{"id":204358,"date":"2025-05-29T09:11:51","date_gmt":"2025-05-29T01:11:51","guid":{"rendered":"https:\/\/server.hk\/cnblog\/204358\/"},"modified":"2025-05-29T09:11:51","modified_gmt":"2025-05-29T01:11:51","slug":"keras%ef%bc%9a%e9%80%9a%e8%bf%87%e8%af%a6%e7%bb%86%e7%a4%ba%e4%be%8b%e4%ba%86%e8%a7%a3%e5%9f%ba%e7%a1%80%e7%9f%a5%e8%af%86","status":"publish","type":"post","link":"https:\/\/server.hk\/cnblog\/204358\/","title":{"rendered":"Keras\uff1a\u901a\u8fc7\u8be6\u7ec6\u793a\u4f8b\u4e86\u89e3\u57fa\u7840\u77e5\u8bc6"},"content":{"rendered":"<p><b><\/b>     <\/p>\n<h1>Keras\uff1a\u901a\u8fc7\u8be6\u7ec6\u793a\u4f8b\u4e86\u89e3\u57fa\u7840\u77e5\u8bc6<\/h1>\n<p>\u5728\u6587\u7ae0\u5b9e\u6218\u5f00\u53d1\u7684\u8fc7\u7a0b\u4e2d\uff0c\u6211\u4eec\u7ecf\u5e38\u4f1a\u9047\u5230\u4e00\u4e9b\u8fd9\u6837\u90a3\u6837\u7684\u95ee\u9898\uff0c\u7136\u540e\u8981\u5361\u597d\u534a\u5929\uff0c\u7b49\u95ee\u9898\u89e3\u51b3\u4e86\u624d\u53d1\u73b0\u539f\u6765\u4e00\u4e9b\u7ec6\u8282\u77e5\u8bc6\u70b9\u8fd8\u662f\u6ca1\u6709\u638c\u63e1\u597d\u3002\u4eca\u5929\u5c31\u6574\u7406\u5206\u4eab\u300aKeras\uff1a\u901a\u8fc7\u8be6\u7ec6\u793a\u4f8b\u4e86\u89e3\u57fa\u7840\u77e5\u8bc6\u300b\uff0c\u804a\u804a\uff0c\u5e0c\u671b\u53ef\u4ee5\u5e2e\u52a9\u5230\u6b63\u5728\u52aa\u529b\u8d5a\u94b1\u7684\u4f60\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.17golang.com\/uploads\/20241105\/17307936256729d099ce91b.jpg\" class=\"aligncenter\"><\/p>\n<p>\u5f00\u53d1\u8005\u4eec\u5927\u5bb6\u597d\uff0c<\/p>\n<p>\u5982\u679c\u60a8\u662f\u6df1\u5ea6\u5b66\u4e60\u65b0\u624b\uff0c\u60a8\u53ef\u80fd\u9047\u5230\u8fc7 <strong>keras<\/strong> \u8fd9\u4e2a\u540d\u5b57\u3002\u4f46\u5b83\u5230\u5e95\u662f\u4ec0\u4e48\uff1f\u5b83\u662f\u5982\u4f55\u5de5\u4f5c\u7684\uff1f\u5728\u8fd9\u7bc7\u6587\u7ae0\u4e2d\uff0c\u6211\u5c06\u4ece\u5934\u5f00\u59cb\u89e3\u91ca\u6240\u6709\u5185\u5bb9\uff0c\u5e76\u5411\u60a8\u5c55\u793a\u4e00\u4e2a\u4f7f\u7528 <strong>keras<\/strong> \u6784\u5efa\u7b80\u5355\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u7684\u5206\u6b65\u793a\u4f8b\u3002\u6211\u8fd8\u5c06\u89e3\u91ca\u8bf8\u5982 <strong>mnist \u6570\u636e\u96c6<\/strong> \u4e4b\u7c7b\u7684\u5173\u952e\u6982\u5ff5\uff0c\u4ee5\u4fbf\u60a8\u53ef\u4ee5\u8f7b\u677e\u7406\u89e3\uff01<\/p>\n<p>keras \u662f\u4e00\u4e2a\u7528 python \u7f16\u5199\u7684\u5f00\u6e90<strong>\u9ad8\u7ea7\u795e\u7ecf\u7f51\u7edc