2.2 GPU\u9a71\u52a8\u4e0e\u4f9d\u8d56\u5e93\u5b89\u88c5<\/h3>\n

\n
1. \u9a71\u52a8\u7248\u672c\u5339\u914d\uff1a\u6839\u636eGPU\u578b\u53f7\u9009\u62e9\u5bf9\u5e94\u7684NVIDIA\u9a71\u52a8\u7248\u672c\u3002\u9700\u907f\u514d\u9a71\u52a8\u7248\u672c\u8fc7\u9ad8\u6216\u8fc7\u4f4e\u5bfc\u81f4\u7684\u6846\u67b6\u8fd0\u884c\u9519\u8bef\uff0c\u53ef\u53c2\u8003NVIDIA\u5b98\u65b9\u6587\u6863\u4e2d\u7684\u9a71\u52a8-GPU\u517c\u5bb9\u6027\u5217\u8868\u3002<\/li>\n
2. \u6838\u5fc3\u4f9d\u8d56\u5e93\u5b89\u88c5\uff1a\n
   \n
   • \u5b89\u88c5CUDA\u5de5\u5177\u5305\uff1a\u9009\u62e9\u4e0e\u76ee\u6807AI\u6846\u67b6\u517c\u5bb9\u7684CUDA\u7248\u672c\u3002\u5efa\u8bae\u4f7f\u7528\u65e5\u672c\u672c\u5730\u955c\u50cf\u6e90\uff08\u5982\u4e1c\u4eac\u5de5\u4e1a\u5927\u5b66\u955c\u50cf\u6e90\uff09\uff0c\u63d0\u5347\u4e0b\u8f7d\u901f\u5ea6\u3002<\/li>\n
   • \u5b89\u88c5cuDNN\uff1a\u4e0b\u8f7d\u4e0e\u5df2\u5b89\u88c5CUDA\u5de5\u5177\u5305\u7248\u672c\u5339\u914d\u7684cuDNN\uff0c\u5e76\u6b63\u786e\u914d\u7f6e\u73af\u5883\u53d8\u91cf\u3002<\/li>\n<\/ul>\n<\/li>\n
   • \u5b89\u88c5\u9a8c\u8bc1\uff1a\u5728\u7ec8\u7aef\u6267\u884c nvidia-smi<\/code> \u547d\u4ee4\u3002\u82e5\u547d\u4ee4\u8f93\u51fa\u4e2d\u6b63\u786e\u663e\u793aGPU\u578b\u53f7\u3001\u9a71\u52a8\u7248\u672c\u53caCUDA\u7248\u672c\uff0c\u5219\u8bf4\u660e\u9a71\u52a8\u5b89\u88c5\u6210\u529f\u3002<\/li>\n<\/ol>\n

     [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””]<\/p>\n

     \u4e09\u3001\u65e5\u672cGPU\u670d\u52a1\u5668\u4e3b\u6d41AI\u6846\u67b6\u5206\u6b65\u914d\u7f6e\u6307\u5357<\/strong><\/h2>\n

     \u672c\u8282\u9488\u5bf9\u76ee\u524d\u5e94\u7528\u6700\u5e7f\u6cdb\u7684\u4e24\u5927AI\u6846\u67b6\u2014\u2014TensorFlow\u4e0ePyTorch\uff0c\u7ed3\u5408\u65e5\u672cGPU\u670d\u52a1\u5668\u7684\u7f51\u7edc\u73af\u5883\u4e0e\u786c\u4ef6\u7279\u6027\uff0c\u63d0\u4f9b\u8be6\u7ec6\u53ef\u843d\u5730\u7684\u914d\u7f6e\u65b9\u6848\u3002<\/p>\n

     3.1 TensorFlow GPU\u914d\u7f6e<\/h3>\n

     \n
     1. \u5b89\u88c5\u65b9\u5f0f\uff1a\u63a8\u8350\u4f7f\u7528Conda\u800c\u975epip\u8fdb\u884c\u5b89\u88c5\uff0c\u501f\u52a9\u73af\u5883\u9694\u79bb\u529f\u80fd\uff0c\u907f\u514d\u4e0d\u540c\u6846\u67b6\u4e0e\u4f9d\u8d56\u5e93\u4e4b\u95f4\u7684\u7248\u672c\u51b2\u7a81\u3002\u53ef\u901a\u8fc7\u65e5\u672c\u672c\u5730\u955c\u50cf\u6e90\u5b89\u88c5Miniconda\u3002<\/li>\n
     2. \u6838\u5fc3\u6b65\u9aa4\uff1a\n
        \n
        • \u521b\u5efa\u4e13\u5c5eConda\u73af\u5883\uff1aconda create -n tf-gpu python=3.9<\/code><\/li>\n
        • \u6fc0\u6d3b\u73af\u5883\uff1aconda activate tf-gpu<\/code><\/li>\n
