf\u4e2a\u5e94\u7528\u7a0b\u5e8f\u90fd\u4f1a\u9644\u5e26 llama.cpp\u3002GGUF \u6587\u4ef6\u683c\u5f0f\u5df2\u7ecf\u6210\u4e3a LLM \u7684\u4e8b\u5b9e\u6807\u51c6\u3002\u672a\u6765\u7684\u64cd\u4f5c\u7cfb\u7edf\u5f88\u53ef\u80fd\u5185\u7f6e llama.cpp\u3002\u4e0e\u6b64\u540c\u65f6\uff0c\u4f60\u7684\u5e94\u7528\u7a0b\u5e8f\u9700\u8981\u643a\u5e26 llama.cpp \u5e76\u6346\u7ed1\u60f3\u8981\u7684\u6a21\u578b\uff08\u5982 \u6216\u5176\u5fb7\u8bed\u8868\u4eb2 \uff0c\u82f1\u6587\u610f\u601d\u662f\u201c\u5c0f\u7f8a\u7f94\u201d\uff09\u3002<\/p>\n
\u5c06\u81ea\u5df1\u7684 LLM \u6346\u7ed1\u5230\u4f60\u7684\u5e94\u7528\u7a0b\u5e8f\u4e2d\u5c06\u4e3a\u5176\u63d0\u4f9b\u65e0\u9650\u7684\u3001\u4e0d\u53d7\u9650\u5236\u7684\u79bb\u7ebf LLM \u529f\u80fd\u3002\u65e0\u9700 API \u4ee4\u724c\uff0c\u65e0\u9700\u8ba2\u9605\u3002\u53ea\u9700\u5c06 TinyLlama \u6216 Phi \u4e0e\u4f60\u7684\u5e94\u7528\u7a0b\u5e8f\u4e00\u8d77\u643a\u5e26\u5373\u53ef\u3002MacBook \u7528\u6237\u5df2\u7ecf\u62e5\u6709\u5f3a\u5927\u7684\u786c\u4ef6\uff0c\u6267\u884c\u8fd9\u4e9b\u6beb\u65e0\u56f0\u96be\u3002\u5bf9\u4e8e Windows \u548c Linux \u7528\u6237\uff1f\u4ed6\u4eec\u53ef\u4ee5\u770b\u7740\u4ed6\u4eec\u7684 CPU \u88ab\u627c\u6740\u800c\u5165\u7761\uff0c\u6216\u8005\u53ea\u9700\u83b7\u5f97\u5408\u9002\u7684 Nvidia GPU\u3002\u65e0\u8bba\u4f60\u6b63\u5728\u6784\u5efa\u4ec0\u4e48\u5e94\u7528\u7a0b\u5e8f\uff0c\u90fd\u53ef\u4ee5\u5c06 LLM \u878d\u5165\u5176\u4e2d\uff0c\u5e76\u5411 OpenAI \u5305\u88c5\u5668\u6784\u5efa\u6280\u672f\u5144\u5f1f\u5c55\u793a\u8c01\u662f\u771f\u6b63\u7684 AI Gigachad\uff0c\u540c\u65f6\u5c06\u5176\u4ef7\u683c\u964d\u4f4e 99.999%\u3002<\/p>\n
1.2 \u4f7f\u7528 Stable Diffusion \u751f\u6210\u56fe\u50cf<\/h3>\n
\u9700\u8981\u4e00\u5f20\u6298\u7eb8\u4e2d\u7684\u201c\u9e1f\u5728\u793e\u4ea4\u5a92\u4f53\u4e0a\u53d1\u5e16\u65f6\u559d\u5496\u5561\u201d\u7684\u56fe\u50cf\u5417\uff1f\u8fd9\u5c31\u662f\u53ca\u5176\u4f17\u591a\u884d\u751f\u4ea7\u54c1\u7684\u7528\u9014\u3002 \u548c\u76ee\u524d\u662f\u6700\u53d7\u6b22\u8fce\u7684\u3002\u8fd9\u4e9b\u6a21\u578b\u7684\u5927\u5c0f\u7ea6\u4e3a 8GB\uff0c\u56e0\u6b64\u4f60\u9700\u8981\u5728 iPhone \u4e0a\u5c0f\u5fc3\u4e00\u70b9\u3002\u867d\u7136\u5df2\u7ecf\u6709\u5b9e\u73b0\uff0c\u4f46\u5b83\u4eec\u9700\u8981\u5f3a\u5927\u7684 iPhone 15 Pro \u6216 M1 iPad\u3002iPhone SE \u4ec0\u4e48\u4e5f\u505a\u4e0d\u4e86\uff08\u4f46\u5b83\u53ef\u4ee5\u50cf TinyLlama \u4e00\u6837\u6267\u884c LLM\uff09\u3002<\/p>\n
 \u4f7f\u7528 RealVis \u5728 macOS \u4e0a\u4f7f\u7528 Swift \u548c Xcode \u7684\u7b80\u5355 Stable Diffusion XL \u7ba1\u9053 <\/p>\n
Apple \u81ea\u5df1\u5728 Github \u4e0a\u53d1\u5e03\u4e86\u3002\u8be5\u5e93\u4f7f\u5728 macOS \u548c iPad \u4e0a\u7684\u63a8\u7406\u53d8\u5f97\u975e\u5e38\u7b80\u5355\u3002\u4f60\u6240\u9700\u8981\u7684\u53ea\u662f\u6a21\u578b\u7684\u5185\u5bb9\uff0c\u53ef\u4ee5\u5c06\u5176\u6346\u7ed1\u5230\u4f60\u7684\u5e94\u7528\u7a0b\u5e8f\u4e2d\u6216\u4ece Huggingface \u4e0b\u8f7d\uff0c\u7136\u540e\u4f7f\u7528\u63d0\u4f9b\u7684\u5e93\u6267\u884c\u5b83\u3002<\/p>\n
