\u4f60\u7684\u81ea\u5efa AI \u804a\u5929\u6a5f\u5668\u4eba\u53ef\u4ee5\u5728\u696d\u52d9\u9ad8\u5cf0\u671f\u4f9d\u820a\u4fdd\u6301\u7a69\u5b9a\u7684\u56de\u61c9\u6642\u9593\uff0c\u4f60\u4e5f\u56e0\u6b64\u7372\u5f97\u5c0d\u6548\u80fd\u8207\u96b1\u79c1\u7684\u5b8c\u5168\u638c\u63a7\u3002<\/p>\n

\u95dc\u9375\u8981\u9ede<\/strong><\/h2>\n

\n
• \u81ea\u5efa AI \u804a\u5929\u52a9\u624b\u53ef\u4ee5\u4fdd\u8b77\u654f\u611f\u6a94\u6848\uff0c\u4e26\u6d88\u9664\u4f9d token \u8a08\u8cbb\u7684 API \u7d93\u5e38\u6027\u6210\u672c\u3002<\/li>\n
• \u672c\u5730\u4f3a\u670d\u5668\u5728\u7e41\u5fd9\u9ad8\u5cf0\u671f\u4f9d\u7136\u80fd\u63d0\u4f9b\u7a69\u5b9a\u56de\u61c9\u6642\u9593\u8207\u53ef\u9810\u671f\u7684\u6548\u80fd\u3002<\/li>\n
• \u5408\u9069\u7684 GPU \u786c\u9ad4\u8207 Docker \u5bb9\u5668\u80fd\u8b93\u4f60\u7684\u81ea\u8a02 AI \u6a21\u578b\u6301\u7e8c\u5e73\u7a69\u904b\u884c\u3002<\/li>\n
• \u672c\u5730\u5411\u91cf\u8cc7\u6599\u5eab\u8b93 AI \u5728\u4e0d\u6d29\u6f0f\u4f01\u696d\u8cc7\u6599\u7684\u524d\u63d0\u4e0b\u641c\u5c0b\u5167\u90e8\u6587\u4ef6\u3002<\/li>\n<\/ul>\n

  \u7f8e\u570b\u4f3a\u670d\u5668\u74b0\u5883\u8207 AI \u90e8\u7f72<\/strong><\/h2>\n

  \u5728\u904b\u884c\u672c\u5730\u8edf\u9ad4\u4e4b\u524d\uff0c\u4f60\u9700\u8981\u5728\u7f8e\u570b\u4f3a\u670d\u5668\u4e0a\u6e96\u5099\u597d\u5408\u9069\u7684\u5be6\u9ad4\u786c\u9ad4\u3002\u5408\u7406\u7684\u914d\u7f6e\u53ef\u4ee5\u8b93\u4f60\u5b8c\u5168\u638c\u63a7\u904b\u7b97\u901f\u5ea6\u8207\u7cfb\u7d71\u96b1\u79c1\u3002<\/p>\n

  \u786c\u9ad4\u898f\u683c\u8207\u4f5c\u696d\u7cfb\u7d71\u8a2d\u5b9a<\/h3>\n

  \u4f60\u76ee\u6a19\u6a21\u578b\u7684\u53c3\u6578\u898f\u6a21\u5c07\u6c7a\u5b9a GPU \u986f\u793a\u8a18\u61b6\u9ad4\u8207\u6574\u9ad4\u7cfb\u7d71\u786c\u9ad4\u9700\u6c42\u3002\u6a21\u578b\u6108\u5927\uff0c\u5c31\u6108\u9700\u8981\u5f37\u52c1\u7684\u7e6a\u5716\u8655\u7406\u5668\u4f86\u9ad8\u6548\u670d\u52d9\u7dda\u4e0a\u667a\u80fd\u9ad4\u3002<\/p>\n

  \n\n\n\n<\/colgroup>\n\n\n\n\n\n

  \u6a21\u578b\u898f\u6a21<\/th>\n	\u5efa\u8b70\u786c\u9ad4<\/th>\n<\/tr>\n
  7B \u6a21\u578b<\/strong><\/td>\n	NVIDIA RTX 5060 Ti 16GB \/ RTX 4060 8GB<\/td>\n<\/tr>\n
  13B\u201334B \u6a21\u578b<\/strong><\/td>\n	AMD RX 9070 XT 16GB \/ NVIDIA RTX 5080 16GB<\/td>\n<\/tr>\n
  70B \u6a21\u578b<\/strong><\/td>\n	Mac Studio M4 Max \u6216 Mac Studio M3 Ultra 192GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u5728 4 \u4f4d\u5143\u7cbe\u5ea6\u4e0b\u904b\u884c 700 \u5104\u53c3\u6578\u6a21\u578b\uff0c\u5927\u7d04\u9700\u8981 40GB \u986f\u793a\u8a18\u61b6\u9ad4\u3002\u4f60\u53ef\u4ee5\u4f7f\u7528\u5169\u5f35 RTX 3090 \u986f\u793a\u5361\uff0c\u6216\u4e00\u5f35 A100 GPU\u3002\u82e5\u50c5\u7528 CPU \u57f7\u884c\uff0c\u81f3\u5c11\u9700\u8981 64GB \u8a18\u61b6\u9ad4\uff0c\u800c 128GB \u8a18\u61b6\u9ad4\u5247\u80fd\u5728\u8907\u96dc AI \u5de5\u4f5c\u6d41\u7a0b\u4e0b\u63d0\u4f9b\u66f4\u6d41\u66a2\u7684\u7cfb\u7d71\u8868\u73fe\u3002\u8acb\u52d9\u5fc5\u5c07 Linux \u5b89\u88dd\u5728 NVMe \u786c\u789f\u4e0a\uff0c\u4ee5\u52a0\u901f\u6587\u4ef6\u77e5\u8b58\u6aa2\u7d22\u3002<\/p>\n Docker \u5bb9\u5668\u57fa\u790e\u8a2d\u65bd\u642d\u5efa<\/h3>\n \ud83d\udca1 \u63d0\u793a\uff1a<\/strong>\u5bb9\u5668\u53ef\u4ee5\u5c07\u8edf\u9ad4\u76f8\u4f9d\u95dc\u4fc2\u6e05\u695a\u9694\u96e2\uff0c\u4e26\u7c21\u5316\u6574\u9ad4\u74b0\u5883\u7684\u5b89\u88dd\u8207\u7ba1\u7406\u3002<\/p><\/blockquote>\n \u4f60\u53ef\u4ee5\u900f\u904e\u4ee5\u4e0b\u516d\u500b\u6b65\u9a5f\uff0c\u5728 Ubuntu \u4e0a\u8a2d\u5b9a Docker \u8207 GPU \u652f\u63f4\uff1a<\/p>\n \n \u4f7f\u7528 sudo apt-get remove -y docker docker-engine docker.io containerd runc<\/code> \u79fb\u9664\u820a\u7248\u672c\u76f8\u95dc\u8edf\u9ad4\u5957\u4ef6\u3002<\/li>\n \u4f7f\u7528 sudo apt-get update && sudo apt-get install -y docker-ce docker-ce-cli containerd.io<\/code> \u5b89\u88dd Docker\u3002<\/li>\n \u900f\u904e sudo usermod -aG docker $USER<\/code> \u5c07\u4e00\u822c\u4f7f\u7528\u8005\u52a0\u5165 Docker \u7fa4\u7d44\uff0c\u4e26\u4f7f\u7528 newgrp docker<\/code> \u66f4\u65b0\u76ee\u524d\u5de5\u4f5c\u968e\u6bb5\u8a2d\u5b9a\u3002<\/li>\n \u532f\u5165 NVIDIA keyring\uff0c\u4e26\u5728 \/etc\/apt\/sources.list.d\/nvidia-container-toolkit.list<\/code> \u4e2d\u52a0\u5165\u5c0d\u61c9\u8edf\u9ad4\u4f86\u6e90\u3002<\/li>\n \u4f7f\u7528 sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit<\/code> \u5b89\u88dd\u52a0\u901f\u5de5\u5177\u3002<\/li>\n \u900f\u904e sudo nvidia-ctk runtime configure --runtime=docker<\/code> \u8a2d\u5b9a\u57f7\u884c\u968e\u6bb5\u9a45\u52d5\uff0c\u4e26\u4f7f\u7528 sudo systemctl restart docker<\/code> \u91cd\u65b0\u555f\u52d5\u670d\u52d9\u3002<\/li>\n<\/ol>\n \u63a5\u4e0b\u4f86\uff0c\u8981\u4fdd\u8b77\u4f60\u7684\u7db2\u8def\u74b0\u5883\u3002\u70ba\u6bcf\u4f4d\u6388\u6b0a\u4f7f\u7528\u8005\u555f\u7528\u9632\u706b\u7246\u898f\u5247\uff0c\u963b\u64cb\u672a\u6388\u6b0a\u9023\u7dda\u3002\u8a2d\u5b9a SSH \u91d1\u9470\u9a57\u8b49\u4e26\u505c\u7528\u5bc6\u78bc\u767b\u5165\u3002\u9019\u4e9b\u6b65\u9a5f\u53ef\u4ee5\u4fdd\u8b77\u4f60\u7684\u672c\u5730 AI \u5f15\u64ce\uff0c\u4f7f\u81ea\u8a02\u667a\u80fd\u9ad4\u548c\u667a\u6167\u4ee3\u7406\u5728\u67e5\u8a62\u672c\u5730\u77e5\u8b58\u5eab\u6642\u4e0d\u6703\u6d29\u9732\u5167\u90e8\u8cc7\u6599\u3002<\/p>\n \u90e8\u7f72\u672c\u5730 