{"id":103555,"date":"2026-09-10T23:09:10","date_gmt":"2026-09-10T15:09:10","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/103555.html"},"modified":"2026-09-10T23:09:10","modified_gmt":"2026-09-10T15:09:10","slug":"%e7%ac%ac38%e8%af%be%ef%bc%9atensorflow%ef%bd%9c%e5%8f%af%e8%a7%86%e5%8c%96%e5%b7%a5%e5%85%b7tensorboard%e5%85%a8%e7%94%a8%e6%b3%95%e3%80%90%e6%97%a5%e5%bf%97%e5%86%99%e5%85%a5%e3%80%81%e6%8c%87","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/103555.html","title":{"rendered":"\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011"},"content":{"rendered":"<p><img decoding=\"async\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260910150908-6aa2c81445d73.jpg\" alt=\"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\" \/><\/p>\n<\/p>\n<h4>\u6587\u7ae0\u76ee\u5f55<\/h4>\n<ul>\n<li>\n<ul>\n<li>1. \u8bfe\u524d\u5bfc\u8bfb<\/li>\n<li>\n<ul>\n<li>1.1 \u672c\u8282\u8bfe\u5b66\u4e60\u76ee\u6807<\/li>\n<li>1.2 \u77e5\u8bc6\u91cd\u96be\u70b9<\/li>\n<li>1.3 \u5b66\u4e60\u524d\u7f6e\u6761\u4ef6<\/li>\n<li>1.4 \u5b66\u5b8c\u53ef\u638c\u63e1\u80fd\u529b<\/li>\n<li>1.5 \u884c\u4e1a\u5e94\u7528\u573a\u666f<\/li>\n<\/ul>\n<\/li>\n<li>2. \u6838\u5fc3\u7406\u8bba\u7cbe\u8bb2<\/li>\n<li>\n<ul>\n<li>2.1 TensorBoard \u67b6\u6784\u4e0e\u65e5\u5fd7<\/li>\n<li>2.2 \u6807\u91cf\u4eea\u8868\u677f&#xff08;Scalar&#xff09;<\/li>\n<li>2.3 \u8ba1\u7b97\u56fe\u53ef\u89c6\u5316<\/li>\n<li>2.4 \u76f4\u65b9\u56fe\u4eea\u8868\u677f&#xff08;Histogram&#xff09;<\/li>\n<li>2.5 \u5d4c\u5165\u6295\u5f71\u4eea&#xff08;Embedding Projector&#xff09;<\/li>\n<li>2.6 \u8d85\u53c2\u6570\u8c03\u4f18&#xff08;HParams&#xff09;<\/li>\n<li>2.7 \u6027\u80fd\u5206\u6790&#xff08;Profile&#xff09;<\/li>\n<\/ul>\n<\/li>\n<li>3. \u73af\u5883\u642d\u5efa\u4e0e\u5de5\u5177\u914d\u7f6e<\/li>\n<li>4. \u4ee3\u7801\u5b9e\u6218\u6559\u5b66<\/li>\n<li>\n<ul>\n<li>4.1 \u57fa\u672c\u7528\u6cd5&#xff1a;Keras\u56de\u8c03<\/li>\n<li>4.2 \u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d\u624b\u52a8\u5199\u5165\u65e5\u5fd7<\/li>\n<li>4.3 \u53ef\u89c6\u5316\u8bcd\u5411\u91cf&#xff08;\u5d4c\u5165\u6295\u5f71\u4eea&#xff09;<\/li>\n<li>4.4 \u8d85\u53c2\u6570\u8c03\u4f18&#xff08;HParams&#xff09;<\/li>\n<li>4.5 \u6027\u80fd\u5206\u6790&#xff08;Profile&#xff09;<\/li>\n<\/ul>\n<\/li>\n<li>5. \u6848\u4f8b\u5b9e\u64cd\u6f14\u7ec3<\/li>\n<li>\n<ul>\n<li>5.1 \u6570\u636e\u52a0\u8f7d\u4e0e\u6a21\u578b\u6784\u5efa<\/li>\n<li>5.2 \u8d85\u53c2\u6570\u641c\u7d22\u4e0e\u8bb0\u5f55<\/li>\n<li>5.3 \u7ed3\u679c\u5206\u6790<\/li>\n<\/ul>\n<\/li>\n<li>6. \u5e38\u89c1\u5751\u70b9\u4e0e\u6392\u9519\u603b\u7ed3<\/li>\n<li>\n<ul>\n<li>6.1 \u65e5\u5fd7\u5199\u5165\u5751\u70b9<\/li>\n<li>6.2 \u76f4\u65b9\u56fe\u7406\u89e3\u8bef\u533a<\/li>\n<li>6.3 \u8ba1\u7b97\u56fe\u53ef\u89c6\u5316<\/li>\n<li>6.4 \u5d4c\u5165\u6295\u5f71\u4eea<\/li>\n<li>6.5 Profile<\/li>\n<\/ul>\n<\/li>\n<li>7. \u77e5\u8bc6\u70b9\u603b\u7ed3 &#043; \u8bfe\u540e\u4f5c\u4e1a<\/li>\n<li>\n<ul>\n<li>7.1 \u6838\u5fc3\u77e5\u8bc6\u70b9\u68b3\u7406<\/li>\n<li>7.2 \u57fa\u7840\u4f5c\u4e1a<\/li>\n<li>7.3 \u8fdb\u9636\u5b9e\u64cd\u4f5c\u4e1a<\/li>\n<li>7.4 \u601d\u8003\u62d3\u5c55\u9898<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/li>\n<li>&#x1f517;\u300aTensorFlow2.x: \u6df1\u5ea6\u5b66\u4e60\u5165\u95e8\u5230\u9ad8\u9636\u5b9e\u6218\u6559\u7a0b\u300b\u7cfb\u5217\u8bfe\u7a0b\u5bfc\u822a<\/li>\n<\/ul>\n<hr \/>\n<h3>1. \u8bfe\u524d\u5bfc\u8bfb<\/h3>\n<h4>1.1 \u672c\u8282\u8bfe\u5b66\u4e60\u76ee\u6807<\/h4>\n<ul>\n<li>\u638c\u63e1TensorBoard\u7684\u542f\u52a8\u65b9\u5f0f&#xff08;\u547d\u4ee4\u884c\u3001Notebook\u5185\u5d4c&#xff09;\u548c\u65e5\u5fd7\u76ee\u5f55\u7ed3\u6784\u3002<\/li>\n<li>\u5b66\u4f1a\u5728Keras\u7684model.fit\u4e2d\u4f7f\u7528TensorBoard\u56de\u8c03&#xff0c;\u8bb0\u5f55\u635f\u5931\u3001\u81ea\u5b9a\u4e49\u6307\u6807\u3001\u5b66\u4e60\u7387\u7b49\u3002<\/li>\n<li>\u7406\u89e3\u8ba1\u7b97\u56fe\u53ef\u89c6\u5316\u7684\u610f\u4e49&#xff0c;\u80fd\u591f\u67e5\u770b\u6a21\u578b\u7684\u7ed3\u6784\u548c\u6bcf\u5c42\u7684\u8f93\u5165\u8f93\u51fa\u5f62\u72b6\u3002<\/li>\n<li>\u638c\u63e1\u76f4\u65b9\u56fe\u7684\u4f7f\u7528&#xff0c;\u89c2\u5bdf\u6743\u91cd\u3001\u68af\u5ea6\u3001\u6fc0\u6d3b\u503c\u7684\u5206\u5e03\u53d8\u5316&#xff0c;\u8bca\u65ad\u68af\u5ea6\u6d88\u5931\/\u7206\u70b8\u3002<\/li>\n<li>\u5b66\u4f1a\u4f7f\u7528\u5d4c\u5165\u6295\u5f71\u4eea&#xff08;Embedding Projector&#xff09;\u53ef\u89c6\u5316\u9ad8\u7ef4\u7279\u5f81\u5411\u91cf&#xff08;\u5982\u8bcd\u5411\u91cf&#xff09;\u3002<\/li>\n<li>\u80fd\u591f\u4f7f\u7528HParams\u4eea\u8868\u677f\u8fdb\u884c\u8d85\u53c2\u6570\u8c03\u4f18\u5bf9\u6bd4\u3002<\/li>\n<li>\u4f7f\u7528Profile\u5de5\u5177\u5206\u6790\u8bad\u7ec3\u6027\u80fd\u74f6\u9888&#xff08;\u6570\u636e\u52a0\u8f7d vs \u8ba1\u7b97&#xff09;\u3002<\/li>\n<\/ul>\n<h4>1.2 \u77e5\u8bc6\u91cd\u96be\u70b9<\/h4>\n<table>\n<tr>\u7c7b\u522b\u5185\u5bb9<\/tr>\n<tbody>\n<tr>\n<td>\u91cd\u70b9<\/td>\n<td>TensorBoard\u56de\u8c03\u914d\u7f6e&#xff1b;\u6807\u91cf\u76d1\u63a7&#xff08;\u635f\u5931\u3001\u51c6\u786e\u7387&#xff09;&#xff1b;\u8ba1\u7b97\u56fe\u5bfc\u51fa&#xff1b;\u76f4\u65b9\u56fe\u7684\u4f7f\u7528&#xff1b;\u5d4c\u5165\u6295\u5f71\u4eea\u7684\u57fa\u672c\u6d41\u7a0b<\/td>\n<\/tr>\n<tr>\n<td>\u96be\u70b9<\/td>\n<td>\u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d\u624b\u52a8\u5199\u5165\u65e5\u5fd7&#xff08;tf.summary&#xff09;&#xff1b;Profile\u7684\u89e3\u8bfb\u4e0e\u6027\u80fd\u74f6\u9888\u5b9a\u4f4d&#xff1b;\u76f4\u65b9\u56fe\u4e0e\u5206\u5e03\u7684\u5bf9\u5e94\u5173\u7cfb<\/td>\n<\/tr>\n<tr>\n<td>\u6613\u6df7\u6dc6\u70b9<\/td>\n<td>tf.summary.scalar\u4e0etf.summary.histogram\u7684\u4f5c\u7528\u57df&#xff1b;\u65e5\u5fd7\u76ee\u5f55\u6bcf\u6b21\u8fd0\u884c\u9700\u4e0d\u540c&#xff0c;\u5426\u5219\u66f2\u7ebf\u6df7\u6dc6&#xff1b;Profile\u4ec5\u9002\u7528\u4e8eGPU\/TPU<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h4>1.3 \u5b66\u4e60\u524d\u7f6e\u6761\u4ef6<\/h4>\n<ul>\n<li>\u5df2\u5b8c\u6210\u524d\u51e0\u8bfe\u7684\u6a21\u578b\u8bad\u7ec3&#xff08;\u5982CNN\u3001RNN&#xff09;\u3002<\/li>\n<li>\u80fd\u591f\u642d\u5efa\u7b80\u5355\u7684Keras\u6a21\u578b\u5e76\u8bad\u7ec3\u3002<\/li>\n<li>\u4e86\u89e3\u8bcd\u5411\u91cf\u7684\u57fa\u672c\u6982\u5ff5&#xff08;\u7b2c30\u8bfe&#xff09;\u3002<\/li>\n<\/ul>\n<h4>1.4 \u5b66\u5b8c\u53ef\u638c\u63e1\u80fd\u529b<\/h4>\n<ul>\n<li>\u72ec\u7acb\u4f7f\u7528TensorBoard\u76d1\u63a7\u8bad\u7ec3\u8fc7\u7a0b&#xff0c;\u5feb\u901f\u8bc6\u522b\u8fc7\u62df\u5408\u3001\u6b20\u62df\u5408\u3002<\/li>\n<li>\u901a\u8fc7\u76f4\u65b9\u56fe\u89c2\u5bdf\u6743\u91cd\u521d\u59cb\u5316\u662f\u5426\u5408\u7406\u3001\u662f\u5426\u51fa\u73b0\u68af\u5ea6\u6d88\u5931\u3002<\/li>\n<li>\u53ef\u89c6\u5316\u6a21\u578b\u7ed3\u6784&#xff0c;\u8f85\u52a9\u6587\u6863\u64b0\u5199\u548c\u8c03\u8bd5\u3002<\/li>\n<li>\u4f7f\u7528\u5d4c\u5165\u6295\u5f71\u4eea\u5206\u6790\u5206\u7c7b\u5668\u7684\u7279\u5f81\u8868\u793a\u8d28\u91cf\u3002<\/li>\n<li>\u5b9a\u4f4d\u6570\u636e\u52a0\u8f7d\u74f6\u9888&#xff0c;\u4f18\u5316\u8bad\u7ec3\u901f\u5ea6\u3002<\/li>\n<\/ul>\n<h4>1.5 \u884c\u4e1a\u5e94\u7528\u573a\u666f<\/h4>\n<ul>\n<li>\u5b9e\u9a8c\u7ba1\u7406&#xff1a;\u5bf9\u6bd4\u591a\u7ec4\u8d85\u53c2\u6570\u4e0b\u7684\u6a21\u578b\u6027\u80fd\u3002<\/li>\n<li>\u8c03\u8bd5&#xff1a;\u89c2\u5bdf\u68af\u5ea6\u6d88\u5931\/\u7206\u70b8&#xff0c;\u8c03\u6574\u5b66\u4e60\u7387\u6216\u521d\u59cb\u5316\u3002<\/li>\n<li>\u53ef\u89c6\u5316&#xff1a;\u5411\u975e\u6280\u672f\u4eba\u5458\u5c55\u793a\u6a21\u578b\u7ed3\u6784\u548c\u7279\u5f81\u63d0\u53d6\u6548\u679c\u3002<\/li>\n<li>\u6027\u80fd\u4f18\u5316&#xff1a;Profile\u5206\u6790\u627e\u51fa\u8bad\u7ec3\u4e2d\u7684\u74f6\u9888\u6b65\u9aa4\u3002<\/li>\n<\/ul>\n<h3>2. \u6838\u5fc3\u7406\u8bba\u7cbe\u8bb2<\/h3>\n<h4>2.1 TensorBoard \u67b6\u6784\u4e0e\u65e5\u5fd7<\/h4>\n<p>TensorBoard\u901a\u8fc7\u8bfb\u53d6TensorFlow\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u751f\u6210\u7684\u65e5\u5fd7\u6587\u4ef6&#xff08;events\u6587\u4ef6&#xff09;\u6765\u5c55\u793a\u56fe\u8868\u3002\u65e5\u5fd7\u4e2d\u8bb0\u5f55\u4e86\u6807\u91cf\u3001\u76f4\u65b9\u56fe\u3001\u56fe\u7ed3\u6784\u3001\u5d4c\u5165\u5411\u91cf\u7b49\u6570\u636e\u3002<\/p>\n<p>\u65e5\u5fd7\u76ee\u5f55\u7ed3\u6784&#xff1a;\u5efa\u8bae\u6bcf\u6b21\u8fd0\u884c\u4f7f\u7528\u4e00\u4e2a\u72ec\u7acb\u7684\u5b50\u76ee\u5f55&#xff0c;\u5982logs\/run1\u3001logs\/run2&#xff0c;\u4ee5\u4fbf\u5728TensorBoard\u4e2d\u5207\u6362\u5bf9\u6bd4\u3002<\/p>\n<p>TensorBoard\u670d\u52a1\u7684\u542f\u52a8&#xff1a;<\/p>\n<p>tensorboard <span class=\"token parameter variable\">&#8211;logdir<\/span> logs\/<\/p>\n<p>\u4e4b\u540e\u5728\u6d4f\u89c8\u5668\u4e2d\u6253\u5f00http:\/\/localhost:6006\u5373\u53ef\u3002<\/p>\n<p>\u5728Jupyter Notebook\u4e2d\u53ef\u76f4\u63a5\u5185\u5d4c&#xff1a;<\/p>\n<p><span class=\"token operator\">%<\/span>load_ext tensorboard<br \/>\n<span class=\"token operator\">%<\/span>tensorboard <span class=\"token operator\">&#8211;<\/span><span class=\"token operator\">&#8211;<\/span>logdir logs<span class=\"token operator\">\/<\/span><\/p>\n<h4>2.2 \u6807\u91cf\u4eea\u8868\u677f&#xff08;Scalar&#xff09;<\/h4>\n<p>\u8bb0\u5f55\u968f\u8fed\u4ee3\u6b21\u6570&#xff08;step&#xff09;\u53d8\u5316\u7684\u6570\u503c&#xff0c;\u5982\u635f\u5931\u3001\u51c6\u786e\u7387\u3001\u5b66\u4e60\u7387\u3002Keras\u7684TensorBoard\u56de\u8c03\u81ea\u52a8\u8bb0\u5f55\u6bcfepoch\u7684\u635f\u5931\u548c\u6307\u6807\u3002\u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d\u9700\u624b\u52a8\u8c03\u7528tf.summary.scalar()\u3002<\/p>\n<h4>2.3 \u8ba1\u7b97\u56fe\u53ef\u89c6\u5316<\/h4>\n<p>TensorBoard\u53ef\u4ee5\u5c55\u793a\u6a21\u578b\u7684\u8ba1\u7b97\u56fe&#xff08;Graph&#xff09;&#xff0c;\u5e2e\u52a9\u7406\u89e3\u6570\u636e\u6d41\u5411\u3001\u68c0\u67e5\u5404\u5c42\u8fde\u63a5\u662f\u5426\u6b63\u786e\u3002\u5bf9\u4e8eKeras\u6a21\u578b&#xff0c;\u53ea\u9700\u5728\u56de\u8c03\u4e2d\u8bbe\u7f6ehistogram_freq&#061;1&#xff08;\u6216\u4efb\u610f&gt;0&#xff09;\u5373\u53ef\u8bb0\u5f55\u56fe\u3002\u6ce8\u610f&#xff1a;\u56fe\u53ef\u89c6\u5316\u53ef\u80fd\u975e\u5e38\u590d\u6742&#xff0c;\u53ef\u4ee5\u4f7f\u7528logdir\u7684write_graph\u53c2\u6570\u63a7\u5236\u3002<\/p>\n<h4>2.4 \u76f4\u65b9\u56fe\u4eea\u8868\u677f&#xff08;Histogram&#xff09;<\/h4>\n<p>\u8bb0\u5f55\u5f20\u91cf\u503c\u968f\u65f6\u95f4\u7684\u5206\u5e03&#xff0c;\u4f8b\u5982\u6743\u91cd\u3001\u504f\u7f6e\u3001\u68af\u5ea6\u3001\u6fc0\u6d3b\u503c\u3002\u76f4\u65b9\u56fe\u53ef\u4ee5\u89c2\u5bdf&#xff1a;<\/p>\n<ul>\n<li>\u6743\u91cd\u662f\u5426\u8d8b\u4e8e\u96f6&#xff08;\u53ef\u80fd\u6b7b\u4ea1&#xff09;\u6216\u8fc7\u5927&#xff08;\u68af\u5ea6\u7206\u70b8&#xff09;\u3002<\/li>\n<li>\u6fc0\u6d3b\u503c\u662f\u5426\u9971\u548c&#xff08;Sigmoid\/Tanh&#xff09;\u3002<\/li>\n<li>\u68af\u5ea6\u5206\u5e03\u662f\u5426\u5747\u5300\u3002<\/li>\n<\/ul>\n<p>\u5728Keras\u4e2d\u8bbe\u7f6ehistogram_freq&#061;1\u4f1a\u8bb0\u5f55\u6bcf\u5c42\u6743\u91cd\u7684\u76f4\u65b9\u56fe&#xff1b;\u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d\u4f7f\u7528tf.summary.histogram()\u3002<\/p>\n<h4>2.5 \u5d4c\u5165\u6295\u5f71\u4eea&#xff08;Embedding Projector&#xff09;<\/h4>\n<p>\u7528\u4e8e\u53ef\u89c6\u5316\u9ad8\u7ef4\u5411\u91cf&#xff08;\u5982\u8bcd\u5d4c\u5165\u3001\u7279\u5f81\u5411\u91cf&#xff09;\u5728\u4f4e\u7ef4\u7a7a\u95f4&#xff08;PCA\u3001t-SNE&#xff09;\u4e2d\u7684\u5206\u5e03\u3002\u9700\u8981\u63d0\u4f9b\u4e00\u4e2a\u5143\u6570\u636e\u6587\u4ef6&#xff08;TSV&#xff09;\u5b9a\u4e49\u6807\u7b7e\u3002TensorBoard\u4f1a\u81ea\u52a8\u964d\u7ef4\u5e76\u652f\u6301\u4ea4\u4e92\u5f0f\u63a2\u7d22\u3002<\/p>\n<h4>2.6 \u8d85\u53c2\u6570\u8c03\u4f18&#xff08;HParams&#xff09;<\/h4>\n<p>\u7528\u4e8e\u8bb0\u5f55\u548c\u5bf9\u6bd4\u591a\u7ec4\u8d85\u53c2\u6570\u914d\u7f6e&#xff08;\u5b66\u4e60\u7387\u3001\u6279\u6b21\u5927\u5c0f\u3001\u7f51\u7edc\u5c42\u6570\u7b49&#xff09;\u3002\u9700\u8981\u5b9a\u4e49\u8d85\u53c2\u6570\u7a7a\u95f4&#xff0c;\u5e76\u5728\u6bcf\u6b21\u8fd0\u884c\u65f6\u8bb0\u5f55\u3002TensorBoard\u4f1a\u751f\u6210\u5e73\u884c\u5750\u6807\u56fe\u3001\u6563\u70b9\u56fe\u77e9\u9635\u7b49&#xff0c;\u5e2e\u52a9\u9009\u62e9\u6700\u4f73\u7ec4\u5408\u3002<\/p>\n<h4>2.7 \u6027\u80fd\u5206\u6790&#xff08;Profile&#xff09;<\/h4>\n<p>Profile\u5de5\u5177\u53ef\u4ee5\u6355\u83b7\u4e00\u4e2a\u8bad\u7ec3\u6b65\u9aa4\u4e2d\u5404\u64cd\u4f5c\u7684\u6267\u884c\u65f6\u95f4\u548c\u5185\u5b58\u5206\u914d\u3002\u901a\u8fc7\u65f6\u95f4\u8f74\u56fe\u53ef\u4ee5\u8bc6\u522bCPU\/GPU\u7684\u5229\u7528\u7387\u3001\u6570\u636e\u6d41\u6c34\u7ebf\u74f6\u9888&#xff08;\u5982\u8f93\u5165\u7ba1\u9053\u7a7a\u95f2&#xff09;\u3002Profile\u9700\u8981\u542f\u7528tf.profiler\u3002<\/p>\n<h3>3. \u73af\u5883\u642d\u5efa\u4e0e\u5de5\u5177\u914d\u7f6e<\/h3>\n<p>\u6cbf\u7528\u7b2c37\u8bfe\u73af\u5883&#xff0c;TensorBoard\u901a\u5e38\u968fTensorFlow\u4e00\u8d77\u5b89\u88c5\u3002\u82e5\u672a\u5b89\u88c5\u53ef\u624b\u52a8&#xff1a;<\/p>\n<p>pip <span class=\"token function\">install<\/span> tensorboard<\/p>\n<p>\u5bfc\u5165\u6a21\u5757&#xff1a;<\/p>\n<p><span class=\"token keyword\">import<\/span> tensorflow <span class=\"token keyword\">as<\/span> tf<br \/>\n<span class=\"token keyword\">import<\/span> numpy <span class=\"token keyword\">as<\/span> np<br \/>\n<span class=\"token keyword\">import<\/span> datetime<br \/>\n<span class=\"token keyword\">import<\/span> os<br \/>\n<span class=\"token keyword\">from<\/span> tensorflow<span class=\"token punctuation\">.<\/span>keras <span class=\"token keyword\">import<\/span> layers<span class=\"token punctuation\">,<\/span> models<span class=\"token punctuation\">,<\/span> datasets<span class=\"token punctuation\">,<\/span> callbacks<\/p>\n<h3>4. \u4ee3\u7801\u5b9e\u6218\u6559\u5b66<\/h3>\n<h4>4.1 \u57fa\u672c\u7528\u6cd5&#xff1a;Keras\u56de\u8c03<\/h4>\n<p><span class=\"token comment\"># \u52a0\u8f7dMNIST\u6570\u636e<\/span><br \/>\n<span class=\"token punctuation\">(<\/span>x_train<span class=\"token punctuation\">,<\/span> y_train<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token punctuation\">(<\/span>x_test<span class=\"token punctuation\">,<\/span> y_test<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">&#061;<\/span> datasets<span class=\"token punctuation\">.<\/span>mnist<span class=\"token punctuation\">.<\/span>load_data<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\nx_train <span class=\"token operator\">&#061;<\/span> x_train<span class=\"token punctuation\">.<\/span>reshape<span class=\"token punctuation\">(<\/span><span class=\"token operator\">&#8211;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">28<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">28<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>astype<span class=\"token punctuation\">(<\/span>np<span class=\"token punctuation\">.<\/span>float32<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">\/<\/span> <span class=\"token number\">255.0<\/span><br \/>\nx_test <span class=\"token operator\">&#061;<\/span> x_test<span class=\"token punctuation\">.<\/span>reshape<span class=\"token punctuation\">(<\/span><span class=\"token operator\">&#8211;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">28<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">28<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>astype<span class=\"token punctuation\">(<\/span>np<span class=\"token punctuation\">.<\/span>float32<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">\/<\/span> <span class=\"token number\">255.0<\/span><br \/>\ny_train <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>utils<span class=\"token punctuation\">.<\/span>to_categorical<span class=\"token punctuation\">(<\/span>y_train<span class=\"token punctuation\">,<\/span> <span class=\"token number\">10<\/span><span class=\"token punctuation\">)<\/span><br \/>\ny_test <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>utils<span class=\"token punctuation\">.