\/\/\/ initializes the pipeline on launch of the app\nfunc initPipeline() async throws {\n    writeStatus(text: \"Initializing model configuration\")\n    let configuration = MLModelConfiguration()\n    configuration.computeUnits = .cpuAndGPU\n    \n    writeStatus(text: \"Instanciating pipeline with model...\")\n    let resourceUrl = URL(fileURLWithPath: ContentView.modelPath)\n    self.pipeline = try StableDiffusionPipeline(resourcesAt: resourceUrl,\n                           controlNet: [],\n                           configuration: configuration,\n                           disableSafety: false,\n                           reduceMemory: false)\n    \n    writeStatus(text: \"Pipeline is now loading resources...\")\n    try pipeline!.loadResources()\n    \n    writeStatus(text: \"Pipeline ready\")\n}\n    \n\/\/\/ performs image generation with a prompt\nfunc generateImage(){\n    writeStatus(text: \"Configuring pipeline for prompt\")\n    \n    var pipelineConfig = StableDiffusionPipeline.Configuration\n                                              (prompt: promptText)\n    pipelineConfig.negativePrompt = \"\"\n    pipelineConfig.seed = UInt32.random(in: (0..<UInt32.max))\n    pipelineConfig.guidanceScale = 3\n    pipelineConfig.stepCount = 25\n    pipelineConfig.imageCount = 1\n    \n    pipelineConfig.originalSize = 1024\n    pipelineConfig.targetSize = 1024\n    \n    writeStatus(text: \"Pipeline executing prompt...\")\n    let result = try! pipeline!.generateImages\n                      (configuration: pipelineConfig)\n\n    if result.count > 0 {\n        writeStatus(text: \"Ready, assigning result\")\n        generatedImage = result[0]\n    } else {\n        writeStatus(text: \"Pipeline error: no image returned\")\n    }\n}<\/code><\/pre>\n
\u9884\u8ba1\u5728\u4e0d\u4e45\u7684\u5c06\u6765\u4f1a\u51fa\u73b0\u5fae\u5c0f\u7684\u56fe\u50cf\u751f\u6210\u6a21\u578b\u3002\u56fe\u50cf\u751f\u6210\u5668\u7684\u7528\u4f8b\u51e0\u4e4e\u662f\u65e0\u7a77\u65e0\u5c3d\u7684\u3002\u60f3\u8c61\u4e00\u4e0b\u8f6f\u4ef6\u7ecf\u5e38\u9644\u5e26\u7684\u6240\u6709\u9ed8\u8ba4\u56fe\u50cf\uff1f\u662f\u7684\uff0c\u4f60\u53ef\u4ee5\u5728\u5c06\u6765\u8df3\u8fc7\u8fd9\u4e9b\u6b65\u9aa4\uff0c\u53ea\u9700\u4f7f\u7528\u7ed9\u5b9a\u7684\u7528\u6237\u8f93\u5165\u751f\u6210\u5b83\u4eec\u5373\u53ef\u3002<\/p>\n
1.3 \u4f7f\u7528 sherpa-onnx \u8fdb\u884c\u6587\u672c\u8f6c\u8bed\u97f3<\/h3>\n
\u6587\u672c\u8f6c\u8bed\u97f3\u6216\u8bed\u97f3\u5408\u6210\u65e9\u5728 1984 \u5e74\u5c31\u5df2\u5728 MacSpeak \u4e2d\u51fa\u73b0\uff0c\u542c\u8d77\u6765\u5c31\u50cf\u51b0\u7bb1\u95e8\u574f\u4e86\u53d1\u51fa\u7684\u5431\u5431\u58f0\u3002\u4ece\u90a3\u65f6\u8d77\uff0cTTS \u5df2\u7ecf\u53d6\u5f97\u4e86\u957f\u8db3\u7684\u8fdb\u6b65\uff0c\u5982\u4eca\u7684\u58f0\u97f3\u542c\u8d77\u6765\u975e\u5e38\u81ea\u7136\u3002 \u5e93\u652f\u6301\u5404\u79cd\u6a21\u578b\uff0c\u5305\u62ec\u4ee4\u4eba\u5370\u8c61\u6df1\u523b\u7684 Kokoro-82M \u6a21\u578b\u3002\u8bed\u97f3\u8bc6\u522b\u5df2\u5185\u7f6e\u5728\u5927\u591a\u6570\u64cd\u4f5c\u7cfb\u7edf\u4e2d\uff0c\u5305\u62ec Windows\u3001Linux\u3001macOS\u3001iOS \u548c Android\u3002<\/p>\n
 \u6211\u7684 Botcast \u5e94\u7528\u7a0b\u5e8f\u4f7f\u7528 sherpa-onnx \u548c Kokoro \u521b\u5efa\u6574\u4e2a\u64ad\u5ba2 <\/p>\n
TTS \u548c\u8bed\u97f3\u8bc6\u522b\u5141\u8bb8\u5bf9\u4f60\u7684\u5e94\u7528\u7a0b\u5e8f\u8fdb\u884c\u5b8c\u5168\u4ea4\u4e92\u5f0f\u8bed\u97f3\u63a7\u5236\u3002\u7ed3\u5408 LLM\uff0c\u5bf9\u5938\u5f20\u7684 GUI \u7684\u9700\u6c42\u53d8\u5f97\u975e\u5e38\u503c\u5f97\u6000\u7591\u3002\u5728\u6240\u6709\u652f\u6301\u7684\u5e73\u53f0\u4e0a\uff0c\u4f7f\u7528\u8fd9\u4e9b\u6a21\u578b\u548c sherpa-onnx \u5e93\u90fd\u975e\u5e38\u7b80\u5355\u3002<\/p>\n