api<\/strong>\u3002\u5b83\u5141\u8bb8\u5f00\u53d1\u4eba\u5458\u4f7f\u7528\u7528\u6237\u53cb\u597d\u7684\u754c\u9762\u5feb\u901f\u8f7b\u677e\u5730\u6784\u5efa\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u3002 keras \u4f4d\u4e8e <strong>tensorflow<\/strong> \u7b49\u66f4\u590d\u6742\u7684\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u4e4b\u4e0a\uff0c\u8ba9\u60a8\u53ef\u4ee5\u4e13\u6ce8\u4e8e\u6784\u5efa\u6a21\u578b\uff0c\u800c\u4e0d\u4f1a\u88ab\u5e95\u5c42\u590d\u6742\u6027\u6240\u56f0\u6270\u3002<\/p>\n<ul>\n<li> <strong>\u6613\u4e8e\u4f7f\u7528<\/strong>\uff1akeras \u7684\u8bbe\u8ba1\u6613\u4e8e\u9605\u8bfb\u548c\u7406\u89e3\uff0c\u8fd9\u975e\u5e38\u9002\u5408\u521d\u5b66\u8005\u3002<\/li>\n<li> <strong>\u6a21\u5757\u5316<\/strong>\uff1a\u5b83\u662f\u9ad8\u5ea6\u6a21\u5757\u5316\u7684\uff0c\u8fd9\u610f\u5473\u7740\u60a8\u53ef\u4ee5\u50cf\u79ef\u6728\u4e00\u6837\u5c06\u6a21\u578b\u7ec4\u5408\u5728\u4e00\u8d77\u3002<\/li>\n<li> <strong>\u591a\u540e\u7aef\u652f\u6301<\/strong>\uff1akeras \u53ef\u4ee5\u5728 tensorflow\u3001theano \u6216 cntk \u4e4b\u4e0a\u8fd0\u884c\uff0c\u4f7f\u5176\u975e\u5e38\u7075\u6d3b\u3002<\/li>\n<li> <strong>\u5feb\u901f\u539f\u578b<\/strong>\uff1a\u53ea\u9700\u51e0\u884c\u4ee3\u7801\u5373\u53ef\u6784\u5efa\u3001\u7f16\u8bd1\u548c\u8bad\u7ec3\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u3002<\/li>\n<\/ul>\n<p><strong>mnist \u6570\u636e\u96c6<\/strong> \u662f\u673a\u5668\u5b66\u4e60\u9886\u57df\u6700\u8457\u540d\u7684\u6570\u636e\u96c6\u4e4b\u4e00\u3002\u5b83\u5305\u542b <strong>70,000 \u5f20\u624b\u5199\u6570\u5b57 (0-9) \u56fe\u50cf<\/strong>\u3002\u6bcf\u4e2a\u56fe\u50cf\u90fd\u662f\u7070\u5ea6\u56fe\u7247\uff0c\u5927\u5c0f\u4e3a 28&#215;28 \u50cf\u7d20\u3002\u76ee\u6807\u662f\u5c06\u8fd9\u4e9b\u56fe\u50cf\u5206\u7c7b\u4e3a\u5341\u4e2a\u6570\u5b57\u7c7b\u522b\u4e4b\u4e00\u3002<\/p>\n<p>\u4ee5\u4e0b\u662f mnist \u6570\u636e\u96c6\u4e2d\u7684\u4e00\u4e9b\u6570\u5b57\u793a\u4f8b\uff1a<\/p>\n<pre>[0] [1] [2] [3] [4] [5] [6] [7] [8] [9]\n<\/pre>\n<p>\u4f7f\u7528 keras \u65f6\uff0c\u60a8\u7ecf\u5e38\u4f1a\u770b\u5230\u6559\u7a0b\u4e2d\u4f7f\u7528 mnist \u6570\u636e\u96c6\uff0c\u56e0\u4e3a\u5b83\u7b80\u5355\u3001\u6613\u4e8e\u7406\u89e3\uff0c\u5e76\u4e14\u975e\u5e38\u9002\u5408\u6d4b\u8bd5\u65b0\u6a21\u578b\u3002<\/p>\n<hr>\n<p>\u73b0\u5728\u8ba9\u6211\u4eec\u4f7f\u7528 keras \u6784\u5efa\u4e00\u4e2a\u7b80\u5355\u7684\u795e\u7ecf\u7f51\u7edc\u6765\u5bf9\u8fd9\u4e9b\u624b\u5199\u6570\u5b57\u8fdb\u884c\u5206\u7c7b\u3002\u6211\u4eec\u5c06\u4e00\u6b65\u6b65\u8fdb\u884c\u3002<\/p>\n<h4> \u7b2c 1 \u6b65\uff1a\u5b89\u88c5 tensorflow\uff08keras \u4e0e tensorflow \u6346\u7ed1\u5728\u4e00\u8d77\uff09 <\/h4>\n<p>\u9996\u5148\uff0c\u60a8\u9700\u8981\u5b89\u88c5 <strong>tensorflow<\/strong>\uff0c\u56e0\u4e3a keras \u662f\u6700\u65b0\u7248\u672c\u4e2d tensorflow \u7684\u4e00\u90e8\u5206\u3002\u60a8\u53ef\u4ee5\u901a\u8fc7 pip \u5b89\u88c5\u5b83\uff1a<\/p>\n<pre>pip install tensorflow\n<\/pre>\n<h4> \u7b2c2\u6b65\uff1a\u5bfc\u5165\u6240\u9700\u7684\u5e93 <\/h4>\n<p>\u6211\u4eec\u5c06\u5bfc\u5165\u6784\u5efa\u548c\u8bad\u7ec3\u6a21\u578b\u6240\u9700\u7684 tensorflow \u548c keras \u7279\u5b9a\u5e93\u3002<\/p>\n<pre>import tensorflow as tf\nfrom tensorflow.keras import layers, models\n<\/pre>\n<p>\u8fd9\u91cc\uff0ctensorflow.keras \u662f tensorflow \u4e2d\u7684 keras api\u3002<\/p>\n<h4> \u6b65\u9aa4 3\uff1a\u52a0\u8f7d mnist \u6570\u636e\u96c6 <\/h4>\n<p>keras \u63d0\u4f9b\u4e86\u5bf9 mnist \u7b49\u6570\u636e\u96c6\u7684\u8f7b\u677e\u8bbf\u95ee\u3002\u6211\u4eec\u5c06\u52a0\u8f7d\u6570\u636e\u96c6\u5e76\u5c06\u5176\u5206\u4e3a\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\u3002<\/p>\n<pre># load the mnist dataset\nmnist = tf.keras.datasets.mnist\n(train_images, train_labels), (test_images, test_labels) = mnist.load_data()\n<\/pre>\n<p>\u5728\u6b64\u6b65\u9aa4\u4e2d\uff0ctrain_images \u548c train_labels \u4fdd\u5b58\u8bad\u7ec3\u6570\u636e\uff0c\u800c test_images \u548c test_labels \u4fdd\u5b58\u6d4b\u8bd5\u6570\u636e\u3002<\/p>\n<p>train_images\u4e2d\u7684\u6bcf\u5f20\u56fe\u50cf\u90fd\u662f28&#215;28\u50cf\u7d20\u7684\u7070\u5ea6\u56fe\u50cf\uff0ctrain_labels\u5305\u542b\u6bcf\u5f20\u56fe\u50cf\u5bf9\u5e94\u7684\u6570\u5b57\u6807\u7b7e\uff080-9\uff09\u3002<\/p>\n<h4> \u6b65\u9aa4 4\uff1a\u9884\u5904\u7406\u6570\u636e <\/h4>\n<p>\u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u9700\u8981\u5bf9\u56fe\u50cf\u7684\u50cf\u7d20\u503c\u8fdb\u884c\u5f52\u4e00\u5316\uff0c\u4ee5\u4f7f\u6a21\u578b\u8bad\u7ec3\u66f4\u52a0\u9ad8\u6548\u3002\u56fe\u50cf\u4e2d\u7684\u6bcf\u4e2a\u50cf\u7d20\u503c\u90fd\u5728 0 \u5230 255 \u4e4b\u95f4\u3002\u6211\u4eec\u5c06\u56fe\u50cf\u9664\u4ee5 255\uff0c\u5c06\u8fd9\u4e9b\u503c\u7f29\u653e\u5230 