<\/span>to_categorical<span class=\"token punctuation\">(<\/span>y_test<span class=\"token punctuation\">,<\/span> <span class=\"token number\">10<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token comment\"># \u7b80\u5355\u6a21\u578b<\/span><br \/>\nmodel <span class=\"token operator\">&#061;<\/span> models<span class=\"token punctuation\">.<\/span>Sequential<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">[<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Conv2D<span class=\"token punctuation\">(<\/span><span class=\"token number\">32<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">3<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;relu&#039;<\/span><span class=\"token punctuation\">,<\/span> input_shape<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">(<\/span><span class=\"token number\">28<\/span><span class=\"token punctuation\">,<\/span><span class=\"token number\">28<\/span><span class=\"token punctuation\">,<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>MaxPooling2D<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Conv2D<span class=\"token punctuation\">(<\/span><span class=\"token number\">64<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">3<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;relu&#039;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>MaxPooling2D<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Flatten<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Dense<span class=\"token punctuation\">(<\/span><span class=\"token number\">64<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;relu&#039;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Dense<span class=\"token punctuation\">(<\/span><span class=\"token number\">10<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;softmax&#039;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token comment\"># \u65e5\u5fd7\u76ee\u5f55<\/span><br \/>\nlog_dir <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#034;logs\/mnist\/&#034;<\/span> <span class=\"token operator\">&#043;<\/span> datetime<span class=\"token punctuation\">.<\/span>datetime<span class=\"token punctuation\">.<\/span>now<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>strftime<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;%Y%m%d-%H%M%S&#034;<\/span><span class=\"token punctuation\">)<\/span><br \/>\ntensorboard_callback <span class=\"token operator\">&#061;<\/span> callbacks<span class=\"token punctuation\">.<\/span>TensorBoard<span class=\"token punctuation\">(<\/span><br \/>\n    log_dir<span class=\"token operator\">&#061;<\/span>log_dir<span class=\"token punctuation\">,<\/span><br \/>\n    histogram_freq<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">,<\/span>      <span class=\"token comment\"># \u6bcf\u4e2aepoch\u8bb0\u5f55\u76f4\u65b9\u56fe<\/span><br \/>\n    write_graph<span class=\"token operator\">&#061;<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">,<\/span>      <span class=\"token comment\"># \u8bb0\u5f55\u8ba1\u7b97\u56fe<\/span><br \/>\n    write_images<span class=\"token operator\">&#061;<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">,<\/span>     <span class=\"token comment\"># \u8bb0\u5f55\u6a21\u578b\u6743\u91cd\u56fe\u7247<\/span><br \/>\n    update_freq<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;epoch&#039;<\/span><span class=\"token punctuation\">,<\/span>   <span class=\"token comment\"># \u6bcf\u4e2aepoch\u66f4\u65b0<\/span><br \/>\n    profile_batch<span class=\"token operator\">&#061;<\/span><span class=\"token number\">0<\/span>        <span class=\"token comment\"># \u4e0d\u542f\u7528profile&#xff08;\u540e\u9762\u5355\u72ec\u6f14\u793a&#xff09;<\/span><br \/>\n<span class=\"token punctuation\">)<\/span><\/p>\n<p>model<span class=\"token punctuation\">.