\/* instanciate the kokoro tts model *\/\nlet kokoro = sherpaOnnxOfflineTtsKokoroModelConfig(\n    model: model.path(),\n    voices: voices.path(),\n    tokens: tokens.path(),\n    dataDir: dataDir\n)\n\n\/* setup the serpa onnx offline tts wrapper *\/\nlet modelConfig = sherpaOnnxOfflineTtsModelConfig(kokoro: kokoro, debug: 0)\nvar ttsConfig = sherpaOnnxOfflineTtsConfig(model: modelConfig)\nself.tts = SherpaOnnxOfflineTtsWrapper(config: &ttsConfig)\n\n\/* generate the audio for the provided text *\/\nlet text = \"Just say it out loud, no need to click\"\nlet speed: Float = 1.0\nlet audio = self.tts?.generate(text: text, sid: Int(voiceId), speed: speed)<\/code><\/pre>\n
\u662f\u6211\u4f7f\u7528 Botcast \u5e94\u7528\u751f\u6210\u7684\u64ad\u5ba2\u3002\u5b83\u5c55\u793a\u4e86 TTS \u5982\u4f55\u751f\u6210 10 \u5206\u949f\u7684\u8bed\u97f3\u97f3\u9891\u5185\u5bb9\u3002<\/p>\n
\u4f20\u7edf\u7684\u8f6f\u4ef6\u8bbe\u8ba1\u4f9d\u8d56\u4e8e\u5e26\u6709\u6587\u672c\u548c\u6309\u94ae\u7684 GUI\u3002TTS\u3001\u8bed\u97f3\u8bc6\u522b\u548c LLM \u7684\u51fa\u73b0\u5f7b\u5e95\u98a0\u8986\u4e86\u8fd9\u4e00\u70b9\u3002\u4efb\u4f55\u4ee5\u524d\u5728\u5e94\u7528\u8bbe\u8ba1\u4e0a\u5e94\u7528\u7684\u89c4\u5219\u90fd\u4e0d\u518d\u9002\u7528\u3002<\/p>\n
1.4 \u4f7f\u7528 Create ML \u81ea\u5b9a\u4e49\u6a21\u578b<\/h3>\n
\u8fd8\u6709\u5927\u91cf\u5176\u4ed6\u6a21\u578b\uff0c\u4ece\u97f3\u4e50\u751f\u6210\u5668\u5230\u56fe\u50cf\u5206\u7c7b\u5668\u3002\u6211\u751a\u81f3\u8fd8\u6ca1\u6709\u4ecb\u7ecd\u8ba1\u7b97\u673a\u89c6\u89c9\u548c\u56fe\u50cf\u5206\u7c7b\uff0c\u56e0\u4e3a\u8fd9\u5df2\u7ecf\u6210\u4e3a\u5e38\u6001\u3002\u6211\u65e0\u6cd5\u6df1\u5165\u7814\u7a76\u6240\u6709\u7684\u4eba\u5de5\u667a\u80fd\u548c\u673a\u5668\u5b66\u4e60\u6a21\u578b\u3002\u8fd9\u4f1a\u8ba9\u8fd9\u7bc7\u6587\u7ae0\u53d8\u6210\u4e00\u672c\u5b8c\u6574\u7684\u4e66\uff0c\u800c\u8fd9\u672c\u4e66\u4e00\u51fa\u7248\u5c31\u4f1a\u8fc7\u65f6\u3002<\/p>\n
 \u82f9\u679c\u7684 Create ML \u7528\u4e8e\u7b80\u5355\u7684\u6a21\u578b\u8bad\u7ec3 <\/p>\n
\u5bf9\u4e8e iOS\u3001iPhone\u3001iPad \u548c macOS \u7528\u6237\u6765\u8bf4\uff0c\u4e00\u4e2a\u91cd\u8981\u7684\u5de5\u5177\u662f\u82f9\u679c\u81ea\u5df1\u7684 \u3002\u4e0e TensorFlow \u4ee5\u53ca AWS\u3001Google Cloud \u548c Azure \u4e0a\u7684\u6240\u6709\u673a\u5668\u5b66\u4e60\u5de5\u5177\u76f8\u6bd4\uff0c\u5b83\u770b\u8d77\u6765\u5c31\u50cf\u4e00\u4e2a\u73a9\u5177\u3002\u5c24\u5176\u662f\u7528 Python\u3002\u8fd9\u5f88\u597d\uff0c\u56e0\u4e3a\u8fd9\u4e9b\u6a21\u578b\u53ef\u4ee5\u4f7f\u7528\u82f9\u679c\u7684 \u8f6c\u6362\u4e3a CoreML\u3002<\/p>\n
\u4f46\u8fd9\u4e0d\u662f Create ML \u7684\u91cd\u70b9\u3002\u5b83\u770b\u8d77\u6765\u50cf\u4e00\u4e2a\u73a9\u5177\uff0c\u56e0\u4e3a\u5b83\u7684\u76ee\u6807\u6b63\u662f\uff1a\u4e00\u4e2a\u7528\u4e8e\u76f8\u5bf9\u57fa\u672c\u7684\u673a\u5668\u5b66\u4e60\u6a21\u578b\u7684\u6781\u5176\u7b80\u5355\u7684\u8bad\u7ec3\u5e94\u7528\u7a0b\u5e8f\u3002\u4e3a\u4ec0\u4e48\uff1f\u56e0\u4e3a\u4f60\u5230\u5904\u90fd\u9700\u8981\u5b83\u4eec\u3002\u770b\u770b Create ML \u63d0\u4f9b\u7684\u6a21\u677f\uff0c\u8fd9\u662f\u65e5\u5e38\u4efb\u52a1\u3002<\/p>\n
\u4e0b\u6b21\u4f60\u6784\u5efa\u4e00\u4e2a\u7528\u6237\u53ef\u4ee5\u5bf9\u9879\u76ee\u8fdb\u884c\u5206\u7c7b\u7684 iOS \u5e94\u7528\u65f6\uff0c\u53ea\u9700\u4f7f\u7528\u6587\u672c\u5206\u7c7b\u5668\u5e76\u81ea\u52a8\u5bf9\u5176\u8fdb\u884c\u5206\u7c7b\uff0c\u6216\u8005\u81f3\u5c11\u6839\u636e ML \u6587\u672c\u5206\u7c7b\u5668\u63d0\u51fa\u5efa\u8bae\u3002\u5728 Xcode \u4e2d\u4f7f\u7528 Create ML \u4e2d\u7684\u6a21\u578b\u5f88\u5bb9\u6613\uff1a\u53ea\u9700\u62d6\u653e\u6a21\u578b\u5305\u6587\u4ef6\u5373\u53ef\u3002<\/p>\n