        • \u5b89\u88c5TensorFlow GPU\u7248\u672c\uff1a\u4f7f\u7528\u4e0eCUDA\u7248\u672c\u9002\u914d\u7684\u5b98\u65b9\u547d\u4ee4\u8fdb\u884c\u5b89\u88c5\uff0c\u540c\u65f6\u914d\u7f6e\u65e5\u672cPyPI\u955c\u50cf\u6e90\uff0c\u89e3\u51b3\u6d77\u5916\u4e0b\u8f7d\u901f\u5ea6\u6162\u7684\u95ee\u9898\u3002<\/li>\n
        • \u914d\u7f6e\u73af\u5883\u53d8\u91cf\uff1a\u5c06 LD_LIBRARY_PATH<\/code> \u73af\u5883\u53d8\u91cf\u6307\u5411CUDA\u4e0ecuDNN\u7684\u5e93\u6587\u4ef6\u76ee\u5f55\u3002<\/li>\n<\/ul>\n<\/li>\n
        • \u914d\u7f6e\u9a8c\u8bc1\uff1a\u8fd0\u884c\u4e00\u6bb5\u7b80\u5355\u7684TensorFlow\u4ee3\u7801\uff0c\u68c0\u67e5GPU\u662f\u5426\u88ab\u6b63\u5e38\u8bc6\u522b\u3002\u793a\u4f8b\u4ee3\u7801\u5982\u4e0b\uff1a\n

          import tensorflow as tf\r\nprint(tf.config.list_physical_devices('GPU'))<\/code><\/pre>\n
          \u82e5\u4ee3\u7801\u8f93\u51faGPU\u8bbe\u5907\u4fe1\u606f\uff0c\u5219\u8bf4\u660e\u914d\u7f6e\u6210\u529f\u3002<\/li>\n<\/ol>\n
          3.2 PyTorch GPU\u914d\u7f6e<\/h3>\n
          \n
          \u7248\u672c\u9009\u62e9\uff1a\u9700\u4e25\u683c\u4fdd\u8bc1PyTorch\u7248\u672c\u4e0e\u5df2\u5b89\u88c5\u7684CUDA\u7248\u672c\u76f8\u5339\u914d\u3002\u53ef\u53c2\u8003PyTorch\u5b98\u65b9\u63d0\u4f9b\u7684\u7248\u672c\u517c\u5bb9\u6027\u5bf9\u7167\u8868\uff0c\u907f\u514d\u51fa\u73b0\u4e0d\u517c\u5bb9\u95ee\u9898\u3002<\/li>\n
          \u5b89\u88c5\u4f18\u5316\uff1a\u4f7f\u7528\u65e5\u672c\u672c\u5730\u7684PyPI\u6216Conda\u955c\u50cf\u6e90\u52a0\u901f\u5b89\u88c5\u6d41\u7a0b\uff0c\u964d\u4f4e\u8de8\u533a\u57df\u6570\u636e\u4f20\u8f93\u5e26\u6765\u7684\u5ef6\u8fdf\u3002<\/li>\n
          \u529f\u80fd\u6d4b\u8bd5\uff1a\u8fd0\u884c\u4e00\u4e2a\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc\u8bad\u7ec3\u4efb\u52a1\uff0c\u9a8c\u8bc1GPU\u7b97\u529b\u7684\u5229\u7528\u7387\u3002\u4f8b\u5982\uff0c\u5728\u6837\u672c\u6570\u636e\u96c6\u4e0a\u8bad\u7ec3\u4e00\u4e2a\u57fa\u7840CNN\u6a21\u578b\uff0c\u5e76\u901a\u8fc7 nvidia-smi<\/code> \u547d\u4ee4\u76d1\u63a7GPU\u4f7f\u7528\u7387\u3002<\/li>\n<\/ol>\n
          3.3 \u5c0f\u4f17AI\u6846\u67b6\u914d\u7f6e\u6280\u5de7\uff08\u5982MindSpore\u3001MXNet\uff09<\/h3>\n
          \n
          \u9a71\u52a8\u517c\u5bb9\u6027\u6ce8\u610f\u4e8b\u9879\uff1a\u5c0f\u4f17\u6846\u67b6\u5bf9\u9a71\u52a8\u7248\u672c\u7684\u6700\u4f4e\u8981\u6c42\u53ef\u80fd\u4e0e\u4e3b\u6d41\u6846\u67b6\u5b58\u5728\u5dee\u5f02\uff0c\u9700\u7279\u522b\u7559\u610f\u3002<\/li>\n