<\/span><span class=\"token builtin\">compile<\/span><span class=\"token punctuation\">(<\/span>optimizer<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;adam&#039;<\/span><span class=\"token punctuation\">,<\/span> loss<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;categorical_crossentropy&#039;<\/span><span class=\"token punctuation\">,<\/span> metrics<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">[<\/span><span class=\"token string\">&#039;accuracy&#039;<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><br \/>\nmodel<span class=\"token punctuation\">.<\/span>fit<span class=\"token punctuation\">(<\/span>x_train<span class=\"token punctuation\">,<\/span> y_train<span class=\"token punctuation\">,<\/span> epochs<span class=\"token operator\">&#061;<\/span><span class=\"token number\">5<\/span><span class=\"token punctuation\">,<\/span> batch_size<span class=\"token operator\">&#061;<\/span><span class=\"token number\">128<\/span><span class=\"token punctuation\">,<\/span> validation_split<span class=\"token operator\">&#061;<\/span><span class=\"token number\">0.1<\/span><span class=\"token punctuation\">,<\/span><br \/>\n          callbacks<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">[<\/span>tensorboard_callback<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">,<\/span> verbose<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p>\u542f\u52a8TensorBoard&#xff1a;<\/p>\n<p>tensorboard <span class=\"token parameter variable\">&#8211;logdir<\/span> logs\/<\/p>\n<p>\u5728\u6d4f\u89c8\u5668\u4e2d\u67e5\u770bScalars\u3001Graph\u3001Histograms\u7b49\u3002<\/p>\n<h4>4.2 \u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d\u624b\u52a8\u5199\u5165\u65e5\u5fd7<\/h4>\n<p><span class=\"token comment\"># \u51c6\u5907\u6570\u636e<\/span><br \/>\ntrain_ds <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>data<span class=\"token punctuation\">.<\/span>Dataset<span class=\"token punctuation\">.<\/span>from_tensor_slices<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">(<\/span>x_train<span class=\"token punctuation\">,<\/span> y_train<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>batch<span class=\"token punctuation\">(<\/span><span class=\"token number\">128<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>shuffle<span class=\"token punctuation\">(<\/span><span class=\"token number\">10000<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>prefetch<span class=\"token punctuation\">(<\/span>tf<span class=\"token punctuation\">.<\/span>data<span class=\"token punctuation\">.<\/span>AUTOTUNE<span class=\"token punctuation\">)<\/span><\/p>\n<p>model <span class=\"token operator\">&#061;<\/span> models<span class=\"token punctuation\">.