2\u3001\u4e3a\u4ec0\u4e48\u4f1a\u6709\u6dd8\u91d1\u70ed<\/h2>\n
\u4eba\u5de5\u667a\u80fd\u4e0d\u4ec5\u6539\u53d8\u4e86\u6211\u4eec\u7f16\u5199\u8f6f\u4ef6\u7684\u65b9\u5f0f\uff0c\u8fd8\u6539\u53d8\u4e86\u8f6f\u4ef6\u672c\u8eab\u3002\u957f\u4e0b\u62c9\u5217\u8868\uff1f\u4e0d\uff0c\u6587\u672c\u5206\u7c7b\u5668\u3002\u62cd\u6444\u4ea7\u54c1\u5e76\u624b\u52a8\u8f93\u5165\u4ea7\u54c1\u8be6\u7ec6\u4fe1\u606f\uff1f\u5f53\u7136\u4e0d\u662f\uff0c\u90a3\u5c06\u662f\u4e00\u4e2a\u56fe\u50cf\u5206\u7c7b\u5668\u3002\u60f3\u5728\u6267\u884c\u67d0\u9879\u64cd\u4f5c\u65f6\u901a\u77e5\u7528\u6237\uff1f\u5fd8\u8bb0\u8ba1\u65f6\u5668\u5427\uff0c\u90a3\u5c06\u662f\u4e00\u4e2a\u9884\u6d4b\u65f6\u95f4\u7684\u56de\u5f52\u6a21\u578b\u3002<\/p>\n
 \u6211\u7684\u81ea\u5b9a\u4e49 Create ML \u6a21\u578b\uff0c\u5b83\u4e3a\u7ed9\u5b9a\u6807\u9898\u5b9a\u4e49\u64ad\u5ba2\u80cc\u666f\u97f3\u4e50\u6d41\u6d3e <\/p>\n
\u6709\u4e9b\u7528\u4f8b\u751a\u81f3\u66f4\u7b80\u5355\uff1a\u6211\u7684 Botcast \u5e94\u7528\u9644\u5e26 16 \u4e2a\u80cc\u666f\u97f3\u4e50\u6587\u4ef6\uff0c\u7528\u4e8e\u751f\u6210\u7684\u64ad\u5ba2\u3002\u5b83\u53ef\u4ee5\u968f\u673a\u9009\u62e9\u4e00\u4e2a\uff0c\u8fd9\u4f1a\u5f88\u5c34\u5c2c\uff0c\u56e0\u4e3a\u5b83\u4e0e\u5185\u5bb9\u4e0d\u5339\u914d\u3002\u76f8\u53cd\uff0c\u5b83\u4f7f\u7528\u6765\u81ea Create ML \u7684\u81ea\u5b9a\u4e49\u8bad\u7ec3\u6a21\u578b\u3002\u8be5\u6a21\u578b\u672c\u8eab\u975e\u5e38\u5c0f\uff0c\u53ea\u6709 73 KB\u3002\u63a8\u7406\uff08\u6267\u884c\uff09\u4ec5\u9700\u51e0\u6beb\u79d2\u5373\u53ef\u5b8c\u6210\u3002\u8bad\u7ec3\u4ec5\u9700 5 \u5206\u949f\u3002\u8fd9\u6bd4\u6b63\u5219\u8868\u8fbe\u5f0f\u9b54\u6cd5\u8981\u597d\u5f97\u591a\u3002<\/p>\n
\/\/\/ returns the background music file to use for this podcast\nfunc getBackgroundMusicFile() -> URL? {\n    var musicGenre = self.defaultBackgroundMusic\n    \n    do {\n        let genreOuput = try self.musicClassifier?.prediction\n                                              (text: self.title)\n        if genreOuput != nil {\n            musicGenre = genreOuput?.label ?? self.defaultBackgroundMusic\n        }\n    } catch {\n        print(\"Failed to predict background music: fallback to default\")\n        musicGenre = self.defaultBackgroundMusic\n    }\n    \n    let fileIndex = Int.random(in: 1...2)\n    let audioFileUrl = Bundle.main.url(\n                forResource: \"\\(musicGenre)\\(fileIndex)\", \n                withExtension: \"mp3\"\n    )\n\n    if audioFileUrl != nil {\n        return audioFileUrl\n    }\n    \n    return nil\n}<\/code><\/pre>\n
\u4f7f\u7528 Xcode \u81ea\u52a8\u4e3a mlmodel \u6587\u4ef6\u751f\u6210\u5305\u88c5\u5668\u7c7b\uff0c\u6267\u884c\u6a21\u578b\u672c\u8eab\u53ea\u9700 3 \u884c\u975e\u5e38\u57fa\u672c\u7684\u4ee3\u7801\u3002\u73b0\u5728\uff0c\u4f5c\u4e3a\u5f00\u53d1\u4eba\u5458\uff0c\u6211\u4eec\u771f\u7684\u9700\u8981\u91cd\u65b0\u8003\u8651\u6211\u4eec\u7684\u65b9\u6cd5\u3002\u6b63\u5219\u8868\u8fbe\u5f0f\u548c\u968f\u673a\u5316\u5668\u4ee5\u53ca\u5927\u578b\u8868\u5355\u519c\u573a\u548c\u9884\u9009\u4e0d\u518d\u53ef\u884c\u3002\u5373\u4f7f\u662f\u51e0\u767e KB \u7684\u57fa\u672c\u6a21\u578b\u4e5f\u53ef\u4ee5\u7acb\u5373\u9884\u6d4b\u503c\u3002<\/p>\n
3\u3001ML \u8fdb\u5165\u5e94\u7528\u7684\u540e\u679c<\/h2>\n