0 \u5230 1 \u4e4b\u95f4\u3002<\/p>\n<pre># normalize pixel values to be between 0 and 1\ntrain_images = train_images \/ 255.0\ntest_images = test_images \/ 255.0\n<\/pre>\n<h4> \u7b2c 5 \u6b65\uff1a\u6784\u5efa\u6a21\u578b <\/h4>\n<p>\u73b0\u5728\u8ba9\u6211\u4eec\u4f7f\u7528 keras \u6784\u5efa\u6211\u4eec\u7684\u795e\u7ecf\u7f51\u7edc\u3002\u6211\u4eec\u5c06\u521b\u5efa\u4e00\u4e2a<strong>\u987a\u5e8f<\/strong>\u6a21\u578b\uff0c\u5b83\u5141\u8bb8\u6211\u4eec\u5c06\u5c42\u5806\u53e0\u5728\u53e6\u4e00\u4e2a\u4e4b\u4e0a\u3002<\/p>\n<pre># build the model\nmodel = models.sequential([\n    layers.flatten(input_shape=(28, 28)),      # flatten the 28x28 images into a 1d vector of 784 pixels\n    layers.dense(128, activation='relu'),      # add a fully-connected (dense) layer with 128 neurons\n    layers.dense(10, activation='softmax')     # output layer with 10 neurons (one for each digit 0-9)\n])\n<\/pre>\n<ul>\n<li> <strong>flatten<\/strong>\uff1aflatten \u5c42\u5c06 28&#215;28 2d \u56fe\u50cf\u8f6c\u6362\u4e3a 784 \u4e2a\u503c\u7684 1d \u6570\u7ec4\u3002<\/li>\n<li> <strong>dense<\/strong>\uff1adense \u5c42\u662f\u5168\u8fde\u63a5\u5c42\u3002\u8fd9\u91cc\u6211\u4eec\u7684\u9690\u85cf\u5c42\u6709 128 \u4e2a\u795e\u7ecf\u5143\uff0c\u8f93\u51fa\u5c42\u6709 10 \u4e2a\u795e\u7ecf\u5143\uff08\u56e0\u4e3a\u6211\u4eec\u6709 10 \u4e2a\u6570\u5b57\u7c7b\uff09\u3002\u6211\u4eec\u4f7f\u7528 <strong>relu<\/strong> \u4f5c\u4e3a\u9690\u85cf\u5c42\u7684\u6fc0\u6d3b\u51fd\u6570\uff0c\u4f7f\u7528 <strong>softmax<\/strong> \u4f5c\u4e3a\u8f93\u51fa\u5c42\u3002<\/li>\n<\/ul>\n<h4> \u7b2c 6 \u6b65\uff1a\u7f16\u8bd1\u6a21\u578b <\/h4>\n<p>\u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u9700\u8981\u7f16\u8bd1\u6a21\u578b\u3002\u8fd9\u662f\u6211\u4eec\u6307\u5b9a<strong>\u4f18\u5316\u5668<\/strong>\u3001<strong>\u635f\u5931\u51fd\u6570<\/strong>\u548c<strong>\u8bc4\u4f30\u6307\u6807<\/strong>\u3002<br \/>\u7684\u5730\u65b9 <\/p>\n<pre># compile the model\nmodel.compile(optimizer='adam',\n              loss='sparse_categorical_crossentropy',\n              