{"id":1307,"date":"2023-04-19T13:44:51","date_gmt":"2023-04-19T05:44:51","guid":{"rendered":"http:\/\/www.wayln.com\/?p=1307"},"modified":"2023-04-19T17:28:37","modified_gmt":"2023-04-19T09:28:37","slug":"keras%e6%a8%a1%e5%9e%8b%e4%b8%8e%e5%b1%82","status":"publish","type":"post","link":"https:\/\/www.wayln.com\/?p=1307","title":{"rendered":"Keras\u6a21\u578b\u4e0e\u5c42"},"content":{"rendered":"<div id=\"toc_container\" class=\"toc_transparent no_bullets\"><p class=\"toc_title\">Contents<\/p><ul class=\"toc_list\"><li><a href=\"#Keras\"><span class=\"toc_number toc_depth_1\">1<\/span> Keras\u5e38\u7528\u7684\u6a21\u578b\u4e0e\u5c42<\/a><ul><li><a href=\"#i\"><span class=\"toc_number toc_depth_2\">1.1<\/span> \u57fa\u7840\u5c42<\/a><\/li><li><a href=\"#i-2\"><span class=\"toc_number toc_depth_2\">1.2<\/span> \u5377\u79ef\u5c42<\/a><\/li><li><a href=\"#i-3\"><span class=\"toc_number toc_depth_2\">1.3<\/span> \u5faa\u73af\u7f51\u7edc\u76f8\u5173\u5c42<\/a><\/li><\/ul><\/li><li><a href=\"#i-4\"><span class=\"toc_number toc_depth_1\">2<\/span> \u81ea\u5b9a\u4e49\u5c42<\/a><\/li><li><a href=\"#i-5\"><span class=\"toc_number toc_depth_1\">3<\/span> \u81ea\u5b9a\u4e49\u6a21\u578b<\/a><\/li><li><a href=\"#API\"><span class=\"toc_number toc_depth_1\">4<\/span> \u6a21\u578b\u5e38\u7528API<\/a><\/li><li><a href=\"#KerasAPI\"><span class=\"toc_number toc_depth_1\">5<\/span> Keras\u5e38\u7528API\u8bf4\u660e<\/a><ul><li><a href=\"#Callbacks\"><span class=\"toc_number toc_depth_2\">5.1<\/span> \u56de\u8c03\u51fd\u6570Callbacks<\/a><\/li><li><a href=\"#i-6\"><span class=\"toc_number toc_depth_2\">5.2<\/span> \u57fa\u7840\u7528\u6cd5<\/a><\/li><li><a href=\"#i-7\"><span class=\"toc_number toc_depth_2\">5.3<\/span> \u5df2\u5b9a\u4e49\u7684\u56de\u8c03\u51fd\u6570\u7c7b<\/a><\/li><\/ul><\/li><\/ul><\/div>\n<p>[TOC]<\/p>\n<h1><span id=\"Keras\">Keras\u5e38\u7528\u7684\u6a21\u578b\u4e0e\u5c42<\/span><\/h1>\n<p>\u6a21\u578b\u4e0e\u5c42\u662fKeras\u4e2d\u7684\u4e24\u4e2a\u6700\u57fa\u672c\u7684\u6982\u5ff5\u3002<br \/>\nKeras\u5728Tensorflow.Keras.Layers\u4e0b\u5185\u7f6e\u4e86\u4e30\u5bcc\u7684\u6df1\u5ea6\u5b66\u4e60\u5e38\u7528\u7684\u5404\u79cd\u529f\u80fd\u7684\u9884\u5b9a\u4e49\u5c42\uff0c\u4e3e\u4f8b\u5982\u4e0b\uff1a<br \/>\n&#8211; Layers.Dense<br \/>\n&#8211; Layers.Flatten<br \/>\n&#8211; Layers.RNN<br \/>\n&#8211; Layers.BatchNormalization<br \/>\n&#8211; Layers.Dropout<br \/>\n&#8211; Layers.Conv2D<br \/>\n&#8211; Layers.MaxPooling2D<br \/>\n&#8211; Layers.Conv1D<br \/>\n&#8211; Layers.Embedding<br \/>\n&#8211; Layers.GRU<br \/>\n&#8211; Layers.LSTM<br \/>\n&#8211; Layers.Bidirectional<br \/>\n&#8211; Layers.InputLayer<br \/>\n\u5982\u679c\u4e0a\u8ff0\u7684\u5185\u7f6e\u7f51\u7edc\u5c42\u65e0\u6cd5\u6ee1\u8db3\u9700\u6c42\uff0c\u5219Keras\u5141\u8bb8\u6211\u4eec\u81ea\u5b9a\u4e49\u5c42\uff0c\u901a\u8fc7\u7ee7\u627fTensorflow.Keras.Engine.Layer\u57fa\u7c7b\u521b\u5efa\u81ea\u5b9a\u4e49\u7f51\u7edc\u5c42<\/p>\n<h2><span id=\"i\">\u57fa\u7840\u5c42<\/span><\/h2>\n<ul>\n<li>Dense:<br \/>\n\u5168\u8fde\u63a5\u5c42\uff0c\u903b\u8f91\u4e0a\u7b49\u4ef7\u4e8e\u8fd9\u6837\u4e00\u4e2a\u51fd\u6570\uff0c\u6743\u91cdW\u4e3amxn\u7684\u77e9\u9635\uff0c\u8f93\u5165\u4e3ax\u4e3anw\u7ef4\u5411\u91cf\uff0c\u6fc0\u6d3b\u51fd\u6570\u4e3aActivation,\u504f\u7f6e\u9879\u4e3aBias,\u8f93\u51fa\u5411\u91cfOut\u4e3am\u7ef4\u5411\u91cf\u3002\u51fd\u6570\u5982\u4e0b\uff0cOut=Activation(Wx+Bias),\u5373\u4e00\u4e2a\u7ebf\u6027\u53d8\u5316\u52a0\u4e00\u4e2a\u975e\u7ebf\u6027\u53d8\u5316\u4ea7\u751f\u8f93\u51fa<\/li>\n<li>Activation<br \/>\n\u6fc0\u6d3b\u51fd\u6570\u66fe\uff0c\u4e00\u822c\u653e\u5728\u5168\u8fde\u63a5\u5c42\u540e\u9762\uff0c\u7b49\u4ef7\u4e8e\u5728\u5168\u8fde\u63a5\u5c42\u4e2d\u6307\u5b9a\u6fc0\u6d3b\u51fd\u6570\uff0c\u6fc0\u6d3b\u51fd\u6570\u53ef\u4ee5\u63d0\u9ad8\u6a21\u578b\u7684\u975e\u7ebf\u6027\u8868\u8fbe\u80fd\u529b\u3002<\/li>\n<li>Dropout<br \/>\n\u968f\u673a\u4e22\u5f03\u5c42\uff0c\u8bad\u7ec3\u671f\u95f4\u4ee5\u4e00\u5b9a\u6982\u7387\u5c06\u8f93\u5165\u7f6e\u96f6\uff08\u7b49\u4ef7\u4e8e\u6309\u7167\u4e00\u5b9a\u7684\u6982\u7387\u5c06CNN\u5355\u5143\u6682\u65f6\u4ece\u7f51\u7edc\u4e2d\u4e22\u5f03\uff09\uff0c\u662f\u4e00\u79cd\u6b63\u5219\u5316\u624b\u6bb5\uff0c\u53ef\u4ee5\u4f5c\u4e3aCNN\u4e2d\u9632\u6b62\u8fc7\u62df\u5408\u3001\u63d0\u9ad8\u8bad\u7ec3\u6548\u679c\u7684\u795e\u5668<\/li>\n<li>BatchNormalization<br \/>\n\u6279\u6807\u51c6\u5316\u5c42\uff0c\u901a\u8fc7\u7ebf\u6027\u53d8\u6362\u5c06\u8f93\u5165\u6279\u6b21\u7f29\u653e\u5e73\u79fb\u6210\u7a33\u5b9a\u7684\u5747\u503c\u548c\u6807\u51c6\u5dee\uff0c\u53ef\u4ee5\u589e\u5f3a\u6a21\u578b\u5bf9\u8f93\u5165\u6279\u6b21\u7684\u4e0d\u540c\u5206\u5e03\u7684\u9002\u5e94\u6027\uff0c\u52a0\u5feb\u8bad\u7ec3\u901f\u5ea6\uff0c\u6709\u8f7b\u5fae\u6b63\u5219\u5316\u6548\u679c\u3002\u4e00\u822c\u5728\u6fc0\u6d3b\u51fd\u6570\u4e4b\u524d\u4f7f\u7528<\/li>\n<li>SpatialDropout2D<br \/>\n\u7a7a\u95f4\u968f\u673a\u7f6e\u96f6\u5c42\uff0c\u8bad\u7ec3\u671f\u95f4\u4ee5\u4e00\u5b9a\u6982\u7387\u5c06\u6574\u4e2a\u7279\u5f81\u56fe\u7f6e\u96f6\uff0c\u662f\u4e00\u79cd\u6b63\u5219\u5316\u624b\u6bb5\uff0c\u6709\u5229\u4e8e\u907f\u514d\u7279\u5f81\u56fe\u4e4b\u95f4\u8fc7\u9ad8\u7684\u76f8\u5173\u6027<\/li>\n<li>InputLayer<br \/>\n\u8f93\u5165\u5c42\uff0c\u901a\u5e38\u5728\u4f7f\u7528Function API\u65b9\u5f0f\u6784\u5efa\u6a21\u578b\u65f6\u4f5c\u4e3a\u7b2c\u4e00\u5c42<\/li>\n<li>DenseFeature<br \/>\n\u7279\u5f81\u8f6c\u6362\u5c42\uff0c\u7528\u4e8e\u63a5\u6536\u4e00\u4e2a\u7279\u5f81\u5217\u8868\u5e76\u4ea7\u751f\u4e00\u4e2a\u5168\u8fde\u63a5\u5c42<\/li>\n<li>Flatten<br \/>\n\u5c55\u5e73\u5c42\uff0c\u7528\u4e8e\u5c06\u591a\u7ef4\u5f20\u91cf\u5c55\u5e73\u5f00\u538b\u5e73\u6210\u4e00\u7ef4\u5f20\u91cf<\/li>\n<li>Reshape<br \/>\n\u5f62\u72b6\u53d8\u6362\u5c42\uff0c\u53ef\u4ee5\u6539\u53d8\u5f20\u91cf\u7684\u5f62\u72b6<\/li>\n<li>Concatenate<br \/>\n\u62fc\u63a5\u5c42\uff0c\u53ef\u4ee5\u5c06\u591a\u4e2a\u5f20\u91cf\u5728\u67d0\u4e2a\u7ef4\u5ea6\u4e0a\u62fc\u63a5<\/li>\n<li>Add<br \/>\n\u52a0\u6cd5\u5c42<\/li>\n<li>Subtract<br \/>\n\u51cf\u6cd5\u5c42<\/li>\n<li>Maximum<br \/>\n\u53d6\u6700\u5927\u503c\u5c42<\/li>\n<li>Minimum<br \/>\n\u53d6\u6700\u5c0f\u503c\u5c42<\/li>\n<\/ul>\n<h2><span id=\"i-2\">\u5377\u79ef\u5c42<\/span><\/h2>\n<ul>\n<li>Conv1d<br \/>\n\u666e\u901a1\u7ef4\u5377\u79ef\u5c42\uff0c\u5e38\u7528\u4e8e\u6587\u672c\u548c\u65f6\u95f4\u5e8f\u5217\u7684\u5377\u79ef<\/li>\n<li>Conv2d<br \/>\n\u666e\u901a2\u7ef4\u5377\u79ef\u5c42\uff0c\u5e38\u7528\u4e8e\u56fe\u50cf\u7684\u7a7a\u95f4\u5377\u79ef<\/li>\n<li>Conv3d<br \/>\n\u666e\u901a3\u7ef4\u5377\u79ef\u5c42\uff0c\u5e38\u7528\u4e8e\u89c6\u9891\u6216\u4f53\u79ef\u4e0a\u7684\u7a7a\u95f4\u5377\u79ef<\/li>\n<li>SeparableConv2D<br \/>\n2\u7ef4\u6df1\u5ea6\u53ef\u5206\u79bb\u5377\u79ef\u5c42\uff0c\u4e0d\u540c\u4e8e\u666e\u901a\u5377\u79ef\u5c42\uff0c\u53ef\u4ee5\u540c\u4e8b\u5bf9\u533a\u57df\u548c\u901a\u9053\u8fdb\u884c\u64cd\u4f5c<\/li>\n<li>DepthwiseConv2D<br \/>\n2\u7ef4\u6df1\u5ea6\u5377\u79ef\u5c42\uff0c\u4ec5\u67092\u7ef4\u6df1\u5ea6\u53ef\u5206\u79bb\u5377\u79ef\u5c42\u7684\u524d\u534a\u90e8\u5206\u64cd\u4f5c<\/li>\n<li>Conv2DTranspose<br \/>\n2\u7ef4\u5377\u79ef\u8f6c\u7f6e\u5c42\uff0c\u4e5f\u79f0\u53cd\u5411\u5377\u79ef\u5c42<\/li>\n<li>LocallyConnected2D<br \/>\n2\u7ef4\u5c40\u90e8\u8fde\u63a5\u5c42<\/li>\n<li>MaxPool2D<br \/>\n2\u7ef4\u6700\u5927\u6c60\u5316\u5c42<\/li>\n<li>AveragePooling2D<br \/>\n2\u7ef4\u5e73\u5747\u6c60\u5316\u5c42<\/li>\n<li>GlobalMaxPool2D<br \/>\n\u5168\u5c40\u6700\u5927\u6c60\u5316\u5c42<\/li>\n<li>GlobalAveragePooling3D<br \/>\n\u5168\u5c40\u5e73\u5747\u6c60\u5316\u5c42<\/li>\n<\/ul>\n<h2><span id=\"i-3\">\u5faa\u73af\u7f51\u7edc\u76f8\u5173\u5c42<\/span><\/h2>\n<ul>\n<li>Embedding<br \/>\n\u5d4c\u5165\u5c42<\/li>\n<li>LSTM<br \/>\n\u957f\u77ed\u8bb0\u5fc6\u5faa\u73af\u7f51\u7edc\u5c42<\/li>\n<li>GRU<br \/>\n\u95e8\u63a7\u5faa\u73af\u5355\u5143\u5c42<\/li>\n<li>SimpleRNN<br \/>\n\u7b80\u5355\u5faa\u73af\u7f51\u7edc\u5c42<\/li>\n<li>ConvLSTM2D<br \/>\n\u5377\u79ef\u957f\u77ed\u8bb0\u5fc6\u5faa\u73af\u7f51\u7edc\u5c42<\/li>\n<li>Bidirectional<br \/>\n\u53cc\u5411\u5faa\u73af\u7f51\u7edc\u5305\u88c5\u5668<\/li>\n<li>RNN<br \/>\nRNN\u57fa\u672c\u5c42<\/li>\n<li>Dot-prodcut Attention<br \/>\nDot-product\u7c7b\u578b\u6ce8\u610f\u529b\u673a\u5236\u5c42<\/li>\n<li>AdditiveAttention<br \/>\nAdditive\u7c7b\u578b\u6ce8\u610f\u529b\u673a\u5236\u5c42<\/li>\n<li>TimeDistributed<br \/>\n\u65f6\u95f4\u5206\u5e03\u5305\u88c5\u5668\u5c42<\/li>\n<\/ul>\n<h1><span id=\"i-4\">\u81ea\u5b9a\u4e49\u5c42<\/span><\/h1>\n<p>\u53ef\u4ee5\u901a\u8fc7\u96c6\u6210Tensorflow.Keras.Engine.Layer\u7c7b\u6765\u7f16\u5199\u81ea\u5b9a\u4e49\u5c42\u3002<br \/>\n\u96c6\u6210\u8be5\u7c7b\uff0c\u91cd\u65b0\u5b9e\u73b0\u5c42\u6784\u9020\u51fd\u6570\uff0c\u5e76\u91cd\u5199Build\u548ccall\u8fd9\u4e24\u4e2a\u65b9\u6cd5\uff08build\u65b9\u6cd5\u4e00\u822c\u5b9a\u4e00\u5c42\u9700\u8981\u88ab\u8bad\u7ec3\u7684\u53c2\u6570\uff0ccall\u65b9\u6cd5\u4e00\u822c\u5b9a\u4e49\u524d\u5411\u4f20\u64ad\u8fd0\u7b97\u903b\u8f91\uff0c\u4e5f\u53ef\u4ee5\u6dfb\u52a0\u81ea\u5b9a\u4e49\u7684\u65b9\u6cd5\uff09<\/p>\n<h1><span id=\"i-5\">\u81ea\u5b9a\u4e49\u6a21\u578b<\/span><\/h1>\n<p>\u81ea\u5b9a\u4e49\u6a21\u578b\u9700\u8981\u5148\u96c6\u6210Tensorflow.Keras.Engine.Model,\u518d\u5728\u6784\u9020\u51fd\u6570\u4e2d\u521d\u59cb\u5316\u6a21\u578b\u6240\u9700\u8981\u7684\u5c42\uff0c\u5e76\u91cd\u8f7dcall()\u65b9\u6cd5\u8fdb\u884c\u6a21\u578b\u7684\u8c03\u7528\uff0c\u5efa\u7acb\u8f93\u5165\u548c\u8f93\u51fa\u4e4b\u95f4\u7684\u51fd\u6570\u5173\u7cfb<\/p>\n<h1><span id=\"API\">\u6a21\u578b\u5e38\u7528API<\/span><\/h1>\n<ol>\n<li>\u6a21\u578b\u7c7b<\/li>\n<\/ol>\n<pre><code class=\"language-CShap line-numbers\">Tensorflow.Keras.Engine.Model(ModelArgs args)\n<\/code><\/pre>\n<p>\u6a21\u578b\u7c7b\u7684\u53c2\u6570\u4e3aModelArgs\u53c2\u6570\u5217\u8868\u7c7b\uff0c\u5305\u542bInputs(\u6a21\u578b\u7684\u8f93\u5165)\u548cOutputs(\u6a21\u578b\u7684\u8f93\u51fa)<\/p>\n<ol start=\"2\">\n<li>\u6a21\u578b\u7c7b\u7684summary\u65b9\u6cd5<\/li>\n<\/ol>\n<pre><code class=\"language-CSharp line-numbers\">Model.summary(int line_length=-1,float[] positions=null)\n<\/code><\/pre>\n<p>line_length:int\u7c7b\u578b\uff0c\u6253\u5370\u7684\u603b\u884c\u6570\uff0c\u9ed8\u8ba4\u503c\u4e3a-1\uff0c\u4ee3\u8868\u6253\u5370\u6240\u6709\u5185\u5bb9<br \/>\npositions:float\u7c7b\u578b\uff0c\u6307\u5b9a\u6bcf\u4e00\u884c\u4e2d\u65e5\u5fd7\u5143\u7d20\u7684\u76f8\u5bf9\u6216\u7edd\u5bf9\u4f4d\u7f6e\uff0c\u9ed8\u8ba4\u503c\u4e3anull<\/p>\n<ol start=\"3\">\n<li>Sequential\u7c7b<\/li>\n<\/ol>\n<pre><code class=\"language-CSharp line-numbers\">Tensorflow.Keras.Engine.Sequential(SequentialArgs args)\n<\/code><\/pre>\n<ol start=\"4\">\n<li>Sequential\u7c7b\u7684add\u65b9\u6cd5<\/li>\n<\/ol>\n<pre><code class=\"language-CSharp line-numbers\">Sequential.add(Layer