\u5728\u4ee5\u524d\u4f7f\u7528\u590d\u6742\u6b63\u5219\u8868\u8fbe\u5f0f\u6216\u968f\u673a\u6570\u751f\u6210\u5668\u7684\u573a\u666f\u4e2d\uff0c\u73b0\u5728\u5b83\u5c5e\u4e8e ML \u6a21\u578b\u7684\u9886\u57df\u3002\u5176\u7ed3\u679c\u662f\uff0c\u4efb\u4f55\u73b0\u6709\u7684\u8f6f\u4ef6\u90fd\u9700\u8981\u5f7b\u5e95\u6539\u9769\u3002\u5c31\u50cf\u6280\u672f\u5144\u5f1f\u4eec\u51e0\u5e74\u524d\u8c08\u8bba\u201c\u79fb\u52a8\u4f18\u5148\u201d\u548c\u201cAPI \u4f18\u5148\u201d\u65b9\u6cd5\u4e00\u6837\uff0c\u6211\u4eec\u73b0\u5728\u9700\u8981\u8003\u8651\u201cAI \u4f18\u5148\u201d\u65b9\u6cd5\u3002<\/p>\n
\u6211\u7684\u610f\u601d\u662f\u4ec0\u4e48\uff1f\u5047\u8bbe\u4f60\u7684\u5e94\u7528\u6709 5 \u4e2a\u6838\u5fc3\u529f\u80fd\uff0c\u7528\u6237\u5c06\u6839\u636e\u4ed6\u5f53\u524d\u7684\u60c5\u51b5\u548c\u4e4b\u524d\u7684\u4efb\u52a1\u9700\u8981\u8fd9\u4e9b\u529f\u80fd\u3002\u65e0\u9700\u7f16\u5199\u590d\u6742\u7684\u6761\u4ef6\u6765\u786e\u5b9a\u4f55\u65f6\u5728\u4f55\u5904\u5448\u73b0\u8fd9\u4e9b\u529f\u80fd\uff0c\u53ea\u9700\u4f7f\u7528\u63a8\u8350\u6a21\u578b\u5373\u53ef\u3002\u63a8\u8350\u5668\u4e0d\u4ec5\u5185\u7f6e\u4e8e Core ML\uff0c\u8fd8\u5185\u7f6e\u4e8e\u8bb8\u591a\u5176\u4ed6\u8bed\u8a00\u7684\u5e93\u4e2d\u5b58\u5728\u3002\u8bf7\u67e5\u770b\u5fae\u8f6f\u7684 \u7684C# \u7248\u4f5c\u4e3a\u793a\u4f8b\u4e4b\u4e00\u3002<\/p>\n
* Recommender engine output from Microsoft.ML with C# and .NET *\n\n=============== Training the model ===============\niter      tr_rmse          obj\n   0       1.5382   3.1213e+05\n   1       0.9223   1.6051e+05\n   2       0.8691   1.5050e+05\n   3       0.8413   1.4576e+05\n   4       0.8145   1.4208e+05\n   5       0.7848   1.3895e+05\n   6       0.7552   1.3613e+05\n   7       0.7259   1.3357e+05\n   8       0.6987   1.3121e+05\n   9       0.6747   1.2949e+05\n  10       0.6533   1.2766e+05\n  11       0.6353   1.2636e+05\n  12       0.6209   1.2561e+05\n  13       0.6072   1.2462e+05\n  14       0.5965   1.2394e+05\n  15       0.5868   1.2352e+05\n  16       0.5782   1.2279e+05\n  17       0.5713   1.2227e+05\n  18       0.5637   1.2190e+05\n  19       0.5604   1.2178e+05\n=============== Evaluating the model ===============\nRms: 0.977175077487166\nRSquared: 0.43233349213192\n=============== Making a prediction ===============\nMovie 10 is recommended for user 6\n=============== Saving the model to a file ===============<\/code><\/pre>\n
\u5c06 ML \u6a21\u578b\u96c6\u6210\u5230\u5e94\u7528\u7a0b\u5e8f\u4e2d\u5e76\u4e0d\u662f\u4e00\u4e2a\u81ea\u6211\u5b9e\u73b0\u7684\u9884\u8a00\u3002\u5b83\u901a\u8fc7\u505a\u51fa\u5408\u7406\u4e14\u6709\u7528\u7684\u9884\u6d4b\u6765\u8282\u7701\u7528\u6237\u7684\u65f6\u95f4\uff0c\u5426\u5219\u4ed6\u4eec\u5fc5\u987b\u624b\u52a8\u51b3\u5b9a\u6216\u8ba1\u7b97\u3002\u8fd9\u5c31\u662f\u6dd8\u91d1\u70ed\u51fa\u73b0\u7684\u539f\u56e0\uff1a\u516c\u53f8\u613f\u610f\u4e3a\u73b0\u4ee3\u673a\u5668\u5b66\u4e60\u548cAI\u9a71\u52a8\u7684\u8f6f\u4ef6\u4ed8\u8d39\uff0c\u56e0\u4e3a\u8fd9\u53ef\u4ee5\u4e3a\u4ed6\u4eec\u8282\u7701\u6570\u5343\u5c0f\u65f6\u7684\u624b\u52a8\u5de5\u4f5c\u3002<\/p>\n
4\u3001\u201c\u6740\u624b\u7ea7\u529f\u80fd\u201d\u7684\u65f6\u4ee3\u5df2\u7ecf\u7ed3\u675f<\/h2>\n
\u8f6f\u4ef6\u5f00\u53d1\u8fc7\u53bb\u975e\u5e38\u6ce8\u91cd\u529f\u80fd\u3002\u5b83\u4e13\u6ce8\u4e8e GUI \u548c\u5c06\u6570\u636e\u63d2\u5165 GUI \u7684\u4eba\uff0c\u65e0\u8bba\u662f\u7f51\u7edc\u3001\u79fb\u52a8\u8fd8\u662f\u684c\u9762\u3002\u8be5\u8f6f\u4ef6\u5c06\u5305\u542b\u4e00\u4e2a\u201c\u795e\u5947\u7b97\u6cd5\u201d\uff0c\u8fd9\u662f\u7528\u6237\u8d2d\u4e70\u6216\u4f7f\u7528\u8be5\u8f6f\u4ef6\u7684\u4e3b\u8981\u539f\u56e0\u3002<\/p>\n