          \u6545\u969c\u6392\u67e5\u65b9\u6cd5\uff1a\u82e5\u5728\u65e5\u672c\u670d\u52a1\u5668\u4e0a\u5b89\u88c5\u5c0f\u4f17\u6846\u67b6\u65f6\u9047\u5230\u95ee\u9898\uff0c\u53ef\u67e5\u9605\u5b98\u65b9\u6587\u6863\u4e0e\u793e\u533a\u8bba\u575b\u3002\u540c\u65f6\uff0c\u53ef\u501f\u52a9\u672c\u5730\u5316\u6280\u672f\u652f\u6301\uff0c\u5feb\u901f\u89e3\u51b3\u95ee\u9898\u3002<\/li>\n<\/ul>\n
          [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””]<\/p>\n
          \u56db\u3001\u914d\u7f6e\u9a8c\u8bc1\u4e0e\u6027\u80fd\u6d4b\u8bd5<\/strong><\/h2>\n
          \u5b8c\u6210AI\u6846\u67b6\u914d\u7f6e\u540e\uff0c\u5f00\u5c55\u9a8c\u8bc1\u4e0e\u6027\u80fd\u6d4b\u8bd5\u5de5\u4f5c\u81f3\u5173\u91cd\u8981\uff0c\u8fd9\u80fd\u786e\u4fdd\u6846\u67b6\u5728\u65e5\u672cGPU\u670d\u52a1\u5668\u4e0a\u9ad8\u6548\u7a33\u5b9a\u8fd0\u884c\u3002<\/p>\n
          \n
          \u529f\u80fd\u9a8c\u8bc1\uff1a\n
          \n
          \u8fd0\u884c\u6846\u67b6\u5185\u7f6e\u7684\u6d4b\u8bd5\u7528\u4f8b\uff0c\u9a8c\u8bc1\u57fa\u7840\u529f\u80fd\u662f\u5426\u6b63\u5e38\u3002<\/li>\n
          \u901a\u8fc7\u6846\u67b6\u4e13\u5c5e\u547d\u4ee4\u68c0\u67e5GPU\u8bc6\u522b\u72b6\u6001\uff08\u4f8b\u5982TensorFlow\u7684 tf.config.list_physical_devices('GPU')<\/code> \u547d\u4ee4\uff0c\u4ee5\u53caPyTorch\u7684 torch.cuda.is_available()<\/code> \u547d\u4ee4\uff09\u3002<\/li>\n<\/ul>\n<\/li>\n
          \u6027\u80fd\u6d4b\u8bd5\uff1a\n
          \n
          \u9488\u5bf9\u540c\u4e00\u4efb\u52a1\uff0c\u5bf9\u6bd4CPU\u4e0eGPU\u7684\u8fd0\u7b97\u901f\u5ea6\uff0c\u91cf\u5316GPU\u52a0\u901f\u5e26\u6765\u7684\u6027\u80fd\u63d0\u5347\u6548\u679c\u3002<\/li>\n
          \u8fd0\u884c\u957f\u65f6\u95f4\u8bad\u7ec3\u4efb\u52a1\uff0c\u6d4b\u8bd5\u65e5\u672c\u670d\u52a1\u5668\u7b97\u529b\u7684\u7a33\u5b9a\u6027\uff0c\u540c\u65f6\u76d1\u63a7GPU\u6e29\u5ea6\u3001\u5185\u5b58\u5360\u7528\u7b49\u5173\u952e\u6307\u6807\u3002<\/li>\n<\/ul>\n<\/li>\n
          \u6545\u969c\u6392\u67e5\uff1a\n
          \n
          \u9a71\u52a8\u51b2\u7a81\u95ee\u9898\uff1a\u91cd\u65b0\u5b89\u88c5\u5339\u914d\u7248\u672c\u7684\u9a71\u52a8\uff0c\u5e76\u66f4\u65b0\u7cfb\u7edf\u4f9d\u8d56\u5e93\u3002<\/li>\n
          \u5185\u5b58\u4e0d\u8db3\u95ee\u9898\uff1a\u5347\u7ea7\u670d\u52a1\u5668\u5185\u5b58\u5bb9\u91cf\uff0c\u6216\u5bf9\u6a21\u578b\u8fdb\u884c\u4f18\u5316\u4ee5\u964d\u4f4e\u5185\u5b58\u5360\u7528\uff08\u4f8b\u5982\u91c7\u7528\u6df7\u5408\u7cbe\u5ea6\u8bad\u7ec3\u6280\u672f\uff09\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ol>\n
          [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””]<\/p>\n
          \u4e94\u3001\u65e5\u672cGPU\u670d\u52a1\u5668AI\u6846\u67b6\u8fd0\u884c\u4f18\u5316\u6280\u5de7<\/strong><\/h2>\n