layer)\n<\/code><\/pre>\n<ol start=\"5\">\n<li>Sequential\u7c7b\u7684pop\u65b9\u6cd5<\/li>\n<\/ol>\n<pre><code class=\"language-CSharp line-numbers\">Sequential.pop()\n<\/code><\/pre>\n<p>\u7528\u4e8e\u5220\u9664\u7f51\u7edc\u6a21\u578b\u7684\u6700\u540e\u4e00\u4e2a\u7f51\u7edc\u5c42<br \/>\n6. compile\uff08\u7f16\u8bd1\uff09\u65b9\u6cd5<\/p>\n<pre><code class=\"language-CSharp line-numbers\">Model.compile(ILossFunc loss,OptimizerV2 optimizer,string[] metrics)\n<\/code><\/pre>\n<p>\u8be5\u65b9\u6cd5\u7528\u4e8e\u914d\u7f6e\u6a21\u578b\u7684\u8bad\u7ec3\u53c2\u6570\u3002<br \/>\n&#8211; optimizer<br \/>\n\u4f18\u5316\u5668\uff0c\u53c2\u6570\u7c7b\u578b\u4e3a\u4f18\u5316\u5668\u540d\u79f0\u7684\u5b57\u7b26\u4e32\u6216\u4f18\u5316\u5668\u5b9e\u4f8b<br \/>\n&#8211; loss<br \/>\n\u635f\u5931\u51fd\u6570\uff0c\u53c2\u6570\u7c7b\u578b\u4e3a\u635f\u5931\u51fd\u6570\u4f53\u6216Tensorflow.Keras.Losses.Loss\u5b9e\u4f8b<br \/>\n&#8211; metrics<br \/>\n\u5728\u6a21\u578b\u8bad\u7ec3\u548c\u6d4b\u8bd5\u8fc7\u7a0b\u4e2d\u91c7\u7528\u8bc4\u4f30\u6307\u5bfc\u5217\u8868\uff0c\u5217\u8868\u5143\u7d20\u7c7b\u578b\u4e3a\u8bc4\u4f30\u51fd\u6570\u540d\u79f0\u7684\u5b57\u7b26\u4e32\uff0c\u8bc4\u4f30\u51fd\u6570\u4f53\u6216Tensorflow.Keras.Metrics.Metric\u5b9e\u4f8b<br \/>\n&#8211; loss_weigths<br \/>\n\u53ef\u9009\u53c2\u6570\uff0c\u7c7b\u578b\u4e3a\u5217\u8868\u6216\u5b57\u5178\uff0c\u901a\u8fc7\u8bbe\u7f6e\u635f\u5931\u51fd\u6570\u7684\u6743\u91cd\u7cfb\u6570\u6765\u786e\u5b9a\u8f93\u51fa\u4e2d\u635f\u5931\u51fd\u6570\u7684\u8d21\u732e\u5ea6<br \/>\n&#8211; weighted_metric:<br \/>\n\u53ef\u9009\u53c2\u6570\uff0c\u7c7b\u578b\u4e3a\u5217\u8868\uff0c\u5305\u62ec\u5728\u8bad\u7ec3\u548c\u6d4b\u8bd5\u671f\u95f4\u8981\u901a\u8fc7sampleweight\u6216classweigth\u8bc4\u4f30\u548c\u52a0\u6743\u7684\u6307\u6807\u5217\u8868<br \/>\n&#8211; run_eagerly<br \/>\n\u53ef\u9009\u53c2\u6570\uff0c\u7c7b\u578b\u4e3abool,\u9ed8\u8ba4\u503c\u4e3afalse,\u5982\u679c\u8bbe\u7f6e\u4e3atrue,\u5219\u6a21\u578b\u7ed3\u6784\u5c06\u4e0d\u4f1a\u88abtf.function\u4fee\u9970\u548c\u8c03\u7528<br \/>\n&#8211; stepsperexecution<br \/>\n\u53ef\u9009\u53c2\u6570\uff0c\u7c7b\u578b\u4e3aint\uff0c\u9ed8\u8ba4\u503c\u4e3a1\uff0c\uff0c\u8868\u793a\u6bcf\u4e2atf.function\u8c03\u7528\u4e2d\u8981\u8fd0\u884c\u7684\u6279\u5904\u7406\u6570<\/p>\n<ol start=\"7\">\n<li>fit(\u8bad\u7ec3)\u65b9\u6cd5<\/li>\n<\/ol>\n<pre><code class=\"language-CSharp line-numbers\">Model.fit(NDArray x,NDArray y,int batch_size=-1,int epochs=1,int verbose=1,float validation_split=0f,bool shuffle=true,int initial_epoch=0,int max_queue_size=10,int workers=1,bool use_multiprocessing=false)\n<\/code><\/pre>\n<p>\u53c2\u6570\u5982\u4e0b\uff1a<br \/>\n&#8211; x\uff1b<br \/>\n\u8f93\u5165\u6570\u636e\uff0c\u7c7b\u578b\u4e3aNumPy\u6570\u7ec4\u6216\u5f20\u91cf\u3002<br \/>\n&#8211; y\uff1b<br \/>\n\u6807\u7b7e\u6570\u636e\uff0c\u7c7b\u578b\u548cx\u4fdd\u6301\u4e00\u81f4<br \/>\n&#8211; batch_size<br \/>\n\u6279\u6b21\u5927\u5c0f\uff0c\u7c7b\u578b\u4e3aint,\u6307\u5b9a\u6bcf\u6b21\u68af\u5ea6\u66f4\u65b0\u65f6\u6bcf\u4e2a\u6837\u672c\u6279\u6b21\u7684\u6570\u91cf<br \/>\n&#8211; epochs<br \/>\n\u8bad\u7ec3\u8f6e\u6b21\u6570\uff0c\u7c7b\u578b\u4e3aint,\u8868\u793a\u6a21\u578b\u8bad\u7ec3\u7684\u8fed\u4ee3\u8f6e\u6570<br \/>\n&#8211; verbose<br \/>\n\u8be6\u7ec6\u5ea6\uff0c\u7c7b\u578b\u4e3a\u6574\u65700,1\u62162\uff0c\u8bbe\u7f6e\u7ec8\u7aef\u8f93\u51fa\u4fe1\u606f\u7684\u8be6\u7ec6\u7a0b\u5ea6\u30020\u4f4d\u6700\u7b80\u6a21\u5f0f\uff0c1\u4f4d\u7b80\u5355\u663e\u793a\u8fdb\u5ea6\u4fe1\u606f\uff1b2\u4f4d\u8be6\u7ec6\u663e\u793a\u6bcf\u4e2a\u8bad\u7ec3\u5468\u671f\u7684\u6570\u636e<br \/>\n&#8211; validation_split<br \/>\n\u9a8c\u8bc1\u96c6\u5212\u5206\u6bd4\uff0c\u7c7b\u578b\u4e3a0~1\u4e4b\u95f4\u7684float,\u81ea\u52a8\u4ece\u8bad\u7ec3\u96c6\u4e2d\u5212\u5206\u51fa\u8be5\u6bd4\u4f8b\u7684\u6570\u636e\u4f5c\u4e3a\u9a8c\u8bc1\u96c6\u3002\u6a21\u578b\u5c06\u5265\u79bb\u8bad\u7ec3\u6570\u636e\u8fd9\u4e00\u90e8\u5206\uff0c\u4e0d\u5bf9\u5176\u8fdb\u884c\u8bad\u7ec3\uff0c\u5e76\u4e14\u5c06\u5728\u6bcf\u4e2a\u8f6e\u6b21\u7ed3\u675f\u65f6\u8bc4\u4f30\u9a8c\u8bc1\u96c6\u6570\u636e\u7684\u635f\u5931<br \/>\n&#8211; shuffle<br \/>\n\u6570\u636e\u96c6\u4e71\u5e8f\uff0c\u7c7b\u578b\u4e3abool\uff0c\u9ed8\u8ba4\u503c\u4e3atrue,\u8868\u793a\u6bcf\u6b21\u8bad\u7ec3\u524d\u90fd\u5c06\u8be5\u6279\u6b21\u4e2d\u7684\u6570\u636e\u968f\u673a\u6253\u4e71\u3002<br \/>\n&#8211; initial_epoch<br \/>\n\u5468\u671f\u8d77\u70b9\uff0c\u7c7b\u578b\u4e3aint,\u9ed8\u8ba4\u503c\u4e3a0,\u8868\u793a\u8bad\u7ec3\u5f00\u59cb\u7684\u8f6e\u6b21\u8d77\u70b9\uff08\u5e38\u7528\u8bed\u91cd\u542f\u6062\u590d\u4e4b\u524d\u8bad\u7ec3\u4e2d\u7684\u6a21\u578b\uff09<br \/>\n&#8211; max_queue_size<br \/>\n\u961f\u5217\u6700\u5927\u5bb9\u91cf\uff0c\u9ed8\u8ba4\u4e3a10\uff0c\u8bbe\u7f6e\u6570\u636e\u751f\u6210\u5668\u961f\u5217\u7684\u6700\u5927\u5bb9\u91cf<br \/>\n&#8211; workers<br \/>\n\u8fdb\u7a0b\u6570\uff0c\u7c7b\u578b\u4e3aint,\u9ed8\u8ba4\u503c\u4e3a1\uff0c\u8bbe\u7f6e\u5728\u4f7f\u7528\u57fa\u4e8e\u8fdb\u7a0b\u7684\u7ebf\u7a0b\u65f6\uff0c\u8981\u542f\u52a8\u7684\u6700\u5927\u8fdb\u7a0b\u6570\u3002\u5982\u679c\u672a\u6307\u5b9a\uff0c\u5219workers\u7684\u9ed8\u8ba4\u503c\u4e3a1\uff1b\u5982\u679c\u4e3a0\uff0c\u5219\u5c06\u5728\u4e3b\u7ebf\u7a0b\u4e0a\u6267\u884c\u6570\u636e\u751f\u6210\u5668<br \/>\n&#8211; use_multipprocessing:<br \/>\n\u662f\u5426\u4f7f\u7528\u5f02\u6b65\u7ebf\u7a0b\uff0c\u7c7b\u578b\u4e3abool,\u9ed8\u8ba4\u503c\u4e3aflase\u3002\u5982\u679ctrue\uff0c\u5219\u9700\u8981\u4f7f\u7528\u57fa\u4e8e\u8fdb\u7a0b\u7684\u7ebf\u7a0b<br \/>\n8. evaluate(\u8bc4\u4f30)\u65b9\u6cd5<\/p>\n<pre><code class=\"language-CSharp line-numbers\">Model.evaluate(NDArray x,NDArray y,int batch_size=-1,int verbose=1,int steps=-1,int max_queue_size=10,int workers=1,bool use_multiprocessing=false,bool return_dict=false)\n<\/code><\/pre>\n<p>evaluate\u65b9\u6cd5\u7528\u4e8e\u8fd4\u56de\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30\u7684\u6a21\u578b\u635f\u5931\u548c\u7cbe\u5ea6\u3002<br \/>\n&#8211; x<br \/>\n\u8f93\u5165\u7684\u6d4b\u8bd5\u6570\u636e\uff0c\u7c7b\u578b\u4e3aNumPy\u6570\u7ec4\u6216\u5f20\u91cf<br \/>\n&#8211; y<br \/>\n\u6d4b\u8bd5\u7684\u6807\u7b7e\u6570\u636e\uff0c\u7c7b\u578b\u548cx\u4fdd\u6301\u4e00\u81f4<br \/>\n&#8211; batch_size<br \/>\n\u6279\u6b21\u5927\u5c0f\uff0c\u7c7b\u578b\u4e3aint,\u6307\u5b9a\u6bcf\u6b21\u8ba1\u7b97\u65f6\u6bcf\u4e2a\u6837\u672c\u6279\u6b21\u7684\u6570\u91cf<br \/>\n&#8211; verbose<br \/>\n\u8be6\u7ec6\u5ea6\uff0c\u7c7b\u578b\u4e3a\u6574\u65700\u62161\uff0c\u8bbe\u7f6e\u7ec8\u7aef\u8f93\u51fa\u4fe1\u606f\u7684\u8be6\u7ec6\u7a0b\u5ea6\u30020\u4f4d\u6700\u7b80\u6a21\u5f0f\uff1b1\u4f4d\u7b80\u5355\u73b0\u5b9e\u8fdb\u5ea6\u4fe1\u606f<br \/>\n&#8211; steps<br \/>\n\u8fed\u4ee3\u6b65\u6570\uff0c\u7c7b\u578b\u4e3aint,\u9ed8\u8ba4\u503c\u4e3a-1\uff0c\u8bbe\u7f6e\u8bc4\u4f30\u5468\u671f\u5185\u7684\u8fed\u4ee3\u6b65\u6570\uff08\u6837\u672c\u6279\u6b21\u6570\u91cf\uff09<br \/>\n&#8211; max_queue_size<br \/>\n\u961f\u5217\u6700\u5927\u5bb9\u91cf\uff0c\u7c7b\u578b\u4e3aint,\u9ed8\u8ba4\u503c\u4e3a10\uff0c\u8bbe\u7f6e\u6570\u636e\u751f\u6210\u5668\u961f\u5217\u7684\u6700\u5927\u5bb9\u91cf\u3002<br \/>\n&#8211; max_queue_size<br \/>\n\u961f\u5217\u6700\u5927\u5bb9\u91cf\uff0c\u7c7b\u578b\u4e3aint,\u9ed8\u8ba4\u503c\u4e3a10\uff0c\u8bbe\u7f6e\u6570\u636e\u751f\u6210\u5668\u961f\u5217\u7684\u6700\u5927\u5bb9\u91cf\u3002<br \/>\n&#8211; workers<br \/>\n\u8fdb\u7a0b\u6570\uff0c\u7c7b\u578b\u4e3aint,\u9ed8\u8ba4\u503c\u4e3a1\uff0c\u8bbe\u7f6e\u5728\u4f7f\u7528\u57fa\u4e8e\u8fdb\u7a0b\u7684\u7ebf\u7a0b\u65f6\uff0c\u8981\u542f\u52a8\u7684\u6700\u5927\u8fdb\u7a0b\u6570\u3002\u5982\u679c\u672a\u6307\u5b9a\uff0c\u5219workers\u7684\u9ed8\u8ba4\u503c\u4e3a1\uff0c\u5982\u679c\u672a0\uff0c\u5219\u5c06\u5728\u4e3b\u7ebf\u7a0b\u4e0a\u6267\u884c\u6570\u636e\u751f\u6210\u5668<br \/>\n&#8211; use_multiprocessing<br \/>\n\u662f\u5426\u4f7f\u7528\u5f02\u6b65\u7ebf\u7a0b\uff0c\u7c7b\u578b\u4e3abool\uff0c\u9ed8\u8ba4\u503c\u4e3aflase\u3002\u5982\u679c\u4e3atrue,\u5219\u9700\u8981\u4f7f\u7528\u57fa\u4e8e\u8fdb\u7a0b\u7684\u7ebf\u7a0b<br \/>\n&#8211; return_dict<br \/>\n\u8fd4\u56de\u5b57\u5178\uff0c\u7c7b\u578b\u4e3abool,\u9ed8\u8ba4\u503c\u4e3aflase.\u5982\u679c\u4e3atrue,\u5219\u8fd4\u56de\u5b57\u5178\u7c7b\u578b\u7684\u635f\u5931\u548c\u6307\u6807\u7ed3\u679c\uff0c\u5b57\u5178\u7684key\u4e3a\u6570\u636e\u7684\u540d\u79f0\u91cc\u5982\u679c\u4e3aflase,\u5219\u6b63\u5e38\u653e\u56de\u5217\u8868\u7c7b\u578b\u7684\u7ed3\u679c<br \/>\n9. predict(\u9884\u6d4b\u65b9\u6cd5)<\/p>\n<pre><code class=\"language-CSharp line-numbers\">Model.predict(Tensor x,int batch_size=32,int verbose=0,int steps=-1,int steps=-1,int max_queue_size=10,int workers=1,bool use_multiprocessing=false)\n<\/code><\/pre>\n<p>predict\u65b9\u6cd5\u7528\u4e8e\u751f\u6210\u8f93\u5165\u6837\u672c\u7684\u9884\u6d4b\u8f93\u51fa<br \/>\n&#8211; x<br \/>\n\u8f93\u5165\u7684\u6d4b\u8bd5\u6570\u636e\uff0c\u7c7b\u578b\u4e3a\u5f20\u91cf<br \/>\n&#8211; batch_size<br \/>\n\u6279\u6b21\u5927\u5c0f\uff0c\u7c7b\u578b\u4e3aint\u3002\u5982\u679c\u672a\u6307\u5b9a\uff0c\u5219batch_size\u7684\u9ed8\u8ba4\u503c\u4e3a32\uff0c\u8be5\u53c2\u6570\u6307\u5b9a\u4e86\u6bcf\u6b21\u8ba1\u7b97\u65f6\u6bcf\u4e2a\u6837\u672c\u6279\u6b21\u7684\u6570\u91cf<br \/>\n&#8211; verbose<br \/>\n\u8be6\u7ec6\u5ea6\uff0c\u7c7b\u578b\u4e3a\u6574\u65700\u62161\uff0c\u9ed8\u8ba4\u503c\u4e3a0\uff0c\u8bbe\u7f6e\u7ec8\u7aef\u8f93\u51fa\u4fe1\u606f\u7684\u8be6\u7ec6\u7a0b\u5ea6\u30020\u4f4d\u6700\u7b80\u6a21\u5f0f\uff0c1\u4e3a\u7b80\u5355\u663e\u793a\u8fdb\u5ea6\u4fe1\u606f<br \/>\n&#8211; steps<br \/>\n\u8fed\u4ee3\u6b65\u6570\uff0c\u7c7b\u578b\u4e3a\u6574\u6570\uff0c\u9ed8\u8ba4\u503c\u4e3a-1\uff0c\u8bbe\u7f6e\u9884\u6d4b\u5468\u671f\u5185\u7684\u8fed\u4ee3\u6b65\u6570\uff08\u6837\u672c\u6279\u6b21\u6570\u91cf\uff09<br \/>\n&#8211; max_queue_size<br \/>\n\u961f\u5217\u6700\u5927\u5bb9\u91cf\uff0c\u7c7b\u578b\u4e3aint,\u9ed8\u8ba4\u503c\u4e3a10\uff0c\u8bbe\u7f6e\u6570\u636e\u751f\u6210\u5668\u961f\u5217\u7684\u6700\u5927\u5bb9\u91cf<br \/>\n&#8211; worers<br \/>\n\u8fdb\u7a0b\u6570\uff0c\u7c7b\u578b\u4e3aint\uff0c\u9ed8\u8ba4\u503c\u4e3a1\uff0c\u8bbe\u7f6e\u5728\u4f7f\u7528\u57fa\u4e8e\u8fdb\u7a0b\u7684\u7ebf\u7a0b\u65f6\uff0c\u8981\u542f\u52a8\u7684\u6700\u5927\u8fdb\u7a0b\u6570\uff0c\u5982\u679c\u672a\u6307\u5b9a\uff0c\u5219workers\u7684\u9ed8\u8ba4\u503c\u4e3a1\uff0c\u5982\u679c\u672a0\u5219\u5c06\u5728\u4e3b\u7ebf\u7a0b\u4e0a\u6267\u884c\u6570\u636e\u751f\u6210\u5668<br \/>\n&#8211; use_multiprocessing<br \/>\n\u662f\u5426\u4f7f\u7528\u5f02\u6b65\u7ebf\u7a0b\uff0c\u7c7b\u578b\u4e3abool,\u9ed8\u8ba4\u503c\u4e3afalse\u3002\u5982\u679c\u4e3atrue,\u5219\u9700\u8981\u4f7f\u7528\u57fa\u4e8e\u8fdb\u7a0b\u7684\u7ebf\u7a0b\u3002<\/p>\n<p>\u8fd4\u56de\u5f20\u91cf\u7c7b\u578b\u7684\u9884\u6d4b\u6570\u7ec4<\/p>\n<ol