\u73b0\u5728\u8fd9\u5df2\u7ecf\u7ed3\u675f\u4e86\uff0c\u6211\u4eec\u53c8\u56de\u5230\u4e86\u201c\u54e6\uff0c\u5b83\u53ef\u4ee5\u505a\u5230\u5417\uff1f\u201d\u65b0\u6280\u672f\u7684\u9636\u6bb5\u3002\u7528\u6237\u4e0d\u77e5\u9053\u4ec0\u4e48\u662f\u53ef\u80fd\u7684\u3002\u56e0\u6b64\uff0c\u65b9\u6cd5\u9700\u8981\u5b8c\u5168\u7406\u89e3\u76ee\u6807\u7528\u6237\u7fa4\u7684\u8fc7\u7a0b\u3002\u4e00\u65e6\u5b8c\u5168\u7406\u89e3\uff0c\u8f6f\u4ef6\u5c31\u4f1a\u81ea\u52a8\u5316\u5230\u7edd\u5bf9\u6781\u9650\uff0c\u53ea\u5141\u8bb8\u7528\u6237\u9a8c\u8bc1\u3001\u8c03\u6574\u548c\u6279\u51c6\u7ed3\u679c\u3002<\/p>\n
\u8ba9\u6211\u4eec\u7528\u4e00\u4e2a\u5b9e\u9645\u7684\u4f8b\u5b50\u6765\u6982\u8ff0\u8fd9\u4e00\u70b9\uff1a\u5047\u8bbe\u4f60\u6709\u4e00\u4e2a\u56fe\u4e66\u9986\u7684\u5e93\u5b58\u7ba1\u7406\u8f6f\u4ef6\uff0c\u672c\u8d28\u4e0a\u662f\u4e00\u4e2a\u201c\u56fe\u4e66\u6570\u636e\u5e93\u201d\u3002\u5728\u673a\u5668\u5b66\u4e60\u65f6\u4ee3\u4e4b\u524d\uff0c\u8f6f\u4ef6\u5141\u8bb8\u7528\u6237\u62cd\u6444\u4e66\u7c4d\u3001\u626b\u63cf ISBN\u3001\u4ece API \u83b7\u53d6\u4e00\u4e9b\u8be6\u7ec6\u4fe1\u606f\uff08\u4f8b\u5982\u4e66\u540d\u3001\u4f5c\u8005\u7b49\uff09\u5e76\u5b9a\u4e49\u4e66\u7c4d\u526f\u672c\u7684\u72b6\u51b5\u548c\u5e74\u4ee3\u3002\u6211\u4eec\u4e0d\u518d\u8fd9\u6837\u505a\u4e86\u3002<\/p>\n
\u5728\u5f53\u4eca\u7684\u4eba\u5de5\u667a\u80fd\u65f6\u4ee3\uff0c\u7528\u6237\u65e0\u9700\u62cd\u6444\u4e66\u7c4d\u3002\u76f8\u53cd\uff0c\u5e94\u7528\u7a0b\u5e8f\u6709\u4e00\u4e2a\u7b80\u5355\u7684 AR \u626b\u63cf\u529f\u80fd\uff0c\u53ef\u4ee5\u63d0\u53d6 ISBN\u3001\u626b\u63cf\u4e66\u7c4d\u7684\u72b6\u51b5\u3001\u63d0\u53d6\u627e\u5230\u7684\u4efb\u4f55\u6587\u672c\u3001\u786e\u5b9a\u7c7b\u578b\uff08\u4f8b\u5982\u8f6f\u5c01\u9762\u6216\u786c\u5c01\u9762\uff09\u5e76\u8bc6\u522b\u635f\u574f\u6216\u4f7f\u7528\u56fe\u50cf\u5206\u7c7b\u5668\u9700\u8981\u6267\u884c\u7684\u4efb\u4f55\u5176\u4ed6\u64cd\u4f5c\u3002<\/p>\n
\u7528\u6237\u6240\u8981\u505a\u7684\u5c31\u662f\u9a8c\u8bc1\u786e\u5b9a\u548c\u9884\u6d4b\u7684\u4fe1\u606f\u662f\u5426\u6b63\u786e\u3002\u501f\u52a9\u817e\u8baf\u7684 \uff0c\u5b83\u751a\u81f3\u53ef\u4ee5\u521b\u5efa\u5b9e\u9645\u4e66\u7c4d\u526f\u672c\u7684 3D \u6e32\u67d3\u3002\u8fd9\u5c06\u4f7f\u5176\u4ed6\u4eba\u5728\u51fa\u79df\u6216\u8d2d\u4e70\u4e4b\u524d\u4ee5 3D \u5f62\u5f0f\u67e5\u770b\u5b9e\u9645\u526f\u672c\u3002\u8fd9\u4e5f\u610f\u5473\u7740\u7535\u5b50\u5546\u52a1\u9700\u8981\u8fdb\u884c\u5f7b\u5e95\u7684\u6280\u672f\u6539\u9020\u3002\u4eca\u5929\uff0c\u7528\u6237\u5bf9\u6b64\u5370\u8c61\u6df1\u523b\u5e76\u611f\u5230\u60ca\u8bb6\uff0c\u4f46\u51e0\u4e2a\u6708\u540e\u4ed6\u4eec\u5c31\u4f1a\u671f\u5f85\u8fd9\u4e9b\u529f\u80fd\u3002<\/p>\n
5\u3001\u5982\u4f55\u6700\u597d\u5730\u52a0\u5165\u6dd8\u91d1\u70ed\u6f6e<\/h2>\n
\u5728\u8fd9\u573a\u6dd8\u91d1\u70ed\u6f6e\u4e2d\uff0c\u6211\u4eec\u7ecf\u5386\u7684\u65e0\u975e\u662f\u6211\u4eec\u6240\u719f\u77e5\u7684\u201c\u4e13\u5bb6\u7cfb\u7edf\u201d\u7684\u6700\u7ec8\u6d88\u4ea1\u3002\u4ec5\u4f9d\u8d56\u9884\u7f16\u7a0b\u6761\u4ef6\uff08\u5373\u201c\u5982\u679c\u2026\u2026\u90a3\u4e48\u201d\u7b97\u6cd5\uff09\u7684\u8f6f\u4ef6\u5df2\u843d\u5165\u9057\u7559\u8f6f\u4ef6\u7c7b\u522b\u3002\u8f6f\u4ef6\u4e0d\u518d\u5e2e\u52a9\u7528\u6237\u5b9e\u73b0\u6d41\u7a0b\u81ea\u52a8\u5316\uff0c\u800c\u662f\u5206\u6790\u6d41\u7a0b\u5e76\u9884\u6d4b\u64cd\u4f5c\u3002\u5b83\u751a\u81f3\u53ef\u4ee5\u7acb\u5373\u81ea\u52a8\u6267\u884c\u64cd\u4f5c\u3002\u60f3\u60f3\u6c7d\u8f66\u7684\u7d27\u6025\u5236\u52a8\u3002\u901a\u8fc7\u627e\u5230\u5c1a\u672a\u6216\u4e0d\u4f1a\u8f6c\u5411 AI \u548c ML \u7684\u201c\u9057\u7559\u8f6f\u4ef6\u201d\u6765\u52a0\u5165\u8fd9\u573a\u9769\u547d\u3002\u5e76\u66ff\u6362\u5b83\u3002<\/p>\n