          \u4e3a\u5145\u5206\u53d1\u6325\u65e5\u672cGPU\u670d\u52a1\u5668\u4e0aAI\u6846\u67b6\u7684\u6027\u80fd\uff0c\u53ef\u91c7\u53d6\u4ee5\u4e0b\u4f18\u5316\u7b56\u7565\uff1a<\/p>\n
          \n
          \u7b97\u529b\u4f18\u5316\uff1a\u5f00\u542fGPU\u5e76\u884c\u8ba1\u7b97\u529f\u80fd\uff0c\u6839\u636e\u5185\u5b58\u5bb9\u91cf\u8c03\u6574\u6279\u6b21\u5927\u5c0f\uff0c\u91c7\u7528\u6df7\u5408\u7cbe\u5ea6\u8bad\u7ec3\u6280\u672f\uff0c\u63d0\u5347\u8fd0\u7b97\u6548\u7387\u3002<\/li>\n
          \u7f51\u7edc\u4f18\u5316\uff1a\u7ed1\u5b9a\u65e5\u672c\u672c\u5730DNS\u670d\u52a1\u5668\uff0c\u7f29\u77ed\u57df\u540d\u89e3\u6790\u65f6\u95f4\uff1b\u542f\u7528TCP\u52a0\u901f\u6280\u672f\uff0c\u63d0\u5347\u6a21\u578b\u4e0b\u8f7d\u4e0e\u6570\u636e\u4f20\u8f93\u901f\u5ea6\u3002<\/li>\n
          \u8fd0\u7ef4\u4f18\u5316\uff1a\u5b9a\u671f\u66f4\u65b0GPU\u9a71\u52a8\u4e0e\u6846\u67b6\u7248\u672c\uff0c\u4fee\u590d\u5df2\u77e5\u6f0f\u6d1e\u5e76\u63d0\u5347\u6027\u80fd\uff1b\u501f\u52a9\u65e5\u672c\u6570\u636e\u4e2d\u5fc3\u7684\u6e29\u63a7\u4f18\u52bf\uff0c\u5b9e\u65f6\u76d1\u63a7GPU\u6e29\u5ea6\uff0c\u9632\u6b62\u8bbe\u5907\u8fc7\u70ed\u3002<\/li>\n
          \u6210\u672c\u4f18\u5316\uff1a\u6839\u636e\u9879\u76ee\u9700\u6c42\uff0c\u7075\u6d3b\u9009\u62e9\u6309\u9700\u8ba1\u8d39\u3001\u5305\u6708\u6216\u5305\u5e74\u7684\u8ba1\u8d39\u6a21\u5f0f\uff0c\u964d\u4f4eAI\u7814\u53d1\u6210\u672c\u3002\u5bf9\u4e8e\u957f\u671f\u9879\u76ee\uff0c\u5305\u6708\u6216\u5305\u5e74\u8ba1\u8d39\u6a21\u5f0f\u7684\u6027\u4ef7\u6bd4\u66f4\u9ad8\u3002<\/li>\n<\/ol>\n
          [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””]<\/p>\n
          \u516d\u3001\u5e38\u89c1\u95ee\u9898\u89e3\u7b54\uff08FAQ\uff09<\/strong><\/h2>\n
          \n
          \u95ee\uff1a\u65e5\u672cGPU\u670d\u52a1\u5668\u5b89\u88c5\u9a71\u52a8\u540e\uff0c\u63d0\u793a\u201c\u672a\u627e\u5230\u8bbe\u5907\u201d\u8be5\u5982\u4f55\u89e3\u51b3\uff1f
          \n\u7b54\uff1a\u68c0\u67e5\u9a71\u52a8\u7248\u672c\u4e0eGPU\u578b\u53f7\u662f\u5426\u5339\u914d\u3002\u5728BIOS\u4e2d\u5173\u95ed\u5b89\u5168\u542f\u52a8\u529f\u80fd\u540e\uff0c\u91cd\u65b0\u5b89\u88c5\u9a71\u52a8\u3002\u82e5\u95ee\u9898\u4ecd\u672a\u89e3\u51b3\uff0c\u53ef\u8054\u7cfb\u65e5\u672c\u670d\u52a1\u5668\u79df\u7528\u670d\u52a1\u5546\u7684\u6280\u672f\u652f\u6301\uff0c\u8fdb\u884c\u786c\u4ef6\u68c0\u6d4b\u3002<\/li>\n
          \u95ee\uff1aTensorFlow\u65e0\u6cd5\u8bc6\u522bGPU\uff0c\u4ec5\u4f7f\u7528CPU\u8fdb\u884c\u8fd0\u7b97\uff0c\u8be5\u5982\u4f55\u5904\u7406\uff1f
          \n\u7b54\uff1a\u9a8c\u8bc1TensorFlow\u7248\u672c\u4e0eCUDA\/cuDNN\u7248\u672c\u7684\u517c\u5bb9\u6027\uff0c\u68c0\u67e5\u73af\u5883\u53d8\u91cf\u662f\u5426\u914d\u7f6e\u6b63\u786e\u3002\u5fc5\u8981\u65f6\uff0c\u53ef\u5728\u5168\u65b0\u7684Conda\u73af\u5883\u4e2d\u91cd\u65b0\u5b89\u88c5TensorFlow\u3002<\/li>\n