AI \u6a21\u578b\u5f15\u64ce<\/strong><\/h2>\n \u5b8c\u6210\u4f3a\u670d\u5668\u8edf\u9ad4\u90e8\u7f72\u5f8c\uff0c\u4f60\u5c31\u53ef\u4ee5\u5728\u4e0d\u66b4\u9732\u654f\u611f\u6a94\u6848\u7684\u524d\u63d0\u4e0b\u904b\u884c\u672c\u5730 AI \u804a\u5929\u6a5f\u5668\u4eba\u3002\u4f60\u53ef\u4ee5\u5229\u7528\u958b\u6e90\u5de5\u5177\uff0c\u5c07\u7f8e\u570b\u4f3a\u670d\u5668\u5be6\u4f8b\u5347\u7d1a\u70ba\u79c1\u6709\u63a8\u7406\u5f15\u64ce\u3002<\/p>\n \u4f7f\u7528 Ollama \u57f7\u884c\u672c\u5730\u6a21\u578b<\/h3>\n Ollama \u662f\u5728\u672c\u5730\u786c\u9ad4\u4e0a\u76f4\u63a5\u904b\u884c\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u7684\u6838\u5fc3\u5f15\u64ce\u3002\u4f60\u53ef\u4ee5\u900f\u904e Docker Compose \u5c0d\u6574\u9ad4\u670d\u52d9\u68e7\u9032\u884c\u512a\u96c5\u7de8\u6392\u3002<\/p>\n \n \u5148\u6c7a\u689d\u4ef6\u8207 GPU \u8a2d\u5b9a<\/strong>\uff1a\u5728\u5bbf\u4e3b\u74b0\u5883\u4e2d\uff0c\u900f\u904e sudo apt-get install -y nvidia-container-toolkit<\/code> \u5b89\u88dd NVIDIA Container Toolkit\uff0c\u7136\u5f8c\u4f7f\u7528 sudo nvidia-ctk runtime configure --runtime=docker<\/code> \u5c07\u8a72\u57f7\u884c\u968e\u6bb5\u8a3b\u518a\u5230\u5bb9\u5668\u5b88\u8b77\u884c\u7a0b\u4e2d\u3002<\/li>\n \u5c08\u6848\u521d\u59cb\u5316<\/strong>\uff1a\u4f7f\u7528 git clone https:\/\/github.com\/mythrantic\/ollama-docker.git<\/code> \u4e0b\u8f09\u8a2d\u5b9a\u5132\u5b58\u5eab\uff1b\u518d\u900f\u904e cd ollama-docker<\/code> \u9032\u5165\u5de5\u4f5c\u76ee\u9304\u3002<\/li>\n \u5728 <\/strong>docker-compose.yml<\/code> \u4e2d\u5b9a\u7fa9\u670d\u52d9<\/strong>\uff1a\u8a2d\u5b9a ollama<\/code> \u5bb9\u5668\u670d\u52d9\uff0c\u6620\u50cf\u70ba ollama\/ollama:latest<\/code>\uff0c\u4e26\u5c07\u5bbf\u4e3b\u9023\u63a5\u57e0 11434:11434<\/code> \u9032\u884c\u5c0d\u61c9\u3002\u5c07\u8cc7\u6599\u76ee\u9304\u6307\u5411 .\/model_files<\/code>\uff0c\u4e26\u6307\u5b9a run_ollama.sh<\/code> \u70ba\u4e3b\u8981\u5165\u53e3\u8173\u672c\u3002\u5c07 Web \u524d\u7aef\u670d\u52d9\u5c0d\u61c9\u70ba 3000:8080<\/code>\uff0c\u4e26\u8a2d\u5b9a OLLAMA_API_BASE_URL=http:\/\/ollama:11434\/api<\/code>\u3002<\/li>\n \u5efa\u7acb\u555f\u52d5\u8173\u672c\uff08<\/strong>model_files\/run_ollama.sh<\/code>\uff09<\/strong>\uff1a\u4f7f\u7528 ollama serve &<\/code> \u555f\u52d5\u5f8c\u7aef\u670d\u52d9\uff0c\u518d\u4f7f\u7528 ollama create finetuned_mistral -f model_files\/Modelfile<\/code> \u548c ollama run finetuned_mistral<\/code> \u5efa\u7f6e\u4e26\u904b\u884c\u81ea\u8a02\u6a21\u578b\u3002<\/li>\n \u8a2d\u5b9a\u6a21\u578b\u76f8\u4f9d\u6027\uff08<\/strong>Modelfile<\/code>\uff09<\/strong>\uff1a\u900f\u904e FROM .\/mistral-7b-instruct-v0.1.Q2_K.gguf<\/code> \u5c07\u6a21\u578b\u6307\u5411\u672c\u5730\u6b0a\u91cd\u6a94\u6848\u3002<\/li>\n \u90e8\u7f72\u8207\u904b\u884c<\/strong>\uff1a\u4f7f\u7528 docker compose up -d<\/code> \u555f\u52d5\u6a19\u6e96 CPU \u90e8\u7f72\uff1b\u5982\u9700\u555f\u7528 GPU \u52a0\u901f\uff0c\u5247\u900f\u904e docker compose -f docker-compose-ollama-gpu.yaml up -d<\/code> \u555f\u52d5\u3002\u96a8\u5f8c\u5728\u700f\u89bd\u5668\u4e2d\u9020\u8a2a http:\/\/localhost:8080<\/code> \u9032\u5165\u4ecb\u9762\u3002<\/li>\n<\/ol>\n \ud83d\udca1 \u63d0\u793a\uff1a<\/strong>\u5408\u7406\u8abf\u6821\u74b0\u5883\u8b8a\u6578\u53ef\u4ee5\u6700\u4f73\u5316\u986f\u793a\u8a18\u61b6\u9ad4\u5360\u7528\uff0c\u4e26\u63d0\u5347\u4e26\u767c\u8acb\u6c42\u8655\u7406\u901f\u5ea6\u3002<\/p><\/blockquote>\n \u4f60\u53ef\u4ee5\u900f\u904e\u5728\u5bb9\u5668\u74b0\u5883\u4e2d\u8a2d\u5b9a\u5c08\u7528\u74b0\u5883\u8b8a\u6578\u4f86\u512a\u5316\u6548\u80fd\uff1a<\/p>\n \n\n\n\n\n<\/colgroup>\n\n\n\n\n\n\n\n\n\n\n\n \u74b0\u5883\u8b8a\u6578<\/th>\n \u985e\u5225<\/th>\n \u6548\u80fd\u8207\u8a18\u61b6\u9ad4\u5f71\u97ff<\/th>\n<\/tr>\n OLLAMA_FLASH_ATTENTION<\/code><\/td>\n GPU \/ \u8a18\u61b6\u9ad4<\/td>\n \u70ba CUDA \u555f\u7528 FlashAttention\uff0c\u52a0\u901f\u9577\u4e0a\u4e0b\u6587\u8655\u7406\u4e26\u964d\u4f4e\u8a18\u61b6\u9ad4\u958b\u92b7\u3002<\/td>\n<\/tr>\n OLLAMA_KV_CACHE_TYPE<\/code><\/td>\n \u8a18\u61b6\u9ad4\u6700\u4f73\u5316<\/td>\n \u5c0d KV cache \u9032\u884c\u91cf\u5316\uff08\u4f8b\u5982 q8_0<\/code> \u6216 q4_0<\/code>\uff09\uff0c\u53ef\u986f\u8457\u964d\u4f4e\u986f\u793a\u8a18\u61b6\u9ad4\u5360\u7528\uff0c\u4e26\u5728\u7d50\u5408 Flash Attention \u6642\u652f\u63f4\u6700\u591a 2 \u500d\u4e0a\u4e0b\u6587\u9577\u5ea6\u3002<\/td>\n<\/tr>\n OLLAMA_GPU_OVERHEAD<\/code><\/td>\n \u986f\u793a\u8a18\u61b6\u9ad4\u9810\u7559<\/td>\n \u70ba\u5bbf\u4e3b\u4f5c\u696d\u7cfb\u7d71\u6216\u5176\u4ed6\u61c9\u7528\u7a0b\u5f0f\u9810\u7559\u6307\u5b9a\u4f4d\u5143\u7d44\u6578\u7684\u986f\u793a\u8a18\u61b6\u9ad4\uff0c\u9632\u6b62 OOM \u7576\u6a5f\u3002<\/td>\n<\/tr>\n OLLAMA_MAX_QUEUE<\/code><\/td>\n \u4f47\u5217\u7ba1\u7406<\/td>\n \u9650\u5236\u4e26\u767c\u5f85\u8655\u7406 API \u8acb\u6c42\u6578\u91cf\uff0c\u78ba\u4fdd\u9ad8\u5cf0\u671f\u7cfb\u7d71\u8ca0\u8f09\u7a69\u5b9a\u3002<\/td>\n<\/tr>\n OLLAMA_NUM_PARALLEL<\/code><\/td>\n \u4e26\u767c\u63a7\u5236<\/td>\n \u8a2d\u5b9a\u6bcf\u500b\u5df2\u8f09\u5165\u6a21\u578b\u5141\u8a31\u7684\u6700\u5927\u4e26\u884c\u8acb\u6c42\u6578\u3002<\/td>\n<\/tr>\n OLLAMA_MAX_LOADED_MODELS<\/code><\/td>\n \u8a18\u61b6\u9ad4\u7ba1\u7406<\/td>\n \u63a7\u5236\u53ef\u540c\u6642\u8f09\u5165\u5728\u8a18\u61b6\u9ad4\u4e2d\u7684\u6a21\u578b\u6578\u91cf\u4e0a\u9650\u3002<\/td>\n<\/tr>\n OLLAMA_KEEP_ALIVE<\/code><\/td>\n \u8a18\u61b6\u9ad4\u4fdd\u7559<\/td>\n \u8a2d\u5b9a\u6a21\u578b\u5728\u88ab\u5378\u8f09\u524d\u4fdd\u6301\u8f09\u5165\u65bc\u8a18\u61b6\u9ad4\u4e2d\u7684\u6642\u9593\u9577\u5ea6\u3002<\/td>\n<\/tr>\n OLLAMA_CONTEXT_LENGTH<\/code><\/td>\n \u8a18\u61b6\u9ad4\u9650\u5236<\/td>\n \u8a2d\u5b9a\u9810\u8a2d\u4e0a\u4e0b\u6587\u8996\u7a97\u5927\u5c0f\uff0c\u76f4\u63a5\u5f71\u97ff\u6bcf\u500b\u6703\u8a71\u7684\u986f\u793a\u8a18\u61b6\u9ad4\u5206\u914d\u3002<\/td>\n<\/tr>\n OMP_NUM_THREADS<\/code><\/td>\n CPU \u591a\u57f7\u884c\u7dd2<\/td>\n \u70ba\u591a\u6838\u5fc3\u5bbf\u4e3b\u7cfb\u7d71\u6307\u5b9a\u5e73\u884c\u57f7\u884c\u6240\u4f7f\u7528\u7684 CPU \u57f7\u884c\u7dd2\u6578\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u5408\u7406\u8a2d\u5b9a\u5f8c\uff0c\u4f60\u7684 AI \u667a\u80fd\u9ad4\u5c31\u80fd\u5728\u4e0d\u8017\u76e1\u7cfb\u7d71\u8cc7\u6e90\u7684\u524d\u63d0\u4e0b\u5feb\u901f\u56de\u61c9\u4f7f\u7528\u8005\u8acb\u6c42\u3002<\/p>\n \u7528\u65bc\u96e2\u7dda\u67e5\u8a62\u7684\u6a21\u578b\u9078\u64c7<\/h3>\n \u9078\u64c7\u5408\u9069\u7684\u6a21\u578b\u67b6\u69cb\uff0c\u6c7a\u5b9a\u4e86\u4f60\u7684\u77e5\u8b58\u9a45\u52d5\u667a\u80fd\u9ad4\u5728\u8655\u7406\u8907\u96dc\u6587\u4ef6\u6642\u7684\u8868\u73fe\u3002\u73fe\u4ee3\u958b\u6e90\u6b0a\u91cd\u6a21\u578b\u53ef\u4ee5\u70ba\u672c\u5730\u77e5\u8b58\u5eab\u63d0\u4f9b\u51fa\u8272\u7684\u63a8\u7406\u80fd\u529b\u3002<\/p>\n \n\n\n\n\n\n\n<\/colgroup>\n\n\n\n\n\n\n \u6a21\u578b\u540d\u7a31<\/th>\n \u4e3b\u8981\u512a\u52e2\u8207\u61c9\u7528\u60c5\u5883<\/th>\n \u4e0a\u4e0b\u6587\u8996\u7a97<\/th>\n \u53c3\u6578\u914d\u7f6e<\/th>\n \u6388\u6b0a\u689d\u6b3e<\/th>\n<\/tr>\n Gemma 4 26B A4B<\/strong><\/td>\n \u9ad8\u6548\u672c\u5730\u6587\u4ef6\u5206\u6790\u3001\u79c1\u6709\u5de5\u4f5c\u6d41\u7a0b\u3001\u8f15\u91cf\u7b46\u96fb\u7aef\u90e8\u7f72<\/td>\n 256K tokens<\/td>\n \u7e3d\u53c3\u6578 25.2B\uff08\u5176\u4e2d 3.8B \u555f\u7528\uff09<\/td>\n Apache 2.0<\/td>\n<\/tr>\n Llama 4 Scout<\/strong><\/td>\n \u8d85\u9577\u4e0a\u4e0b\u6587\u8655\u7406\uff0c\u9069\u5408\u6cd5\u5f8b\u6587\u4ef6\u3001\u7a0b\u5f0f\u78bc\u5132\u5b58\u5eab\u3001\u5927\u578b\u8cc7\u6599\u593e<\/td>\n 10M tokens<\/td>\n 109B \u7e3d\u53c3\u6578 MoE\uff0817B \u555f\u7528\uff09<\/td>\n Llama 4 Community<\/td>\n<\/tr>\n Phi-4-reasoning<\/strong><\/td>\n \u9069\u5408\u8f15\u91cf\u672c\u5730\u63a8\u7406\u3001\u6578\u5b78\u3001\u7406\u5de5\u79d1\u8207\u7a0b\u5f0f\u8f14\u52a9\u60c5\u5883<\/td>\n \u6a19\u6e96\u4e0a\u4e0b\u6587<\/td>\n \u57fa\u65bc Phi-4 \u7684\u63a8\u7406\u5fae\u8abf\u7248\u672c<\/td>\n MIT<\/td>\n<\/tr>\n DeepSeek R1<\/strong><\/td>\n \u9069\u5408\u96e2\u7dda\u9ad8\u968e\u63a8\u7406\uff08\u53ef\u900f\u904e\u84b8\u993e\u8b8a\u9ad4\u5728\u4e00\u822c\u786c\u9ad4\u4e0a\u904b\u884c\uff09<\/td>\n 128K tokens<\/td>\n 671B MoE\uff0837B \u555f\u7528\uff09\/ \u84b8\u993e\u6a21\u578b\uff081.5B\u201370B\uff09<\/td>\n \u958b\u6e90\u6b0a\u91cd<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u5e95\u5c64\u6a21\u578b\u7684\u9078\u64c7\u6703\u6c7a\u5b9a\u667a\u80fd\u9ad4\u5982\u4f55\u5f9e\u5167\u90e8\u6587\u4ef6\u5eab\u4e2d\u89e3\u6790\u8207\u64f7\u53d6\u8cc7\u8a0a\u3002\u4f60\u53ef\u4ee5\u900f\u904e\u91cf\u5316\u7b49\u65b9\u5f0f\u8abf\u6574\u6a21\u578b\u53c3\u6578\uff0c\u5728\u6587\u4ef6\u6e96\u78ba\u5ea6\u8207\u986f\u793a\u8a18\u61b6\u9ad4\u5360\u7528\u4e4b\u9593\u53d6\u5f97\u5e73\u8861\u3002<\/p>\n \n\n\n\n\n\n<\/colgroup>\n\n\n\n\n \u91cf\u5316\u7b49\u7d1a<\/th>\n \u91cf\u5316\u8b8a\u9ad4<\/th>\n \u8a18\u61b6\u9ad4 \/ \u5132\u5b58\u5360\u7528<\/th>\n \u6a21\u578b\u7cbe\u5ea6\u5f71\u97ff<\/th>\n<\/tr>\n 8 \u4f4d\u5143<\/strong><\/td>\n Q8_0<\/td>\n \u9700\u8981\u8f03\u9ad8\u7684\u5132\u5b58\u8207\u8a18\u61b6\u9ad4\u5360\u7528<\/td>\n \u4f5c\u70ba\u9ad8\u4fdd\u771f\u57fa\u6e96\uff0c\u6548\u80fd\u8207\u7cbe\u5ea6\u975e\u5e38\u63a5\u8fd1 FP16<\/td>\n<\/tr>\n 4 \u4f4d\u5143<\/strong><\/td>\n Q4_1\u3001Q4_K_S\/M<\/td>\n \u986f\u8457\u964d\u4f4e\u8a18\u61b6\u9ad4\u9700\u6c42<\/td>\n \u4ee5\u8f03\u5c0f\u7684\u7cbe\u5ea6\u72a7\u7272\u63db\u53d6\u66f4\u9ad8\u7684\u904b\u884c\u901f\u5ea6<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u90e8\u7f72 4 \u4f4d\u5143\u91cf\u5316\u6a21\u578b\u53ef\u4ee5\u8b93\u6a19\u6e96\u786c\u9ad4\u904b\u884c\u9ad8\u968e\u81ea\u6cbb\u667a\u80fd\u9ad4\uff0c\u540c\u6642\u4e0d\u6703\u986f\u8457\u72a7\u7272\u6838\u5fc3\u5206\u6790\u80fd\u529b\u3002\u81ea\u6cbb\u667a\u80fd\u9ad4\u5728\u672c\u5730\u8655\u7406\u6587\u4ef6\u4e0a\u4e0b\u6587\uff0c\u6574\u500b\u904e\u7a0b\u7121\u9700\u5c07\u5167\u5bb9\u66b4\u9732\u7d66\u4efb\u4f55\u5916\u90e8\u7db2\u8def\uff0c\u5f9e\u800c\u78ba\u4fdd\u4f01\u696d\u5167\u90e8\u77e5\u8b58\u5b8c\u5168\u5728\u53ef\u63a7\u7bc4\u570d\u5167\u3002<\/p>\n \u69cb\u5efa AI \u804a\u5929\u52a9\u624b\u8207\u77e5\u8b58\u5eab<\/strong><\/h2>\n \u5c07\u672c\u5730\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u8207\u5c08\u6709\u6a94\u6848\u6253\u901a\uff0c\u53ef\u4ee5\u69cb\u5efa\u4e00\u500b\u667a\u6167\u5de5\u4f5c\u7a7a\u9593\u3002\u4f60\u80fd\u5920\u642d\u5efa\u50c5\u9762\u5411\u5167\u90e8\u554f\u984c\u7684\u79c1\u6709\u77e5\u8b58\u5eab\uff0c\u800c\u4e0d\u6703\u6d29\u9732\u8cc7\u6599\u3002\u81ea\u5efa\u7cfb\u7d71\u53ef\u4ee5\u628a\u975c\u614b\u6a94\u6848\u8f49\u63db\u70ba\u9762\u5411\u6bcf\u4f4d\u6388\u6b0a\u4f7f\u7528\u8005\u7684\u52d5\u614b\u60c5\u5883\u5316\u56de\u61c9\u3002<\/p>\n \u6587\u4ef6\u532f\u5165\u8207\u5411\u91cf\u5d4c\u5165<\/h3>\n \u90e8\u7f72\u6d41\u7a0b\u9996\u5148\u8981\u628a\u5167\u90e8\u6a94\u6848\u8f49\u5316\u70ba\u6578\u503c\u8868\u793a\u3002\u6587\u4ef6\u8f09\u5165\u5668\u5f9e PDF\u3001Markdown \u7b46\u8a18\u3001\u8a66\u7b97\u8868\u7b49\u539f\u59cb\u6a94\u6848\u4e2d\u64f7\u53d6\u6587\u5b57\u3002LangChain \u8207 Hugging Face \u7b49\u6846\u67b6\u53ef\u4ee5\u5c07\u5927\u578b\u6587\u4ef6\u62c6\u5206\u70ba\u66f4\u5c0f\u3001\u66f4\u6613\u6aa2\u7d22\u7684\u6587\u5b57\u5340\u584a\u3002<\/p>\n \u6587\u5b57\u5207\u5206\u7b56\u7565\u6703\u76f4\u63a5\u5f71\u97ff\u7cfb\u7d71\u7684\u6aa2\u7d22\u901f\u5ea6\u3002\u8a2d\u5b9a 20%\u201330% \u7684\u7247\u6bb5\u91cd\u758a\u6bd4\u4f8b\uff0c\u53ef\u4ee5\u5728\u5206\u6bb5\u908a\u754c\u4e0a\u4fdd\u7559\u5fc5\u8981\u7684\u4e0a\u4e0b\u6587\u3002<\/p>\n \n\n\n\n\n\n\n<\/colgroup>\n\n\n\n \u5207\u5206\u65b9\u5f0f<\/th>\n \u5178\u578b\u61c9\u7528<\/th>\n \u5efa\u8b70\u7247\u6bb5\u5927\u5c0f<\/th>\n \u5efa\u8b70\u91cd\u758a\u7bc4\u570d<\/th>\n \u6aa2\u7d22\u8868\u73fe<\/th>\n<\/tr>\n \u8a9e\u7fa9\u5207\u5206<\/td>\n \u79d1\u7814\u8ad6\u6587\u8207\u77e5\u8b58\u5eab<\/td>\n 150\u2013400 tokens<\/td>\n 0\u201330 tokens<\/td>\n \u900f\u904e\u4fdd\u6301\u8a9e\u7fa9\u4e00\u81f4\u6027\uff0c\u7372\u5f97\u6700\u9ad8\u7684\u6aa2\u7d22\u76f8\u95dc\u5ea6<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \ud83d\udca1 \u63d0\u793a\uff1a<\/strong>\u5c07\u8a9e\u7fa9\u5207\u5206\u8207\u8996\u7a97\u91cd\u758a\u7d50\u5408\u4f7f\u7528\uff0c\u53ef\u4ee5\u5728\u6aa2\u7d22\u904e\u7a0b\u4e2d\u964d\u4f4e\u5e7b\u89ba\u98a8\u96aa\u3002<\/p><\/blockquote>\n \u5728\u5b8c\u6210\u6587\u5b57\u5207\u5206\u5f8c\uff0c\u5d4c\u5165\u6a21\u578b\u6703\u5c07\u6587\u5b57\u5340\u584a\u8f49\u63db\u70ba\u6578\u5b78\u5411\u91cf\u3002\u7cfb\u7d71\u6703\u5c07\u9019\u4e9b\u5411\u91cf\u5132\u5b58\u5230\u5c08\u7528\u5411\u91cf\u8cc7\u6599\u5eab\u4e2d\u3002\u5c07\u5411\u91cf\u8cc7\u6599\u5eab\u90e8\u7f72\u5728\u672c\u5730\uff0c\u53ef\u4ee5\u5728\u77e5\u8b58\u5eab\u5c64\u9762\u5be6\u73fe\u5b8c\u6574\u7684\u7db2\u8def\u96b1\u79c1\u3002<\/p>\n \n\n\n\n\n<\/colgroup>\n\n\n\n\n\n\n \u5b89\u5168\u8207\u5408\u898f\u9762\u5411<\/th>\n \u81ea\u5efa \/ \u672c\u5730\u5411\u91cf\u5132\u5b58<\/th>\n \u96f2\u7aef\u5411\u91cf\u670d\u52d9<\/th>\n<\/tr>\n \u8cc7\u6599\u63a7\u5236\u8207\u66b4\u9732<\/strong><\/td>\n \u5c0d\u8cc7\u6599\u4f4d\u7f6e\u3001\u7db2\u8def\u908a\u754c\u8207\u64cd\u4f5c\u5177\u5b8c\u5168\u638c\u63a7\uff0c\u6d88\u9664\u5916\u90e8\u8cc7\u6599\u5916\u6d41\u9014\u5f91\u3002<\/td>\n \u5f15\u5165\u6f5b\u5728\u7684\u5916\u90e8\u50b3\u8f38\u98a8\u96aa\u3001\u88ab\u52d5\u96f2\u7aef\u4f9d\u8cf4\u8207\u8907\u96dc\u8a2d\u5b9a\u9700\u6c42\u3002<\/td>\n<\/tr>\n \u5408\u898f\u6846\u67b6\u5951\u5408\u5ea6<\/strong><\/td>\n \u900f\u904e\u5c07\u654f\u611f\u8cc7\u6599\u56b4\u683c\u4fdd\u7559\u5728\u672c\u5730\u57fa\u790e\u8a2d\u65bd\u4e2d\uff0c\u66f4\u6613\u6eff\u8db3 GDPR\u3001HIPAA\u3001FedRAMP\u3001GLBA\u3001APRA \u7b49\u56b4\u683c\u76e3\u7ba1\u8981\u6c42\u3002<\/td>\n \u9700\u8981\u6301\u7e8c\u5408\u898f\u5be9\u67e5\u3001\u8907\u96dc\u5408\u7d04\u5b89\u6392\u4ee5\u53ca\u5927\u91cf\u6838\u51c6\u6d41\u7a0b\u3002<\/td>\n<\/tr>\n \u7a3d\u6838\u8207\u5730\u7406\u908a\u754c<\/strong><\/td>\n \u900f\u904e\u6e05\u6670\u7684\u5be6\u9ad4 \/ \u5730\u7406\u908a\u754c\uff0c\u7c21\u5316\u5408\u898f\u7a3d\u6838\u6d41\u7a0b\uff0c\u65b9\u4fbf\u76e3\u7ba1\u6a5f\u69cb\u5224\u5b9a\u3002<\/td>\n \u9700\u8981\u8907\u96dc\u7684\u6cd5\u5f8b\u8207\u6280\u8853\u624b\u6bb5\u4f86\u6eff\u8db3\u8de8\u5883\u50b3\u8f38\u9650\u5236\u3002<\/td>\n<\/tr>\n \u98a8\u96aa\u9694\u96e2\u8207\u8ca0\u8f09\u5206\u96e2<\/strong><\/td>\n \u5141\u8a31\u5c07\u9ad8\u98a8\u96aa\u3001\u5f37\u76e3\u7ba1\u8cc7\u6599\u5b89\u5168\u5730\u7559\u5728\u672c\u5730\uff0c\u800c\u5c07\u975e\u654f\u611f\u4efb\u52d9\u90e8\u7f72\u5728\u5176\u4ed6\u74b0\u5883\u3002<\/td>\n \u5982\u679c\u6c92\u6709\u56b4\u683c\u7684\u7ba1\u8f44\u9650\u5236\uff0c\u6240\u6709\u5de5\u4f5c\u8ca0\u8f09\u90fd\u6703\u540c\u6642\u627f\u53d7\u516c\u6709\u96f2\u57fa\u790e\u8a2d\u65bd\u7684\u98a8\u96aa\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u9078\u64c7\u5408\u9069\u7684\u5411\u91cf\u5f15\u64ce\uff0c\u9700\u8981\u5728\u8a18\u61b6\u9ad4\u5360\u7528\u8207\u6aa2\u7d22\u901f\u5ea6\u4e4b\u9593\u9054\u6210\u5e73\u8861\u3002\u4f60\u7684\u786c\u9ad4\u898f\u683c\u6703\u6c7a\u5b9a\u5411\u91cf\u7d22\u5f15\u5728\u8655\u7406\u9ad8\u5bc6\u5ea6\u5411\u91cf\u6642\u7684\u6548\u7387\u3002<\/p>\n \n\n\n\n\n\n<\/colgroup>\n\n\n\n\n\n \u8cc7\u6599\u5eab<\/th>\n \u8a18\u61b6\u9ad4\u6548\u7387\u8207\u4f7f\u7528<\/th>\n \u7d22\u5f15\u901f\u5ea6\u8207\u5efa\u7f6e\u7279\u6027<\/th>\n RAG \u8207\u751f\u7522\u9069\u914d\u6027<\/th>\n<\/tr>\n Qdrant<\/strong><\/td>\n \u5728 HNSW \u7d22\u5f15\u4e0b\uff0c100 \u842c\u689d 768 \u7dad\u5411\u91cf\u7d04\u9700 1.4 GB \u8a18\u61b6\u9ad4\u3002<\/td>\n \u67e5\u8a62\u6548\u80fd\u5feb\u901f\uff08p95 \u5ef6\u9072\u7d04 