 Excel \u516c\u5f0f\u548c VBA \u5df2\u7ecf\u6fd2\u4e34\u6d88\u4ea1\uff1a\u88ab ML \u548c RAG ETL \u7ba1\u9053\u53d6\u4ee3 <\/p>\n
\u8fc7\u53bb\uff0c\u6211\u4eec\u4f7f\u7528\u590d\u6742\u7684\u6570\u5b66\u516c\u5f0f\u8f6c\u5316\u4e3a\u7b97\u6cd5\u6765\u751f\u6210\u7279\u5b9a\u7ed3\u679c\u3002\u7136\u540e\u7b97\u6cd5\u4f1a\u53d8\u5f97\u8d8a\u6765\u8d8a\u590d\u6742\uff0c\u6db5\u76d6\u8d8a\u6765\u8d8a\u591a\u7684\u8fb9\u7f18\u60c5\u51b5\u3002\u6240\u6709\u8fd9\u4e9b\u90fd\u53ef\u4ee5\u8bad\u7ec3\u6210 ML \u6a21\u578b\uff0c\u8be5\u6a21\u578b\u5c06\u6839\u636e\u4e4b\u524d\u8bad\u7ec3\u7684\u6570\u636e\u9884\u6d4b\u8f93\u5165\u53c2\u6570\u7684\u7ed3\u679c\u3002\u5b83\u4f1a\u901a\u8fc7\u5f3a\u5316\u5b66\u4e60\u548c\u6a21\u578b\u518d\u8bad\u7ec3\u4e0d\u65ad\u66f4\u65b0\u3002<\/p>\n
6\u3001\u7ed3\u675f\u8bed<\/h2>\n
\u7f16\u7a0b\u5df2\u6b7b\uff0c\u7f16\u7a0b\u4e07\u5c81\uff01\u867d\u7136\u4eba\u5de5\u667a\u80fd\u662f\u4e00\u573a\u9769\u547d\uff0c\u4f46\u8f6f\u4ef6\u4ea7\u54c1\u53ea\u662f\u7ecf\u5386\u4e86\u4e00\u573a\u6f14\u53d8\uff1a\u4ece\u9759\u6001\u6761\u4ef6\u7b97\u6cd5\u5230\u673a\u5668\u5b66\u4e60\u7b97\u6cd5\u3002\u867d\u7136\u6211\u4eec\u9700\u8981\u66f4\u5c11\u7684\u8f6f\u4ef6\u5f00\u53d1\u4eba\u5458\u6765\u6784\u5efa\u6211\u4eec\u4eca\u5929\u6784\u5efa\u7684\u4e1c\u897f\uff0c\u4f46\u6211\u4eec\u9700\u8981\u66f4\u591a\u7684\u8f6f\u4ef6\u5f00\u53d1\u4eba\u5458\u6765\u6784\u5efa\u672a\u6765\u3002\u6211\u4eec\u6b63\u5904\u4e8e\u53e6\u4e00\u4e2a\u201c\u6280\u80fd\u8fc7\u65f6\u201d\u7684\u5faa\u73af\u4e2d\u3002<\/p>\n
\u4f5c\u4e3a\u4e00\u540d\u8f6f\u4ef6\u5f00\u53d1\u4eba\u5458\uff0c\u4f60\u73b0\u5728\u9700\u8981\u4e86\u89e3 ML \u7684\u57fa\u672c\u65b9\u6cd5\u4ee5\u53ca\u5982\u4f55\u4f7f\u7528\u5b83\u4eec\u3002\u5728 Mac \u4e0a\u521b\u5efa ML \u662f\u4e00\u4e2a\u5f88\u597d\u7684\u8d77\u70b9\u3002\u4f60\u4e0d\u9700\u8981\u80fd\u591f\u81ea\u5df1\u5b9e\u73b0 K-MEANS \u548c K-NN \u7b97\u6cd5\uff0c\u4f46\u60a8\u9700\u8981\u4e86\u89e3\u5b83\u4eec\u662f\u4ec0\u4e48\u3001\u5982\u4f55\u5de5\u4f5c\u4ee5\u53ca\u4f55\u65f6\u4f7f\u7528\u5b83\u4eec\u3002\u4f60\u8fd8\u9700\u8981\u5bf9 LLM\u3001TTS\u3001\u7a33\u5b9a\u6269\u6563\u548c\u8ba1\u7b97\u673a\u89c6\u89c9\u7684\u5de5\u4f5c\u539f\u7406\u6709\u57fa\u672c\u7684\u4e86\u89e3\u3002<\/p>\n
\u867d\u7136\u6211\u4eec\u8fd9\u4e9b\u5927\u56db\u5b66\u751f\u7ecf\u5e38\u5bf9\u5927\u4e09\u5b66\u751f\u4f7f\u7528 LLM \u505a\u6240\u6709\u4e8b\u60c5\u611f\u5230\u597d\u7b11\uff0c\u4f46\u8fd9\u6b63\u5728\u6162\u6162\u6210\u4e3a\u73b0\u5b9e\u3002\u4e0e\u6211\u4eec\u7684\u9759\u6001\u6761\u4ef6\u7b97\u6cd5\u76f8\u6bd4\uff0cLLM \u53ef\u4ee5\u6db5\u76d6\u66f4\u591a\u4fee\u590d\u635f\u574f\u7684 CSV \u6587\u4ef6\u7684\u60c5\u51b5\u3002\u5982\u679c LLM \u5f88\u5feb\u5c31\u80fd\u52a0\u5165\u64cd\u4f5c\u7cfb\u7edf\uff0c\u90a3\u53c8\u4f55\u5fc5\u5462\uff1f\u4e0d\u8981\u4e0e\u672a\u6765\u4f5c\u6597\u4e89\uff0c\u8981\u62e5\u62b1\u5b83\u3002\u4eca\u5929\u770b\u4f3c\u8fc7\u5ea6\u7684\u4e1c\u897f\uff0c\u660e\u5929\u5c31\u4f1a\u6210\u4e3a\u5e38\u6001\u3002<\/p>\n
\u6b22\u8fce\u6765\u5230 AI \u7f16\u7a0b\u6dd8\u91d1\u70ed\uff01<\/p>\n
\n","protected":false},"excerpt":{"rendered":"