start=\"10\">\n<li>save(\u6a21\u578b\u4fdd\u5b58)\u65b9\u6cd5<\/li>\n<\/ol>\n<pre><code class=\"language-CSharp line-numbers\">Model.save(string filepath,bool overwrite=true,bool include_optimizer=true,string save_format=\"tf\",SaveOptions options=null)\n<\/code><\/pre>\n<p>save\u65b9\u6cd5\u7528\u4e8e\u5c06\u6a21\u578b\u4fdd\u5b58\u5230TensorFlow\u7684SavedModel\u6216\u5355\u4e2aH5\u6587\u4ef6\u4e2d<br \/>\n&#8211; filepath<br \/>\n\u6a21\u578b\u6587\u4ef6\u8def\u5f84\uff0c\u7c7b\u578b\u4e3astring,\u6a21\u578b\u4fdd\u5b58SavedModel\u6216H5\u6587\u4ef6\u7684\u8def\u5f84\u3002<br \/>\n&#8211; overwrite<br \/>\n\u662f\u5426\u8986\u76d6\uff0c\u7c7b\u578b\u4e3abool,\u82e5\u8bbe\u7f6e\u4e3atrue,\u5219\u4ee5\u9ed8\u8ba4\u65b9\u5f0f\u8986\u76d6\u76ee\u6807\u4f4d\u7f6e\u4e0a\u7684\u4efb\u4f55\u73b0\u6709\u6587\u4ef6\uff1b\u82e5\u8bbe\u7f6e\u4e3afalse,\u5219\u5411\u7528\u6237\u63d0\u4f9b\u624b\u52a8\u63d0\u793a\u3002<br \/>\n&#8211; include_optimizer<br \/>\n\u662f\u5426\u5305\u542b\u4f18\u5316\u5668\u4fe1\u606f\uff0c\u7c7b\u578b\u4e3abool\uff0c\u5982\u679c\u4e3atrue,\u5219\u5c06\u4f18\u5316\u5668\u7684\u72b6\u6001\u4fdd\u5b58\u5728\u4e00\u8d77<br \/>\n&#8211; save_format<br \/>\n\u4fdd\u5b58\u683c\u5f0f\uff0c\u7c7b\u578b\u4e3astring,\u53ef\u9009\u62e9tf\u6216h5,\u6307\u793a\u662f\u5426\u5c06\u6a21\u578b\u4fdd\u5b58\u5230TensorFlow\u7684SavedModel\u6216H5\u6587\u4ef6\u4e2d\uff0c\u5728TensorFlow2.x\u4e2d\u9ed8\u8ba4\u4e3atf<br \/>\n&#8211; SaveOptions<br \/>\n\u7c7b\u578b\u4e3aTensorFlow.ModelSaveing.SaveOptiongs\u5bf9\u8c61\uff0c\u7528\u4e8e\u6307\u5b9a\u4fdd\u5b58\u5230SavedModel\u7684\u9009\u9879<br \/>\n11. load model(\u6a21\u578b\u8f7d\u5165)\u65b9\u6cd5<br \/>\nload_model\u65b9\u6cd5\u7528\u4e8e\u52a0\u8f7d\u901a\u8fc7save\u65b9\u6cd5\u4fdd\u5b58\u7684\u6a21\u578b<br \/>\n&#8211; get_weights\u65b9\u6cd5<br \/>\n&#8211; set_weights\u65b9\u6cd5<br \/>\n&#8211; save_weights\u65b9\u6cd5<br \/>\n&#8211; load_weights\u65b9\u6cd5<br \/>\n&#8211; get_config\u65b9\u6cd5<br \/>\n&#8211; from_config\u65b9\u6cd5<br \/>\n&#8211; model_from_config\u529f\u80fd<br \/>\n&#8211; to_json\u65b9\u6cd5<br \/>\n&#8211; model_from_json\u529f\u80fd<br \/>\n&#8211; clone_model\u529f\u80fd<\/p>\n<h1><span id=\"KerasAPI\">Keras\u5e38\u7528API\u8bf4\u660e<\/span><\/h1>\n<h2><span id=\"Callbacks\">\u56de\u8c03\u51fd\u6570Callbacks<\/span><\/h2>\n<p>\u7528\u5728Model.fit()\u4e2d\u4f5c\u4e3a\u53c2\u6570\uff0c\u53ef\u4ee5\u5728\u8bad\u7ec3\u7684\u5404\u4e2a\u9636\u6bb5\u6267\u884c\u4e00\u5b9a\u64cd\u4f5c<br \/>\n&#8211; \u6bcf\u4e2a\u6279\u6b21\u8bad\u7ec3\u540e\u5199\u5165TensorBoard\u65e5\u5fd7\u4ee5\u76d1\u63a7\u6307\u6807<br \/>\n&#8211; \u5b9a\u671f\u5c06\u6a21\u578b\u4fdd\u5b58\u5230\u672c\u5730\u6587\u4ef6\u4e2d<br \/>\n&#8211; \u63d0\u524d\u7ed3\u675f\u8bad\u7ec3<br \/>\n&#8211; \u5728\u8bad\u7ec3\u5668\u95f4\u67e5\u770b\u6a21\u578b\u7684\u5185\u90e8\u72b6\u6001\u548c\u7edf\u8ba1\u4fe1\u606f<br \/>\n&#8211; \u66f4\u591a\u5176\u4ed6\u529f\u80fd<\/p>\n<h2><span id=\"i-6\">\u57fa\u7840\u7528\u6cd5<\/span><\/h2>\n<p>\u53ef\u4ee5\u5c06\u56de\u8c03\u5217\u8868\u4f5c\u4e3a\u56de\u8c03\u53c2\u6570\u4f20\u64ad\u7ed9\u6a21\u578b\u7684Model.fit()\u3002\u8fd9\u6837\u5c31\u53ef\u4ee5\u5728\u8bad\u7ec3\u7684\u6bcf\u4e2a\u9636\u6bb5\u81ea\u52a8\u8c03\u7528\u56de\u8c03\u5217\u8868\u5b9a\u4e49\u7684\u76f8\u5173\u65b9\u6cd5<\/p>\n<h2><span id=\"i-7\">\u5df2\u5b9a\u4e49\u7684\u56de\u8c03\u51fd\u6570\u7c7b<\/span><\/h2>\n<p>kertas.callbacks\u5b50\u6a21\u5757\u4e2d\u5df2\u7ecf\u5b9a\u4e49\u597d\u7684\u56de\u8c03\u51fd\u6570\u7c7b<\/p>\n<ol>\n<li>\n<p>ketas.callbacks.Callbacks<br \/>\n\u7528\u4e8e\u5efa\u7acb\u65b0\u56de\u8c03\u51fd\u6570\u7684\u62bd\u8c61\u57fa\u7c7b\uff0c\u6240\u6709\u56de\u8c03\u51fd\u6570\u90fd\u7ee7\u627f\u81eakeras.callbacks.Callbacks\u57fa\u7c7b\uff0c\u4ed6\u62e5\u6709params\u548cmodel\u4e24\u4e2a\u5c5e\u6027\u5176\u4e2d\uff0cparams\u662f\u4e00\u4e2adict,\u8bb0\u5f55\u4e86\u8bad\u7ec3\u76f8\u5173\u53c2\u6570\uff08\u5982verbosity\u3001batch size\u3001number of epochs\u7b49\uff09\uff1bmodel\u5373\u5f53\u524d\u5173\u8054\u6a21\u578b\u7684\u5f15\u7528<\/p>\n<\/li>\n<li>\n<p>ModelCheckpoint<br \/>\n\u7528\u7279\u5b9a\u9891\u7387\u4fdd\u5b58Keras\u6a21\u578b\u6216\u6a21\u578b\u6743\u91cd<\/p>\n<\/li>\n<li>TensorBoard<br \/>\n\u81ea\u5e26\u7684\u53ef\u89c6\u5316\u5de5\u5177<\/li>\n<li>EarlyStopping<br \/>\n\u7528\u4e8e\u5f53\u8d1d\u76d1\u63a7\u6307\u6807\u5728\u8bbe\u5b9a\u7684\u82e5\u5e72\u4e2a\u8f6e\u6b21\u540e\u6ca1\u6709\u63d0\u5347\u65f6\uff0c\u63d0\u524d\u7ec8\u6b62\u8bad\u7ec3<\/li>\n<li>LearningRateScheduler<br \/>\n\u5b66\u4e60\u63a7\u5236\u5668\uff0c\u7ed9\u5b9a\u5b66\u4e60\u7387\u548c\u8f6e\u6b21\u7684\u51fd\u6570\u5173\u7cfb\uff0c\u4f1a\u6839\u636e\u8be5\u51fd\u6570\u5173\u7cfb\u5728\u6bcf\u4e2a\u8f6e\u6b21\u524d\u66f4\u65b0\u5b66\u4e60\u7387<\/li>\n<li>ReduceLROnPlateau<br \/>\n\u7528\u4e8e\u5f53\u8bbe\u7f6e\u7684\u6307\u6807\u505c\u6b62\u6539\u5584\u65f6\uff0c\u81ea\u52a8\u964d\u4f4e\u5b66\u4e60\u7387<\/li>\n<li>RemoteMonitor<br \/>\n\u7528\u4e8e\u5c06\u4e8b\u4ef6\u6d41\u4f20\u8f93\u5230\u670d\u52a1\u5668\u7aef\u8fdb\u884c\u663e\u793a<\/li>\n<li>LambdaCallback<br \/>\n\u7f16\u5199\u8f83\u4e3a\u7b80\u5355\u7684\u81ea\u5b9a\u4e49\u56de\u8c03\u51fd\u6570<\/li>\n<li>TerminateOnNan<br \/>\n\u7528\u4e8e\u5f53\u8bad\u7ec3\u8fc7\u7a0b\u9047\u5230\u635f\u5931\u503c\u4e3aNaN\u65f6\uff0c\u81ea\u52a8\u7ec8\u6b62<\/li>\n<li>CSVLogger<br \/>\n\u5c06\u6bcf\u4e2a\u8f6e\u6b21\u540e\u7684log\u7ed3\u679c\u4ee5streams\u6d41\u683c\u5f0f\u8bb0\u5f55\u5230\u672c\u5730CSV\u6587\u4ef6\u4e2d<\/li>\n<li>ProgbarLogger<br \/>\n\u5c06\u6bcf\u4e2a\u8f6e\u6b21\u7ed3\u675f\u540e\u7684Log\u7ed3\u679c\u6253\u5370\u5230\u6807\u51c6\u8f93\u51fa\u6d41\u4e2d<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>Contents1 Keras\u5e38\u7528\u7684\u6a21\u578b\u4e0e\u5c421.1 \u57fa\u7840\u5c421.2 \u5377\u79ef\u5c421.3 \u5faa\u73af\u7f51\u7edc\u76f8\u5173\u5c422 \u81ea\u5b9a\u4e49\u5c423  [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[50,49,2],"tags":[],"class_list":["post-1307","post","type-post","status-publish","format-standard","hentry","category-tensorflow","category-49","category-2"],"_links":{"self":[{"href":"https:\/\/www.wayln.com\/index.php?rest_route=\/wp\/v2\/posts\/1307","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wayln.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wayln.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wayln.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wayln.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1307"}],"version-history":[{"count":8,"href":"https:\/\/www.wayln.com\/index.php?rest_route=\/wp\/v2\/posts\/1307\/revisions"}],"predecessor-version":[{"id":1315,"href":"https:\/\/www.wayln.com\/index.php?rest_route=\/wp\/v2\/posts\/1307\/revisions\/1315"}],"wp:attachment":[{"href":"https:\/\/www.wayln.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1307"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wayln.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1307"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wayln.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1307"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}