8ms\uff09\uff0cHNSW \u7d22\u5f15\u5efa\u7f6e\u6642\u9593\u4e5f\u5f97\u5230\u6700\u4f73\u5316\u3002<\/td>\n \u975e\u5e38\u9069\u5408\u9700\u8981\u4f4e\u5ef6\u9072\u3001\u9ad8 QPS \u7684\u751f\u7522\u7d1a RAG \u6d41\u6c34\u7dda\u3002<\/td>\n<\/tr>\n pgvector<\/strong><\/td>\n HNSW \u7d22\u5f15\u76f8\u8f03 IVFFlat \u9700\u8981 3\u20134 \u500d\u8a18\u61b6\u9ad4\uff1b\u8a18\u61b6\u9ad4\u6709\u9650\u6642\uff0cIVFFlat \u66f4\u5408\u9069\u3002<\/td>\n HNSW \u7d22\u5f15\u5efa\u7f6e\u901f\u5ea6\u8f03\u6162\uff0c\u4e14\u8a18\u61b6\u9ad4\u958b\u92b7\u660e\u986f\u9ad8\u65bc IVFFlat\u3002<\/td>\n \u9069\u5408\u8207 Postgres \u6df1\u5ea6\u6574\u5408\u7684\u7cfb\u7d71\uff0c\u4f46\u5728\u7d22\u5f15\u5efa\u7f6e\u6642\u9700\u8981\u6b0a\u8861\u8a18\u61b6\u9ad4\u8207\u901f\u5ea6\u3002<\/td>\n<\/tr>\n ChromaDB<\/strong><\/td>\n \u66ab\u7121\u516c\u958b\u8a73\u7d30\u7684\u57fa\u6e96\u8a18\u61b6\u9ad4\u6578\u64da\u3002<\/td>\n \u4ee5\u672c\u5730\u884c\u7a0b\u65b9\u5f0f\u904b\u884c\uff0c\u7121\u9700\u984d\u5916\u5b89\u88dd\u8a2d\u5b9a\uff0c\u9069\u65bc\u5feb\u901f\u4e0a\u624b\u3002<\/td>\n \u975e\u5e38\u9069\u5408\u65e9\u671f RAG \u539f\u578b\u958b\u767c\uff0c\u4f46\u7f3a\u4e4f\u751f\u7522\u904b\u7dad\u80fd\u529b\uff08\u9ad8\u53ef\u7528\u3001\u76e3\u63a7\u7b49\uff09\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u4f60\u53ef\u4ee5\u900f\u904e\u81ea\u52d5\u5316\u4f4e\u7a0b\u5f0f\u78bc\u5de5\u5177\u7c21\u5316\u77e5\u8b58\u63a5\u5165\u6d41\u7a0b\u3002\u5c07 n8n \u6216 Flowise \u7b49\u7de8\u6392\u5668\u5bb9\u5668\u5316\uff0c\u53ef\u4ee5\u5728\u4e0d\u64b0\u5beb\u81ea\u8a02\u7a0b\u5f0f\u78bc\u7684\u524d\u63d0\u4e0b\uff0c\u5c07\u6587\u4ef6\u6536\u96c6\u5de5\u4f5c\u6d41\u7a0b\u8207\u672c\u5730\u5411\u91cf\u8cc7\u6599\u5eab\u6253\u901a\u3002<\/p>\n \n\n\n\n\n<\/colgroup>\n\n\n\n\n\n\n\n \u670d\u52d9 \/ \u5de5\u5177<\/th>\n \u5728\u5bb9\u5668\u4e2d\u7684\u89d2\u8272<\/th>\n \u5728\u6587\u4ef6\u8655\u7406\u8207 AI \u6d41\u7a0b\u4e2d\u7684\u4f5c\u7528<\/th>\n<\/tr>\n n8n<\/strong><\/td>\n \u4e3b\u7de8\u6392\u5668<\/td>\n \u4f4e\u7a0b\u5f0f\u78bc\u81ea\u52d5\u5316\u5e73\u81fa\uff0c\u7528\u65bc\u8a2d\u8a08\u5de5\u4f5c\u6d41\u7a0b\u4e26\u9023\u7d50\u6574\u5408\u7684 AI \u7bc0\u9ede\u3002<\/td>\n<\/tr>\n Flowise<\/strong><\/td>\n AI \u667a\u80fd\u9ad4\u5efa\u69cb\u5668<\/td>\n \u7528\u65bc\u69cb\u5efa\u8907\u96dc AI \u667a\u80fd\u9ad4\uff0c\u53ef\u7531 n8n \u89f8\u767c\uff0c\u6216\u53cd\u5411\u89f8\u767c n8n \u5de5\u4f5c\u6d41\u7a0b\uff08flowise.yourdomain.com<\/code>\uff09\u3002<\/td>\n<\/tr>\n Docling<\/strong><\/td>\n \u6587\u4ef6\u89e3\u6790\u5668<\/td>\n \u5c07 PDF\u3001DOCX\u3001XLSX\u3001\u5716\u7247\u7b49\u591a\u7a2e\u683c\u5f0f\u8f49\u70ba\u4e7e\u6de8\u7684 Markdown\/JSON\uff0c\u53ef\u900f\u904e REST API \u6574\u5408\u9032 n8n\u3002<\/td>\n<\/tr>\n \u5411\u91cf\u5132\u5b58\uff08Qdrant\/Weaviate\uff09<\/strong><\/td>\n \u77e5\u8b58\u5132\u5b58<\/td>\n \u5132\u5b58\u8207\u6aa2\u7d22\u5411\u91cf\u5d4c\u5165\uff0c\u7528\u65bc RAG\uff08\u6aa2\u7d22\u589e\u5f37\u751f\u6210\uff09\u4efb\u52d9\u3002<\/td>\n<\/tr>\n Ollama \/ LLMs<\/strong><\/td>\n \u63a8\u7406\u5f15\u64ce<\/td>\n \u4ee5\u672c\u5730\u6216 API \u65b9\u5f0f\u63d0\u4f9b\u8a9e\u8a00\u6a21\u578b\uff0c\u53ef\u76f4\u63a5\u5728 n8n \u6216 Flowise \u7bc0\u9ede\u4e2d\u8a2d\u5b9a\u547c\u53eb\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \n \u57fa\u65bc Docker \u7684\u5c01\u88dd\u74b0\u5883\uff0c\u70ba\u6240\u6709\u89e3\u6790\u76f8\u4f9d\u63d0\u4f9b\u81ea\u6d3d\u7684\u57f7\u884c\u7a7a\u9593\uff0c\u63d0\u5347\u53ef\u9760\u6027\u3002<\/li>\n \u9810\u5148\u69cb\u5efa\u7684\u5bb9\u5668\u53ef\u4ee5\u5728\u672c\u5730\u57fa\u790e\u8a2d\u65bd\u4e2d\u6253\u901a\u696d\u52d9\u81ea\u52d5\u5316\u8207\u81ea\u8a02\u667a\u80fd\u9ad4\u3002<\/li>\n \u5c08\u9580\u7684\u6587\u4ef6\u89e3\u6790\u5668\u53ef\u4ee5\u5728\u8cc7\u6599\u5165\u5eab\u524d\u81ea\u52d5\u6e05\u6d17\u539f\u59cb\u5167\u5bb9\u3002<\/li>\n<\/ul>\n \u9019\u4e9b\u81ea\u52d5\u5316\u5de5\u4f5c\u6d41\u7a0b\u53ef\u4ee5\u78ba\u4fdd\u77e5\u8b58\u9a45\u52d5\u667a\u80fd\u9ad4\u80fd\u5920\u5373\u6642\u7d22\u5f15\u65b0\u9032\u4f01\u696d\u6587\u4ef6\u3002<\/p>\n \u642d\u5efa Web \u4ecb\u9762\u8207\u4f7f\u7528\u8005\u7ba1\u7406<\/h3>\n \u73fe\u4ee3 Web \u4ecb\u9762\u53ef\u4ee5\u70ba\u5718\u968a\u6210\u54e1\u63d0\u4f9b\u76f4\u89c0\u7684\u5c0d\u8a71\u9ad4\u9a57\u3002Open WebUI \u63d0\u4f9b\u4e86\u985e\u4f3c ChatGPT \u7684\u7c21\u6f54\u5165\u53e3\uff0c\u540c\u6642\u4ecd\u7136\u4fdd\u6301\u5b8c\u6574\u7684\u4f7f\u7528\u8005\u96b1\u79c1\u3002<\/p>\n \u4f60\u53ef\u4ee5\u900f\u904e\u5e7e\u689d\u7c21\u55ae\u7684 Docker \u6307\u4ee4\uff0c\u5c07 Open WebUI \u9023\u63a5\u5230\u672c\u5730\u5bb9\u5668\u74b0\u5883\uff1a<\/p>\n \n\n\n\n\n<\/colgroup>\n\n\n\n\n\n\n\n\n \u90e8\u7f72\u60c5\u5883<\/th>\n \u786c\u9ad4 \/ \u8a2d\u5b9a<\/th>\n Docker \u6307\u4ee4<\/th>\n<\/tr>\n Ollama \u904b\u884c\u5728\u5bbf\u4e3b\u6a5f\u4e0a<\/td>\n \u9810\u8a2d\u8a2d\u5b9a<\/td>\n docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:\/app\/backend\/data --name open-webui --restart always ghcr.io\/open-webui\/open-webui:main<\/code><\/td>\n<\/tr>\n Ollama \u904b\u884c\u5728\u5bbf\u4e3b\u6a5f\u4e0a<\/td>\n Nvidia GPU \u652f\u63f4<\/td>\n docker run -d -p 3000:8080 --gpus all --add-host=host.docker.internal:host-gateway -v open-webui:\/app\/backend\/data --name open-webui --restart always ghcr.io\/open-webui\/open-webui:cuda<\/code><\/td>\n<\/tr>\n Ollama \u904b\u884c\u5728\u5916\u90e8\u4f3a\u670d\u5668<\/td>\n \u81ea\u8a02 URL\uff08OLLAMA_BASE_URL<\/code>\uff09<\/td>\n docker run -d -p 3000:8080 -e OLLAMA_BASE_URL=https:\/\/example.com -v open-webui:\/app\/backend\/data --name open-webui --restart always ghcr.io\/open-webui\/open-webui:main<\/code><\/td>\n<\/tr>\n \u6574\u5408 Ollama \u5bb9\u5668<\/td>\n GPU \u52a0\u901f<\/td>\n docker run -d -p 3000:8080 --gpus=all -v ollama:\/root\/.ollama -v open-webui:\/app\/backend\/data --name open-webui --restart always ghcr.io\/open-webui\/open-webui:ollama<\/code><\/td>\n<\/tr>\n \u6574\u5408 Ollama \u5bb9\u5668<\/td>\n \u50c5 CPU<\/td>\n docker run -d -p 3000:8080 -v ollama:\/root\/.ollama -v open-webui:\/app\/backend\/data --name open-webui --restart always ghcr.io\/open-webui\/open-webui:ollama<\/code><\/td>\n<\/tr>\n \u9023\u7dda\u6545\u969c\u6392\u67e5<\/td>\n Host \u7db2\u8def\u6a21\u5f0f<\/td>\n docker run -d --network=host -v open-webui:\/app\/backend\/data -e OLLAMA_BASE_URL=http:\/\/127.0.0.1:11434 --name open-webui --restart always ghcr.io\/open-webui\/open-webui:main<\/code><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \n \u7576\u76f4\u63a5\u5728\u5bbf\u4e3b\u6a5f\u4e0a\u7ba1\u7406 Ollama \u6642\uff0c\u53ef\u4ee5\u5c07 Docker \u5bb9\u5668\u8a2d\u5b9a\u70ba\u900f\u904e http:\/\/host.docker.internal:11434<\/code> \u9019\u500b\u7aef\u9ede\u8207\u4e4b\u901a\u4fe1\u3002<\/li>\n<\/ul>\n Web \u7ba1\u7406\u63a7\u5236\u53f0\u70ba\u4f60\u7684\u5718\u968a\u74b0\u5883\u63d0\u4f9b\u4e86\u5b8c\u5584\u7684\u5b58\u53d6\u63a7\u5236\u80fd\u529b\uff1a<\/p>\n \n \u57fa\u65bc\u89d2\u8272\u7684\u5b58\u53d6\u63a7\u5236\uff08RBAC\uff09<\/strong>\uff1a\u70ba\u6bcf\u500b\u65b0\u4f7f\u7528\u8005\u5e33\u865f\u5206\u914d\u4e0d\u540c\u7684\u7ba1\u7406\u8005\u6216\u4e00\u822c\u6b0a\u9650\u3002<\/li>\n \u5b58\u53d6\u9694\u96e2<\/strong>\uff1a\u6839\u64da\u7d44\u7e54\u8077\u8cac\uff0c\u70ba\u4e0d\u540c\u5718\u968a\u6388\u4e88\u50c5\u9650\u65bc\u5176\u6240\u5c6c\u77e5\u8b58\u5eab\u7d22\u5f15\u7684\u5b58\u53d6\u6b0a\u3002<\/li>\n \u81ea\u8a02\u6a21\u578b\u5206\u914d<\/strong>\uff1a\u5c07\u7279\u5b9a\u7684\u5c08\u696d\u667a\u80fd\u9ad4\u5206\u914d\u7d66\u5c0d\u61c9\u90e8\u9580\uff0c\u9650\u5236\u5c0d\u5168\u57df\u6a21\u578b\u7684\u6feb\u7528\u3002<\/li>\n \u4e92\u52d5\u5f0f\u6aa2\u7d22\u6e2c\u8a66<\/strong>\uff1a\u5141\u8a31\u5718\u968a\u6210\u54e1\u76f4\u63a5\u8207\u672c\u5730 AI \u804a\u5929\u52a9\u624b\u5c0d\u8a71\uff0c\u9a57\u8b49\u6587\u4ef6\u6aa2\u7d22\u8868\u73fe\u3002<\/li>\n<\/ol>\n \u9019\u500b\u4ecb\u9762\u5c64\u8b93\u4f60\u7684\u79c1\u6709\u4f3a\u670d\u5668\u57fa\u790e\u8a2d\u65bd\u5177\u5099\u4f01\u696d\u7d1a AI \u804a\u5929\u52a9\u624b\u80fd\u529b\u3002\u4f7f\u7528\u8005\u53ef\u4ee5\u8f15\u9b06\u67e5\u8a62\u8907\u96dc\u7684\u5167\u90e8\u77e5\u8b58\uff0c\u540c\u6642\u6574\u500b\u654f\u611f\u6a94\u6848\u9ad4\u7cfb\u4ecd\u7531\u4f60\u5b8c\u5168\u638c\u63a7\u3002\u900f\u904e\u90e8\u7f72\u5c08\u9580\u7684\u77e5\u8b58\u9a45\u52d5\u667a\u80fd\u9ad4\uff0c\u4f60\u53ef\u4ee5\u628a\u539f\u59cb\u4f01\u696d\u8cc7\u8a0a\u8f49\u5316\u70ba\u6574\u500b\u5718\u968a\u96a8\u6642\u53ef\u7528\u7684 AI \u670d\u52d9\u53f0\u3002\u81ea\u6cbb AI \u667a\u80fd\u9ad4\u5728\u672c\u5730\u8655\u7406\u5167\u90e8\u641c\u5c0b\u8acb\u6c42\uff0c\u78ba\u4fdd\u4f01\u696d\u77e5\u8b58\u5eab\u4e0d\u6703\u5c07\u6a5f\u5bc6\u8cc7\u7522\u66b4\u9732\u7d66\u516c\u6709\u96f2\u74b0\u5883\u3002<\/p>\n \u5b89\u5168\u7db2\u8def\u8a2d\u5b9a\u8207\u6700\u4f73\u5316<\/strong><\/h2>\n \u8a2d\u5b9a Nginx \u53cd\u5411\u4ee3\u7406\u8207 SSL<\/h3>\n \u5728\u5c0d\u5916\u767c\u5e03 Web \u670d\u52d9\u524d\uff0c\u4f60\u5fc5\u9808\u5148\u78ba\u4fdd\u7db2\u8def\u67b6\u69cb\u662f\u5b89\u5168\u7684\u3002Nginx \u53cd\u5411\u4ee3\u7406\u53ef\u4ee5\u4fdd\u8b77\u4f60\u7684 AI \u61c9\u7528\u514d\u65bc\u88ab\u5916\u90e8\u76f4\u63a5\u5b58\u53d6\u3002\u5b83\u8ca0\u8cac\u8655\u7406\u6240\u6709\u5c0d\u5916 Web \u8acb\u6c42\uff0c\u4e26\u5b89\u5168\u5730\u5c07\u5176\u8def\u7531\u5230\u672c\u5730\u667a\u80fd\u9ad4\u3002\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u5728\u56de\u61c9\u6642\u5f80\u5f80\u4ee5\u4e32\u6d41\u65b9\u5f0f\u9010\u5b57\u8f38\u51fa\uff0c\u6642\u9593\u53ef\u80fd\u8f03\u9577\u3002\u9810\u8a2d\u4f3a\u670d\u5668\u903e\u6642\u8a2d\u5b9a\u53ef\u80fd\u6703\u5728\u9577\u56de\u61c9\u904e\u7a0b\u4e2d\u4e2d\u65b7\u9023\u7dda\u3002\u56e0\u6b64\uff0c\u4f60\u9700\u8981\u8abf\u6574\u7279\u5b9a\u903e\u6642\u53c3\u6578\uff0c\u4ee5\u78ba\u4fdd\u6bcf\u4f4d\u6388\u6b0a\u4f7f\u7528\u8005\u7684\u9577\u9023\u7dda\u7a69\u5b9a\u6027\u3002<\/p>\n \n\n\n\n\n\n<\/colgroup>\n\n\n\n\n \u8a2d\u5b9a\u6307\u4ee4<\/th>\n \u89e3\u6c7a\u554f\u984c<\/th>\n \u9810\u8a2d\u503c<\/th>\n \u63a8\u85a6\u8a2d\u5b9a\u8207\u4f5c\u7528<\/th>\n<\/tr>\n proxy_read_timeout<\/code><\/td>\n \u9577\u56de\u61c9\u6642\u51fa\u73fe 504 Gateway Timeout<\/td>\n 60 \u79d2<\/td>\n \u8a2d\u5b9a\u70ba 600s<\/code>\uff08\u6216\u66f4\u9577\uff09\uff0c\u4ee5\u5ef6\u9577\u8b80\u53d6\u903e\u6642\u6642\u9593\uff0c\u9069\u61c9\u9577\u6642\u9593\u63a8\u7406\uff1b\u6bcf\u6536\u5230\u65b0\u7684\u8cc7\u6599\u5340\u584a\u6642\uff0c\u8a72\u8a08\u6642\u5668\u6703\u88ab\u91cd\u7f6e\u3002<\/td>\n<\/tr>\n