          \u95ee\uff1a\u540c\u4e00\u53f0\u65e5\u672cGPU\u670d\u52a1\u5668\u4e0a\uff0c\u80fd\u5426\u540c\u65f6\u8fd0\u884c\u591a\u4e2aAI\u6846\u67b6\uff1f
          \n\u7b54\uff1a\u53ef\u4ee5\u3002\u901a\u8fc7Conda\u4e3a\u4e0d\u540c\u6846\u67b6\u521b\u5efa\u76f8\u4e92\u9694\u79bb\u7684\u8fd0\u884c\u73af\u5883\uff0c\u786e\u4fdd\u5404\u6846\u67b6\u7684\u4f9d\u8d56\u5e93\u4e0d\u4f1a\u53d1\u751f\u51b2\u7a81\u3002<\/li>\n
          \u95ee\uff1a\u65e5\u672c\u670d\u52a1\u5668\u5e26\u5bbd\u4e0d\u8db3\uff0c\u5f71\u54cdAI\u6a21\u578b\u8bad\u7ec3\u600e\u4e48\u529e\uff1f
          \n\u7b54\uff1a\u5347\u7ea7\u670d\u52a1\u5668\u5e26\u5bbd\u5957\u9910\uff0c\u9009\u62e9\u642d\u8f7dBGP\u591a\u7ebf\u5e26\u5bbd\u7684\u65e5\u672cGPU\u670d\u52a1\u5668\uff0c\u4fdd\u969c\u6570\u636e\u4f20\u8f93\u7a33\u5b9a\u6027\u3002\u540c\u65f6\uff0c\u53ef\u5bf9\u8bad\u7ec3\u6570\u636e\u8fdb\u884c\u538b\u7f29\u5904\u7406\uff0c\u6216\u91c7\u7528\u672c\u5730\u6570\u636e\u7f13\u5b58\u6280\u672f\uff0c\u51cf\u5c11\u5e26\u5bbd\u5360\u7528\u3002<\/li>\n<\/ol>\n
          [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””]<\/p>\n
          \u4e03\u3001\u603b\u7ed3<\/strong><\/h2>\n
          \u4e3a\u7279\u5b9aAI\u6846\u67b6\u914d\u7f6e\u65e5\u672cGPU\u670d\u52a1\u5668\uff0c\u4e3b\u8981\u5305\u542b\u56db\u5927\u6838\u5fc3\u6b65\u9aa4\uff1a\u670d\u52a1\u5668\u9009\u578b\u3001\u7cfb\u7edf\u73af\u5883\u642d\u5efa\u3001\u6846\u67b6\u5b89\u88c5\u914d\u7f6e\u4ee5\u53ca\u9a8c\u8bc1\u4f18\u5316\u3002\u65e5\u672cGPU\u670d\u52a1\u5668\u51ed\u501f\u4f4e\u5ef6\u8fdf\u3001\u9ad8\u5408\u89c4\u6027\u3001\u672c\u5730\u5316\u652f\u6301\u5b8c\u5584\u7b49\u4f18\u52bf\uff0c\u4e3a\u4e9a\u592a\u5730\u533a\u7684AI\u5e94\u7528\u63d0\u4f9b\u4e86\u53ef\u9760\u7684\u7b97\u529b\u652f\u6491\u3002\u9075\u5faa\u672c\u6587\u7684\u6280\u672f\u6307\u5357\uff0c\u4f60\u80fd\u591f\u9ad8\u6548\u5b8c\u6210\u914d\u7f6e\u6d41\u7a0b\uff0c\u89c4\u907f\u5e38\u89c1\u95ee\u9898\uff0c\u6700\u5927\u5316\u53d1\u6325AI\u6846\u67b6\u7684\u6027\u80fd\u3002\u65e0\u8bba\u4f60\u4ece\u4e8b\u5b66\u672f\u7814\u7a76\u8fd8\u662f\u5de5\u4e1a\u5e94\u7528\uff0c\u4e3aAI\u6846\u67b6\u91cf\u8eab\u5b9a\u5236\u7684\u65e5\u672cGPU\u670d\u52a1\u5668\u914d\u7f6e\u65b9\u6848<\/strong>\uff0c\u90fd\u5c06\u663e\u8457\u63d0\u5347\u4f60\u7684\u7814\u53d1\u6548\u7387\u3002\u66f4\u591aAI\u4e0e\u670d\u52a1\u5668\u914d\u7f6e\u76f8\u5173\u7684\u6280\u672f\u6307\u5357\uff0c\u656c\u8bf7\u5173\u6ce8\u672c\u7f51\u7ad9\u3002<\/p>\n