metrics=['accuracy'])\n<\/pre>\n<ul>\n<li> <strong>adam<\/strong> \u4f18\u5316\u5668\uff1a\u8fd9\u662f\u4e00\u79cd\u7528\u4e8e\u8bad\u7ec3\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u7684\u6d41\u884c\u4f18\u5316\u5668\u3002<\/li>\n<li> <strong>\u7a00\u758f\u5206\u7c7b\u4ea4\u53c9\u71b5<\/strong>\uff1a\u6b64\u635f\u5931\u51fd\u6570\u7528\u4e8e\u50cf\u6211\u4eec\u8fd9\u6837\u7684\u591a\u7c7b\u5206\u7c7b\u95ee\u9898\u3002<\/li>\n<li> <strong>\u51c6\u786e\u6027<\/strong>\uff1a\u6211\u4eec\u5c06\u4f7f\u7528\u51c6\u786e\u6027\u4f5c\u4e3a\u8bc4\u4f30\u6a21\u578b\u6027\u80fd\u7684\u6307\u6807\u3002<\/li>\n<\/ul>\n<h4> \u7b2c 7 \u6b65\uff1a\u8bad\u7ec3\u6a21\u578b <\/h4>\n<p>\u73b0\u5728\uff0c\u6211\u4eec\u51c6\u5907\u597d\u8bad\u7ec3\u6a21\u578b\u4e86\uff01\u6211\u4eec\u5c06\u5bf9\u5176\u8fdb\u884c <strong>5 epochs<\/strong> \u8bad\u7ec3\uff08\u5373\u6a21\u578b\u5c06\u904d\u5386\u6574\u4e2a\u8bad\u7ec3\u6570\u636e\u96c6 5 \u6b21\uff09\u3002<\/p>\n<pre># train the model\nmodel.fit(train_images, train_labels, epochs=5)\n<\/pre>\n<h4> \u7b2c 8 \u6b65\uff1a\u8bc4\u4f30\u6a21\u578b <\/h4>\n<p>\u6a21\u578b\u8bad\u7ec3\u5b8c\u6210\u540e\uff0c\u6211\u4eec\u53ef\u4ee5\u8bc4\u4f30\u5176\u5728\u6d4b\u8bd5\u6570\u636e\u4e0a\u7684\u6027\u80fd\u3002<\/p>\n<pre># Evaluate the model\ntest_loss, test_acc = model.evaluate(test_images, test_labels)\n\nprint(f'Test accuracy: {test_acc}')\n<\/pre>\n<p>\u8fd9\u5c06\u4e3a\u6211\u4eec\u63d0\u4f9b\u6a21\u578b\u5728\u6d4b\u8bd5\u6570\u636e\u96c6\u4e0a\u7684\u51c6\u786e\u6027\u3002<\/p>\n<hr>\n<p>\u7b80\u5355\u6765\u8bf4\uff1a<\/p>\n<ol>\n<li> <strong>\u6570\u636e\u9884\u5904\u7406<\/strong>\uff1a\u6211\u4eec\u5bf9\u6570\u636e\u8fdb\u884c\u5f52\u4e00\u5316\uff0c\u4f7f\u8bad\u7ec3\u66f4\u52a0\u9ad8\u6548\u3002<\/li>\n<li> <strong>\u6a21\u578b\u5b9a\u4e49<\/strong>\uff1a\u6211\u4eec\u4f7f\u7528<strong>\u987a\u5e8f<\/strong> api \u6784\u5efa\u4e86\u4e00\u4e2a\u7b80\u5355\u7684\u524d\u9988\u795e\u7ecf\u7f51\u7edc\u3002<\/li>\n<li> <strong>\u7f16\u8bd1<\/strong>\uff1a\u6211\u4eec\u9009\u62e9\u4e86\u6b63\u786e\u7684\u635f\u5931\u51fd\u6570\u548c\u4f18\u5316\u5668\u6765\u6307\u5bfc\u6a21\u578b\u7684\u5b66\u4e60\u3002<\/li>\n<li> <strong>\u8bad\u7ec3<\/strong>\uff1a\u6a21\u578b\u5b66\u4f1a\u4e86\u901a\u8fc7\u591a\u6b21\u904d\u5386\u6570\u636e\u96c6\u5c06\u56fe\u50cf\u6620\u5c04\u5230\u6570\u5b57\u3002<\/li>\n<li> <strong>\u8bc4\u4f30<\/strong>\uff1a\u6700\u540e\uff0c\u6211\u4eec\u68c0\u67e5\u4e86\u6a21\u578b\u5bf9\u672a\u89c1\u8fc7\u7684\u6570\u636e\u7684\u6cdb\u5316\u7a0b\u5ea6\u3002<\/li>\n<\/ol>\n<hr>\n<p>keras \u7b80\u5316\u4e86\u6784\u5efa\u548c\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u7684\u8fc7\u7a0b\uff0c\u4f7f\u5176\u6210\u4e3a\u521d\u5b66\u8005\u7684\u7406\u60f3\u8d77\u70b9\u3002\u4e00\u65e6\u60a8\u719f\u6089\u4e86\u57fa\u672c\u6a21\u578b\uff0c\u60a8\u5c31\u53ef\u4ee5\u5c1d\u8bd5\u66f4\u590d\u6742\u7684\u67b6\u6784\uff0c\u4f8b\u5982<strong>\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\uff08cnn\uff09<\/strong>\u548c<strong>\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\uff08rnn\uff09<\/strong>\u3002<\/p>\n<p>\u968f\u610f\u4f7f\u7528 keras \u66f4\u6df1\u5165\u5730\u63a2\u7d22\u6df1\u5ea6\u5b66\u4e60\u4e16\u754c\uff0c\u5c1d\u8bd5\u4e0d\u540c\u7684\u6a21\u578b\uff0c\u5e76\u7a81\u7834\u53ef\u80fd\u7684\u754c\u9650\uff01<\/p>\n<hr>\n<p>\u5230\u76ee\u524d\u4e3a\u6b62\uff0c\u60a8\u5bf9 keras \u6709\u4f55\u770b\u6cd5\uff1f <\/p>\n<p>\u7ec8\u4e8e\u4ecb\u7ecd\u5b8c\u5566\uff01\u5c0f\u4f19\u4f34\u4eec\uff0c\u8fd9\u7bc7\u5173\u4e8e\u300aKeras\uff1a\u901a\u8fc7\u8be6\u7ec6\u793a\u4f8b\u4e86\u89e3\u57fa\u7840\u77e5\u8bc6\u300b\u7684\u4ecb\u7ecd\u5e94\u8be5\u8ba9\u4f60\u6536\u83b7\u591a\u591a\u4e86\u5427\uff01\u6b22\u8fce\u5927\u5bb6\u6536\u85cf\u6216\u5206\u4eab\u7ed9\u66f4\u591a\u9700\u8981\u5b66\u4e60\u7684\u670b\u53cb\u5427~\u516c\u4f17\u53f7\u4e5f\u4f1a\u53d1\u5e03\u6587\u7ae0\u76f8\u5173\u77e5\u8bc6\uff0c\u5feb\u6765\u5173\u6ce8\u5427\uff01<\/p>\n<p>      \u7248\u672c\u58f0\u660e \u672c\u6587\u8f6c\u8f7d\u4e8e\uff1adev.to \u5982\u6709\u4fb5\u72af\uff0c\u8bf7\u8054\u7cfb\u5220\u9664<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Keras\uff1a\u901a\u8fc7\u8be6\u7ec6\u793a\u4f8b\u4e86\u89e3\u57fa\u7840&#46;&#46;&#46;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4925],"tags":[],"class_list":["post-204358","post","type-post","status-publish","format-standard","hentry","category-4925"],"_links":{"self":[{"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/posts\/204358","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/comments?post=204358"}],"version-history":[{"count":0,"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/posts\/204358\/revisions"}],"wp:attachment":[{"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/media?parent=204358"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/categories?post=204358"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/server.hk\/cnblog\/wp-json\/wp\/v2\/tags?post=204358"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}