proxy_send_timeout<\/code><\/td>\n \u4e32\u6d41\u50b3\u8f38\u4e2d\u9014\u9023\u7dda\u88ab\u91cd\u7f6e<\/td>\n \u9810\u8a2d\u8f03\u77ed\u6642\u9593<\/td>\n \u9069\u7576\u589e\u52a0\u8a72\u903e\u6642\u6642\u9577\uff0c\u4ee5\u4fdd\u6301\u7528\u6236\u7aef\u9023\u7dda\u5b58\u6d3b\uff0c\u9632\u6b62\u5728\u6162\u901f\u4e32\u6d41\u8f38\u51fa\u6642\u88ab\u904e\u65e9\u4e2d\u65b7\u3002<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u5408\u7406\u7684\u4ee3\u7406\u8a2d\u5b9a\u53ef\u4ee5\u5728\u57f7\u884c\u8907\u96dc\u5167\u90e8\u77e5\u8b58\u5eab\u67e5\u8a62\u6642\uff0c\u78ba\u4fdd\u9023\u7dda\u7684\u7a69\u5b9a\u6027\u3002<\/p>\n \u91dd\u5c0d\u7f8e\u570b\u8def\u7dda\u7684\u5ef6\u9072\u6700\u4f73\u5316<\/h3>\n \u7f8e\u570b\u4f3a\u670d\u5668\u7684\u5be6\u9ad4\u4f4d\u7f6e\u6703\u76f4\u63a5\u5f71\u97ff\u8cc7\u6599\u50b3\u8f38\u901f\u5ea6\u3002\u70ba\u79c1\u6709\u77e5\u8b58\u7cfb\u7d71\u9078\u64c7\u5408\u9069\u7684\u6a5f\u623f\u5340\u57df\uff0c\u53ef\u4ee5\u986f\u8457\u6539\u5584\u67e5\u8a62\u6548\u80fd\u3002<\/p>\n \n\n\n\n\n\n<\/colgroup>\n\n\n\n\n\n \u7f8e\u570b\u4f3a\u670d\u5668\u5340\u57df<\/th>\n \u7f8e\u570b\u672c\u571f\u5ef6\u9072\u8868\u73fe<\/th>\n \u8de8\u5927\u897f\u6d0b \/ \u6d77\u5916\u5f71\u97ff<\/th>\n \u5efa\u8b70\u4f7f\u7528\u65cf\u7fa4<\/th>\n<\/tr>\n \u7f8e\u570b\u6771\u90e8<\/strong>\uff08NY, NJ\uff09<\/td>\n \u6771\u5cb8\u50b3\u8f38\u8868\u73fe\u51fa\u8272\uff08\u81f3 DC \u7d04 15\u201325ms\uff09\uff1b\u5230\u897f\u5cb8\u5ef6\u9072\u4e00\u822c\uff0860\u201375ms\uff09<\/td>\n \u7531\u65bc\u76f4\u9023\u8de8\u5927\u897f\u6d0b\u5149\u7e96\uff0c\u5c0d\u6b50\u6d32\u5b58\u53d6\u5ef6\u9072\u8f03\u4f4e<\/td>\n \u9762\u5411\u7f8e\u570b\u6771\u90e8\u6216\u540c\u6642\u670d\u52d9\u7f8e\u6771\u8207\u6b50\u6d32\u4f7f\u7528\u8005\u7684\u4f01\u696d<\/td>\n<\/tr>\n \u7f8e\u570b\u4e2d\u90e8<\/strong>\uff08\u829d\u52a0\u54e5\u3001\u9054\u62c9\u65af\uff09<\/td>\n \u5c0d\u7f8e\u570b\u5404\u5340\u57df\u63d0\u4f9b\u8f03\u5e73\u8861\u7684\u5168\u5883\u5ef6\u9072\u8868\u73fe<\/td>\n \u829d\u52a0\u54e5\u662f\u5317\u7f8e\u91cd\u8981\u7684\u8de8\u5927\u897f\u6d0b\u5149\u7e96\u6a1e\u7d10\uff0c\u6709\u5229\u65bc\u964d\u4f4e\u6b50\u6d32\u5b58\u53d6\u5ef6\u9072<\/td>\n \u9762\u5411\u5168\u7f8e\u3001\u7121\u660e\u986f\u5340\u57df\u6d41\u91cf\u50be\u659c\u7684\u5168\u570b\u6027\u61c9\u7528<\/td>\n<\/tr>\n \u7f8e\u570b\u897f\u90e8<\/strong>\uff08\u6d1b\u6749\u78ef\u3001\u77fd\u8c37\uff09<\/td>\n \u672c\u5730\u5340\u5b58\u53d6\u5ef6\u9072\u6975\u4f73\uff1b\u5230\u7f8e\u570b\u6771\u90e8\u5ef6\u9072\u4e00\u822c\uff0860\u201375ms\uff09<\/td>\n \u70ba\u74b0\u592a\u5e73\u6d0b\u5730\u5340\u63d0\u4f9b\u6700\u4f73\u7684\u8de8\u592a\u5e73\u6d0b\u5149\u7e96\u8def\u7531\uff08\u5c0d\u6b50\u6d32\u76f8\u5c0d\u4e0d\u5229\uff09<\/td>\n \u9762\u5411\u7f8e\u570b\u897f\u5cb8\u8207\u74b0\u592a\u5e73\u6d0b\u4f7f\u7528\u8005\u7684\u696d\u52d9<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u4f60\u53ef\u4ee5\u63a1\u7528\u4ee5\u4e0b\u95dc\u9375\u7db2\u8def\u8def\u7531\u7b56\u7565\u4f86\u63d0\u5347\u50b3\u8f38\u6548\u80fd\uff1a<\/p>\n \n \u5730\u7406\u8ddd\u96e2\u8207\u5149\u7e96\u6a1e\u7d10<\/strong>\uff1a\u829d\u52a0\u54e5\u662f\u5317\u7f8e\u8de8\u5927\u897f\u6d0b\u6d77\u5e95\u5149\u7e9c\u7684\u6838\u5fc3\u6a1e\u7d10\u4e4b\u4e00\uff0c\u56e0\u6b64\u76f8\u8f03\u65bc\u5167\u9678\u57ce\u5e02\uff08\u4f8b\u5982 Roswell\uff09\uff0c\u5c0d\u6b50\u6d32\u7684\u5ef6\u9072\u66f4\u4f4e\u3002<\/li>\n \u57fa\u790e\u8a2d\u65bd\u63a5\u5165<\/strong>\uff1a\u6cbf\u4e3b\u5e79\u5149\u7e96\u901a\u9053\u90e8\u7f72\u4f3a\u670d\u5668\uff0c\u53ef\u4ee5\u6e1b\u5c11\u6b50\u6d32\u8acb\u6c42\u7684\u4e2d\u9593\u8df3\u6578\uff0c\u76f8\u8f03\u65bc\u9060\u96e2\u6771\u5cb8\u767b\u9678\u9ede\u7684\u5357\u90e8\u5167\u9678\u5730\u5340\uff0c\u5ef6\u9072\u66f4\u5177\u512a\u52e2\u3002<\/li>\n \u908a\u7de3\u904b\u7b97<\/strong>\uff1a\u5728\u9760\u8fd1\u7d42\u7aef\u4f7f\u7528\u8005\u7684\u5340\u57df\u90e8\u7f72\u63a8\u7406\u7bc0\u9ede\uff0c\u53ef\u4ee5\u7e2e\u77ed\u5be6\u9ad4\u50b3\u8f38\u8ddd\u96e2\u8207\u5f80\u8fd4\u6642\u9593\uff08RTT\uff09\uff0c\u5c0d\u65bc\u5373\u6642\u61c9\u7528\uff0c\u56de\u61c9\u901f\u5ea6\u6700\u591a\u53ef\u63d0\u5347\u7d04 50%\u3002<\/li>\n \u73fe\u4ee3\u8cc7\u6599\u50b3\u8f38\u5354\u5b9a<\/strong>\uff1a\u4f7f\u7528 gRPC\u3001HTTP\/2 \u7b49\u73fe\u4ee3\u5354\u5b9a\u53d6\u4ee3\u50b3\u7d71 HTTP\uff0c\u53ef\u4ee5\u964d\u4f4e\u7db2\u8def\u958b\u92b7\uff0c\u5c07\u56de\u61c9\u6642\u9593\u63d0\u5347\u7d04 30%\u3002<\/li>\n<\/ul>\n \u900f\u904e\u9019\u4e9b\u90e8\u7f72\u8207\u8def\u7531\u7b56\u7565\uff0c\u4f60\u7684\u81ea\u6cbb\u667a\u80fd\u9ad4\u53ef\u4ee5\u5feb\u901f\u67e5\u8a62\u5167\u90e8\u77e5\u8b58\u5f15\u64ce\uff0c\u4e26\u4ee5\u8f03\u9ad8\u7cbe\u5ea6\u5c07\u9818\u57df\u77e5\u8b58\u56de\u50b3\u7d66\u7b2c\u4e00\u7dda\u5718\u968a\uff0c\u540c\u6642\u6301\u7e8c\u4fdd\u8b77\u4f01\u696d\u77e5\u8b58\u8cc7\u7522\u3002 \n\u81f3\u6b64\uff0c\u4f60\u5df2\u7d93\u5728\u4f3a\u670d\u5668\u4e0a\u638c\u63a7\u4e86\u4e00\u5957\u5b8c\u6574\u7684\u672c\u5730\u90e8\u7f72\u67b6\u69cb\u3002\u4f60\u5c07 AI \u804a\u5929\u52a9\u624b\u8207\u672c\u5730\u5411\u91cf\u5132\u5b58\u7d50\u5408\u5728\u4e00\u8d77\uff0c\u4f7f AI \u667a\u80fd\u9ad4\u53ef\u4ee5\u5728\u4e0d\u4f9d\u8cf4\u7b2c\u4e09\u65b9\u7684\u60c5\u6cc1\u4e0b\u8655\u7406\u4f7f\u7528\u8005\u67e5\u8a62\uff0c\u4e26\u9ad8\u6548\u5206\u6790\u672c\u5730\u6a94\u6848\u3002<\/p>\n \ud83d\udca1 \u63d0\u793a\uff1a<\/strong>\u8acb\u5b9a\u671f\u70ba\u5411\u91cf\u5eab\u57f7\u884c\u5099\u4efd\uff0c\u4ee5\u4fbf\u5728\u7cfb\u7d71\u5347\u7d1a\u671f\u9593\u4fdd\u8b77\u7d22\u5f15\u7d00\u9304\u3002<\/p><\/blockquote>\n \u4f60\u53ef\u4ee5\u900f\u904e\u589e\u52a0\u66f4\u591a\u81ea\u52d5\u5316\u5de5\u4f5c\u6d41\u7a0b\u4f86\u64f4\u5c55\u9019\u5957\u7cfb\u7d71\u3002\u540c\u6642\uff0c\u5f8c\u7e8c\u4e5f\u53ef\u4ee5\u5347\u7d1a\u786c\u9ad4\u914d\u7f6e\uff0c\u4ee5\u904b\u884c\u66f4\u5927\u53c3\u6578\u898f\u6a21\u7684\u667a\u80fd\u9ad4\uff0c\u4e26\u6301\u7e8c\u64f4\u5145\u4e2d\u592e\u77e5\u8b58\u5eab\u3002<\/p>\n \n\n\n\n\n<\/colgroup>\n\n\n\n\n\n \u6210\u672c\u7dad\u5ea6<\/th>\n \u81ea\u5efa AI \u77e5\u8b58\u5eab<\/th>\n \u8a17\u7ba1\u96f2\u7aef LLM \u5e73\u81fa<\/th>\n<\/tr>\n \u4e3b\u8981\u6210\u672c\u4f86\u6e90<\/strong><\/td>\n \u524d\u671f\u786c\u9ad4\u63a1\u8cfc\u8207\u6301\u7e8c\u7684\u96fb\u529b\u6d88\u8017<\/td>\n \u4f9d token \u8a08\u8cbb\u7684 API \u4f7f\u7528\u6210\u672c<\/td>\n<\/tr>\n \u908a\u969b\u55ae\u6b21\u67e5\u8a62\u6210\u672c<\/strong><\/td>\n \u6975\u4f4e\uff08\u4e3b\u8981\u662f\u96fb\u529b\u6d88\u8017\uff09<\/td>\n \u6bcf\u8655\u7406\u4e00\u500b token \u90fd\u6703\u7522\u751f\u6301\u7e8c\u8cbb\u7528<\/td>\n<\/tr>\n \u9810\u4f30\u6708\u5ea6\u958b\u652f<\/strong><\/td>\n \u6578\u767e\u7f8e\u5143\uff08\u4e3b\u8981\u70ba\u96fb\u8cbb\uff09<\/td>\n \u5c0d\u65bc\u91cd\u5ea6\u8acb\u6c42\u60c5\u5883\uff0c\u53ef\u80fd\u6bcf\u6708\u9700 1 \u842c\u81f3 5 \u842c\u7f8e\u5143\u751a\u81f3\u66f4\u9ad8<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n \u69cb\u5efa\u79c1\u6709\u77e5\u8b58\u5eab\u53ef\u4ee5\u5728\u4fdd\u8b49\u9ad8\u6548\u80fd\u7684\u540c\u6642\uff0c\u964d\u4f4e\u5c0d\u6d3b\u8e8d\u4f01\u696d\u5718\u968a\u7684\u71df\u904b\u6210\u672c\u3002<\/p>\n \u5e38\u898b\u554f\u984c FAQ<\/strong><\/h2>\n \u5982\u4f55\u4fdd\u8b77\u79c1\u6709\u77e5\u8b58\u5eab\u4e2d\u7684\u4f7f\u7528\u8005\u8cc7\u6599\uff1f<\/h3>\n \u4f60\u53ef\u4ee5\u5c07\u6240\u6709\u6a94\u6848\u5132\u5b58\u5728\u672c\u5730\u7cfb\u7d71\u4e2d\u3002\u4f3a\u670d\u5668\u6703\u5c0d\u4f01\u696d\u6a94\u6848\u9032\u884c\u5b8c\u6574\u52a0\u5bc6\uff0c\u4e26\u900f\u904e\u7db2\u8def\u9694\u96e2\u78ba\u4fdd\u8cc7\u6599\u4e0d\u6703\u5916\u6d29\u3002\u6bcf\u4f4d\u4f7f\u7528\u8005\u5728\u67e5\u8a62\u9818\u57df\u77e5\u8b58\u6642\uff0c\u90fd\u53ef\u4ee5\u5728\u56b4\u683c\u7684\u96b1\u79c1\u4fdd\u8b77\u4e0b\u5b8c\u6210\u64cd\u4f5c\u3002<\/p>\n \u672c\u5730 AI \u667a\u80fd\u9ad4\u662f\u5426\u80fd\u8655\u7406\u8907\u96dc\u6587\u4ef6\u683c\u5f0f\uff1f<\/h3>\n \u53ef\u4ee5\u3002\u5c08\u9580\u7684\u6587\u4ef6\u89e3\u6790\u5668\u53ef\u4ee5\u81ea\u52d5\u6e05\u6d17\u6587\u5b57\u8207\u8868\u683c\u8cc7\u6599\u3002\u9019\u4e9b\u8edf\u9ad4\u667a\u80fd\u9ad4\u6703\u628a\u539f\u59cb\u6587\u5b57\u8f49\u63db\u70ba\u6578\u503c\u5411\u91cf\uff0c\u9ad8\u6548\u8655\u7406\u8f38\u5165\u8cc7\u6599\u6d41\uff0c\u4e26\u5728\u5167\u90e8\u77e5\u8b58\u4e2d\u641c\u5c0b\u7b54\u6848\uff0c\u5f9e\u800c\u751f\u6210\u5177\u6709\u4e0a\u4e0b\u6587\u7684\u5feb\u901f\u56de\u61c9\u3002<\/p>\n \u70ba\u4ec0\u9ebc\u8981\u9078\u64c7\u81ea\u5efa\u804a\u5929\u6a5f\u5668\u4eba\uff0c\u800c\u4e0d\u662f\u96f2\u7aef API\uff1f<\/h3>\n \u5728\u9ad8\u8ca0\u8f09\u60c5\u5883\u4e0b\uff0c\u81ea\u5efa\u53ef\u4ee5\u5fb9\u5e95\u907f\u514d\u4f9d token \u8a08\u8cbb\u7684 API \u6210\u672c\uff0c\u540c\u6642\u8b93\u4f60\u5b8c\u5168\u638c\u63a7\u6548\u80fd\u8868\u73fe\u3002\u57fa\u65bc\u7f8e\u570b\u672c\u5730\u5be6\u9ad4\u4f3a\u670d\u5668\u7684\u65b9\u6848\u53ef\u4ee5\u5728\u9ad8\u5cf0\u6642\u6bb5\u4fdd\u6301\u53ef\u9810\u671f\u7684\u5ef6\u9072\u3002\u6b64\u5916\uff0c\u79c1\u6709\u804a\u5929\u6a5f\u5668\u4eba\u6703\u5c07\u6a5f\u5bc6\u77e5\u8b58\u8207\u7b2c\u4e09\u65b9\u670d\u52d9\u5b8c\u5168\u9694\u96e2\u3002<\/p>\n \u600e\u6a23\u7684\u786c\u9ad4\u53ef\u4ee5\u8b93 AI \u804a\u5929\u52a9\u624b\u904b\u884c\u9806\u66a2\uff1f<\/h3>\n \u4e00\u5f35\u64c1\u6709 16GB \u986f\u793a\u8a18\u61b6\u9ad4\u7684\u73fe\u4ee3 GPU \u5373\u53ef\u8f15\u9b06\u904b\u884c\u5c0f\u578b\u6a21\u578b\u3002\u82e5\u50c5\u7528 CPU \u63a8\u7406\uff0c\u5247\u5efa\u8b70\u81f3\u5c11\u914d\u5099 64GB \u8a18\u61b6\u9ad4\u3002\u9019\u985e\u786c\u9ad4\u914d\u7f6e\u8db3\u4ee5\u652f\u6490 AI \u804a\u5929\u52a9\u624b\u7a69\u5b9a\u904b\u884c\uff0c\u5e6b\u52a9\u4f60\u5c07\u539f\u59cb\u6a94\u6848\u8f49\u5316\u70ba\u53ef\u76f4\u63a5\u4f7f\u7528\u7684\u77e5\u8b58\u8cc7\u7522\u3002[\/vc_column_text][\/vc_column][\/vc_row]<\/p>\n<\/div>","protected":false},"excerpt":{"rendered":" [vc_row][vc_column][vc_column_text css=””]\u4f60 […]<\/p>\n Read More…<\/a><\/p>\n","protected":false},"author":1,"featured_media":33754,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[4815],"tags":[12534,12535,12536,12537,8703,5229],"class_list":["post-33757","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-america-dedicated-server-tc","tag-ai-chat-assistant","tag-knowledge-base","tag-chatbot-deployment","tag-custom-ai","tag-us-server-tc-2","tag-us-servers-tc"],"acf":[],"yoast_head":"\n