\u65b0\u4e00\u8f6e\u7f16\u7a0b\u6dd8\u91d1\u70ed\u6b63\u5728\u5982\u706b\u5982\u837c\u5730\u5c55\u5f00\u3002\u53ea\u662f\u8fd9\u4e00\u6b21\uff0c\u4f60\u4e0d\u9700\u8981\u6602\u8d35\u7684\u94f2\u5b50\uff0c\u4f60\u4f1a\u5f97\u5230\u514d\u8d39\u7684\u6316\u6398\u673a\u3002\u6211\u5728 90 \u5e74\u4ee3\u7ecf\u5386\u8fc7\u4e92\u8054\u7f51\u6dd8\u91d1\u70ed\uff0c\u5728 21 \u4e16\u7eaa\u521d\u7ecf\u5386\u8fc7\u79fb\u52a8\u6dd8\u91d1\u70ed\uff0c\u73b0\u5728\uff0c\u6211\u4eec\u7ec8\u4e8e\u8fce\u6765\u4e86 AI \u6dd8\u91d1\u70ed\u3002\u542c\u8d77\u6765\u6709\u4e9b\u5938\u5f20\uff0c\u4f46\u4e8b\u5b9e\u5e76\u975e\u5982\u6b64\u3002 OpenAI\u3001Anthropic\u3001DeepSeek \u548c X \u7b49\u516c\u53f8\u90fd\u9700\u8981\u5927\u91cf\u7684 Nvidia GPU\u3002\u50cf\u6211\u8fd9\u6837\u7684\u72ec\u7acb\u5f00\u53d1\u8005\u6839\u672c\u8d1f\u62c5\u4e0d\u8d77\u6216\u6295\u8d44\u3002\u6b63\u5982\u4f60\u4ece\u6211\u7684\u6587\u7ae0\u4e2d\u4e86\u89e3\u5230\u7684\u90a3\u6837\uff0c\u53ea\u6709 10% \u7684\u7528\u6237\u53ef\u80fd\u4f1a\u5728\u514d\u8d39\u589e\u503c\u6a21\u5f0f\u4e0b\u4ed8\u8d39\u3002\u8fd9\u610f\u5473\u7740\u4f60\u5fc5\u987b\u4e3a 90% \u7684\u514d\u8d39\u7528\u6237\u4fdd\u6301 AI \u63a8\u7406\u80fd\u529b\u3002\u5bf9\u4e8e\u72ec\u7acb\u5f00\u53d1\u8005\u6765\u8bf4\u8fd9\u662f\u4e0d\u53ef\u80fd\u7684\u3002 \u73b0\u5728\u60c5\u51b5\u5df2\u7ecf\u6539\u53d8\u3002\u4f7f\u7528 llama.cpp \u5e93\u7684 Ollama \u7b49\u8f6f\u4ef6\u4e0a\u7684\u672c\u5730 AI \u548c\u672c\u5730 LLM \u6b63\u5728\u5174\u8d77\u3002\u6587\u672c\u8f6c\u8bed\u97f3\u3001\u6587\u672c\u8f6c\u6587\u672c\u3001\u6587\u672c\u8f6c\u56fe\u50cf\u3001\u56fe\u50cf\u8f6c\u6587\u672c\u548c\u5927\u91cf\u5176\u4ed6\u6a21\u578b\u73b0\u5728\u53ef\u4ee5\u5728\u6700\u7ec8\u7528\u6237\u8bbe\u5907\u4e0a\u8fd0\u884c\uff0c\u751a\u81f3\u53ef\u4ee5\u5728 iPhone \u548c iPad \u4e0a\u8fd0\u884c\u3002\u6211\u6700\u8fd1\u521a\u521a\u63a8\u51fa\u4e86\u6211\u7684\uff0c\u8fd9\u662f\u4e00\u6b3e\u5373\u65f6 AI \u64ad\u5ba2\u751f\u6210\u5668\uff0c\u53ef\u5728 iOS \u4e0a 100% \u672c\u5730\u751f\u6210\u3002\u5b83\u4f7f\u7528 TinyLlama \u548c Kokoro \u6587\u672c\u8f6c\u8bed\u97f3\u6a21\u578b\u6765\u751f\u6210\u64ad\u5ba2\u3002\u6211\u8fd9\u8fb9\u4e0d\u9700\u8981\u57fa\u7840\u8bbe\u65bd\uff0c\u90fd\u662f BYOG\uff08\u201c\u81ea\u5e26 GPU\u201d\uff09\u3002BYOG \u662f\u72ec\u7acb\u5f00\u53d1\u8005\u8fdb\u5165 AI \u7684\u5927\u95e8\u3002 1\u3001\u6dd8\u91d1\u6240\u9700\u7684\u5de5\u5177 \u8bf4\u6559\u5df2\u7ecf\u591f\u591a\u4e86\uff0c\u8ba9\u6211\u4eec\u6765\u770b\u770b\u6211\u4eec\u624b\u5934\u4e0a\u6709\u54ea\u4e9b\u5de5\u5177\u53ef\u4ee5\u52a0\u5165\u672c\u5730 AI \u6dd8\u91d1\u70ed\u3002\u6211\u4eec\u5c06\u91cd\u70b9\u4ecb\u7ecd\u5728 Windows\u3001Linux \u548c Mac \u4e0a\u63d0\u4f9b […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[13],"tags":[],"class_list":["post-53756","post","type-post","status-publish","format-standard","hentry","category-ai"],"_links":{"self":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts\/53756","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/comments?post=53756"}],"version-history":[{"count":0,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts\/53756\/revisions"}],"wp:attachment":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/media?parent=53756"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/categories?post=53756"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/tags?post=53756"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}