          [\/vc_column_text][\/vc_column][\/vc_row]<\/p>\n<\/div>","protected":false},"excerpt":{"rendered":"
          [vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””] \u968f\u7740\u5927\u8bed\u8a00\u6a21\u578b\u4e0e\u6df1\u5ea6\u5b66\u4e60\u5e94\u7528\u7684\u7206\u53d1\u5f0f\u589e\u957f\uff0cGPU\u670d\u52a1\u5668\u5df2\u6210\u4e3a\u7b97\u529b\u6838\u5fc3\u3002TensorFlow\u3001PyTorch\u7b49\u7279\u5b9aAI\u6846\u67b6\u5bf9\u670d\u52a1\u5668\u914d\u7f6e\u6709\u7740\u4e25\u82db\u8981\u6c42\u3002\u65e5\u672cGPU\u670d\u52a1\u5668\u51ed\u501f\u4f4e\u5ef6\u8fdf\u3001\u5e26\u5bbd\u7a33\u5b9a\u3001\u4e9a\u592a\u5730\u533a\u5408\u89c4\u6027\u5f3a\u7b49\u4f18\u52bf\uff0c\u6210\u4e3a\u8de8\u5883AI\u7814\u53d1\u56e2\u961f\u7684\u9996\u9009\u3002\u672c\u6587\u63d0\u4f9b\u4e00\u4efd\u5728\u65e5\u672cGPU\u670d\u52a1\u5668\u4e0a\u914d\u7f6eAI\u6846\u67b6\u7684\u5206\u6b65\u6280\u672f\u6307\u5357\uff0c\u5185\u5bb9\u8986\u76d6\u4ece\u786c\u4ef6\u9009\u578b\u3001\u73af\u5883\u642d\u5efa\u5230\u6d4b\u8bd5\u4f18\u5316\u7684\u5168\u6d41\u7a0b\u3002\u65e5\u672cGPU\u670d\u52a1\u5668\u914d\u7f6e\u4e0eAI\u6846\u67b6\u90e8\u7f72\u662f\u672c\u6587\u7684\u6838\u5fc3\u91cd\u70b9\uff0c\u5c06\u9488\u5bf9\u6027\u89e3\u51b3\u9a71\u52a8\u517c\u5bb9\u6027\u3001\u7b97\u529b\u6d6a\u8d39\u7b49\u5173\u952e\u75db\u70b9\u3002 [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””] \u4e00\u3001\u524d\u671f\u51c6\u5907\uff1a\u6311\u9009\u5408\u9002\u7684\u65e5\u672cGPU\u670d\u52a1\u5668 \u9009\u62e9\u4e00\u53f0\u9002\u914d\u7684\u65e5\u672cGPU\u670d\u52a1\u5668\uff0c\u662fAI\u6846\u67b6\u914d\u7f6e\u6210\u529f\u7684\u57fa\u7840\u6b65\u9aa4\u3002\u9700\u4eceAI\u6846\u67b6\u7684\u9700\u6c42\u89d2\u5ea6\u51fa\u53d1\uff0c\u8003\u91cf\u4ee5\u4e0b\u786c\u4ef6\u9009\u578b\u6807\u51c6\u4e0e\u5730\u57df\u4f18\u52bf\uff1a 1.1 \u786c\u4ef6\u9009\u578b\u6807\u51c6 GPU\u578b\u53f7\uff1a\u4e0d\u540cAI\u6846\u67b6\u7684\u9002\u914d\u6027\u5b58\u5728\u5dee\u5f02\u3002\u4f8b\u5982\uff0cPyTorch\u5bf9\u9ad8\u6027\u80fdGPU\u67b6\u6784\u7684\u4f18\u5316\u6548\u679c\u66f4\u4f73\u3002\u6311\u9009\u670d\u52a1\u5668\u65f6\uff0c\u5e94\u4f18\u5148\u9009\u62e9\u83b7\u5f97\u4e3b\u6d41\u6846\u67b6\u5e7f\u6cdb\u652f\u6301\u7684\u578b\u53f7\u3002 \u8f85\u52a9\u786c\u4ef6\uff1a\u591a\u6838CPU\u662f\u5b9e\u73b0\u5e76\u884c\u6570\u636e\u9884\u5904\u7406\u7684\u5fc5\u5907\u6761\u4ef6\uff1b\u5185\u5b58\u5bb9\u91cf\u5efa\u8bae\u81f3\u5c11\u8fbe\u523064GB\uff0c\u907f\u514d\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u6027\u80fd\u74f6\u9888\uff1b\u63a8\u8350\u914d\u5907\u9ad8\u901f\u5b58\u50a8\u8bbe\u5907\uff08NVMe SSD\uff09\uff0c\u4ee5\u52a0\u901f\u6a21\u578b\u52a0\u8f7d\u4e0e\u6570\u636e\u8bfb\u5199\u901f\u5ea6\u3002 \u5e26\u5bbd\u8981\u6c42\uff1aAI\u6a21\u578b\u7684\u8bad\u7ec3\u4e0e\u63a8\u7406\u6d89\u53ca\u6d77\u91cf\u6570\u636e\u4f20\u8f93\u3002\u642d\u8f7dBGP\u591a\u7ebf\u5e26\u5bbd\u7684\u65e5\u672cGPU\u670d\u52a1\u5668\uff0c\u53ef\u63d0\u4f9b\u7a33\u5b9a\u9ad8\u901f\u7684\u6570\u636e\u4f20\u8f93\u670d\u52a1\uff0c\u8fd9\u5bf9\u8de8\u5883AI\u9879\u76ee\u81f3\u5173\u91cd\u8981\u3002 1.2 \u65e5\u672cGPU\u670d\u52a1\u5668\u7684\u72ec\u7279\u4f18\u52bf \u5408\u89c4\u6027\u4f18\u52bf\uff1a\u65e5\u672c\u7684\u6570\u636e\u9690\u79c1\u653f\u7b56\uff0c\u975e\u5e38\u9002\u5408\u9762\u5411\u65e5\u672c\u3001\u97e9\u56fd\u4e43\u81f3\u6574\u4e2a\u4e9a\u592a\u5e02\u573a\u7684AI\u5e94\u7528\uff0c\u80fd\u591f\u4fdd\u969c\u6570\u636e\u5904\u7406\u73af\u8282\u7684\u5408\u6cd5\u5408\u89c4\u3002 \u672c\u5730\u5316\u6280\u672f\u652f\u6301\uff1a7\u00d724\u5c0f\u65f6\u8fd0\u7ef4\u670d\u52a1\u53ef\u6d88\u9664\u65f6\u533a\u6c9f\u901a\u58c1\u5792\uff0c\u9488\u5bf9\u6d77\u5916\u914d\u7f6e\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684\u95ee\u9898\uff0c\u63d0\u4f9b\u53ca\u65f6\u9ad8\u6548\u7684\u6280\u672f\u652f\u6301\u3002 [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””] \u4e8c\u3001\u7cfb\u7edf\u73af\u5883\u642d\u5efa\uff1a\u4e3aAI\u6846\u67b6\u94fa\u8def\u642d\u6865 \u7a33\u5b9a\u4e14\u517c\u5bb9\u7684\u7cfb\u7edf\u73af\u5883\uff0c\u662fAI\u6846\u67b6\u987a\u7545\u8fd0\u884c\u7684\u524d\u63d0\u3002\u672c\u8282\u5c06\u7ed3\u5408\u65e5\u672c\u670d\u52a1\u5668\u7684\u7279\u70b9\uff0c\u8be6\u7ec6\u4ecb\u7ecd\u64cd\u4f5c\u7cfb\u7edf\u9009\u578b\u4e0eGPU\u9a71\u52a8\u5b89\u88c5\u6d41\u7a0b\u3002 2.1 \u64cd\u4f5c\u7cfb\u7edf\u9009\u578b \u63a8\u8350\u7248\u672c\uff1aUbuntu\u662f\u9996\u9009\u7cfb\u7edf\uff0c\u5b83\u4e0e\u7edd\u5927\u591a\u6570AI\u6846\u67b6\u7684\u517c\u5bb9\u6027\u6700\u4f73\u3002\u591a\u6570\u65e5\u672c\u670d\u52a1\u5668\u79df\u7528\u670d\u52a1\u5546\u90fd\u63d0\u4f9b\u8fd9\u4e24\u4e2a\u7248\u672c\u7684\u9884\u88c5\u670d\u52a1\uff0c\u91c7\u8d2d\u65f6\u53ef\u786e\u8ba4\u9884\u88c5\u6d41\u7a0b\u3002 \u7cfb\u7edf\u4f18\u5316\uff1a\u5173\u95ed\u975e\u5fc5\u8981\u7684\u540e\u53f0\u670d\u52a1\uff0c\u91ca\u653e\u7cfb\u7edf\u8d44\u6e90\uff1b\u5408\u7406\u914d\u7f6e\u4ea4\u6362\u5206\u533a\uff0c\u907f\u514d\u6a21\u578b\u8bad\u7ec3\u65f6\u51fa\u73b0\u5185\u5b58\u4e0d\u8db3\u7684\u9519\u8bef\u3002 2.2 GPU\u9a71\u52a8\u4e0e\u4f9d\u8d56\u5e93\u5b89\u88c5 \u9a71\u52a8\u7248\u672c\u5339\u914d\uff1a\u6839\u636eGPU\u578b\u53f7\u9009\u62e9\u5bf9\u5e94\u7684NVIDIA\u9a71\u52a8\u7248\u672c\u3002\u9700\u907f\u514d\u9a71\u52a8\u7248\u672c\u8fc7\u9ad8\u6216\u8fc7\u4f4e\u5bfc\u81f4\u7684\u6846\u67b6\u8fd0\u884c\u9519\u8bef\uff0c\u53ef\u53c2\u8003NVIDIA\u5b98\u65b9\u6587\u6863\u4e2d\u7684\u9a71\u52a8-GPU\u517c\u5bb9\u6027\u5217\u8868\u3002 \u6838\u5fc3\u4f9d\u8d56\u5e93\u5b89\u88c5\uff1a \u5b89\u88c5CUDA\u5de5\u5177\u5305\uff1a\u9009\u62e9\u4e0e\u76ee\u6807AI\u6846\u67b6\u517c\u5bb9\u7684CUDA\u7248\u672c\u3002\u5efa\u8bae\u4f7f\u7528\u65e5\u672c\u672c\u5730\u955c\u50cf\u6e90\uff08\u5982\u4e1c\u4eac\u5de5\u4e1a\u5927\u5b66\u955c\u50cf\u6e90\uff09\uff0c\u63d0\u5347\u4e0b\u8f7d\u901f\u5ea6\u3002 \u5b89\u88c5cuDNN\uff1a\u4e0b\u8f7d\u4e0e\u5df2\u5b89\u88c5CUDA\u5de5\u5177\u5305\u7248\u672c\u5339\u914d\u7684cuDNN\uff0c\u5e76\u6b63\u786e\u914d\u7f6e\u73af\u5883\u53d8\u91cf\u3002 \u5b89\u88c5\u9a8c\u8bc1\uff1a\u5728\u7ec8\u7aef\u6267\u884c nvidia-smi \u547d\u4ee4\u3002\u82e5\u547d\u4ee4\u8f93\u51fa\u4e2d\u6b63\u786e\u663e\u793aGPU\u578b\u53f7\u3001\u9a71\u52a8\u7248\u672c\u53caCUDA\u7248\u672c\uff0c\u5219\u8bf4\u660e\u9a71\u52a8\u5b89\u88c5\u6210\u529f\u3002 [\/vc_column_text][\/vc_column][\/vc_row][vc_row el_class=”blog-detail-section”][vc_column][vc_column_text css=””] \u4e09\u3001\u65e5\u672cGPU\u670d\u52a1\u5668\u4e3b\u6d41AI\u6846\u67b6\u5206\u6b65\u914d\u7f6e\u6307\u5357 \u672c\u8282\u9488\u5bf9\u76ee\u524d\u5e94\u7528\u6700\u5e7f\u6cdb\u7684\u4e24\u5927AI\u6846\u67b6\u2014\u2014TensorFlow\u4e0ePyTorch\uff0c\u7ed3\u5408\u65e5\u672cGPU\u670d\u52a1\u5668\u7684\u7f51\u7edc\u73af\u5883\u4e0e\u786c\u4ef6\u7279\u6027\uff0c\u63d0\u4f9b\u8be6\u7ec6\u53ef\u843d\u5730\u7684\u914d\u7f6e\u65b9\u6848\u3002 3.1 TensorFlow GPU\u914d\u7f6e \u5b89\u88c5\u65b9\u5f0f\uff1a\u63a8\u8350\u4f7f\u7528Conda\u800c\u975epip\u8fdb\u884c\u5b89\u88c5\uff0c\u501f\u52a9\u73af\u5883\u9694\u79bb\u529f\u80fd\uff0c\u907f\u514d\u4e0d\u540c\u6846\u67b6\u4e0e\u4f9d\u8d56\u5e93\u4e4b\u95f4\u7684\u7248\u672c\u51b2\u7a81\u3002\u53ef\u901a\u8fc7\u65e5\u672c\u672c\u5730\u955c\u50cf\u6e90\u5b89\u88c5Miniconda\u3002 \u6838\u5fc3\u6b65\u9aa4\uff1a \u521b\u5efa\u4e13\u5c5eConda\u73af\u5883\uff1aconda create -n tf-gpu python=3.9 \u6fc0\u6d3b\u73af\u5883\uff1aconda activate tf-gpu […]<\/p>\n
          Read More…<\/a><\/p>\n","protected":false},"author":1,"featured_media":29576,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[8117],"tags":[9832,9833,9489,8163],"class_list":["post-29582","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-japan-dedicated-server-sc","tag-ai-framework-deployment","tag-apac-ai-server","tag-gpu-server-configuration","tag-japan-server-hosting-sc"],"acf":[],"yoast_head":"\n