<\/span>Sequential<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">[<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Conv2D<span class=\"token punctuation\">(<\/span><span class=\"token number\">32<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">3<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;relu&#039;<\/span><span class=\"token punctuation\">,<\/span> input_shape<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">(<\/span><span class=\"token number\">28<\/span><span class=\"token punctuation\">,<\/span><span class=\"token number\">28<\/span><span class=\"token punctuation\">,<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>MaxPooling2D<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Flatten<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Dense<span class=\"token punctuation\">(<\/span><span class=\"token number\">10<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;softmax&#039;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><br \/>\noptimizer <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>optimizers<span class=\"token punctuation\">.<\/span>Adam<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\nloss_fn <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>losses<span class=\"token punctuation\">.<\/span>CategoricalCrossentropy<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token comment\"># \u65e5\u5fd7\u76ee\u5f55<\/span><br \/>\ncurrent_time <span class=\"token operator\">&#061;<\/span> datetime<span class=\"token punctuation\">.<\/span>datetime<span class=\"token punctuation\">.<\/span>now<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>strftime<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;%Y%m%d-%H%M%S&#034;<\/span><span class=\"token punctuation\">)<\/span><br \/>\ntrain_log_dir <span class=\"token operator\">&#061;<\/span> <span class=\"token string-interpolation\"><span class=\"token string\">f&#039;logs\/custom\/<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>current_time<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">\/train&#039;<\/span><\/span><br \/>\nval_log_dir <span class=\"token operator\">&#061;<\/span> <span class=\"token string-interpolation\"><span class=\"token string\">f&#039;logs\/custom\/<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>current_time<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">\/val&#039;<\/span><\/span><br \/>\ntrain_summary_writer <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>summary<span class=\"token punctuation\">.<\/span>create_file_writer<span class=\"token punctuation\">(<\/span>train_log_dir<span class=\"token punctuation\">)<\/span><br \/>\nval_summary_writer <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>summary<span class=\"token punctuation\">.<\/span>create_file_writer<span class=\"token punctuation\">(<\/span>val_log_dir<span class=\"token punctuation\">)<\/span><\/p>\n<p>epochs <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">3<\/span><br \/>\nglobal_step <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><br \/>\n<span class=\"token keyword\">for<\/span> epoch <span class=\"token keyword\">in<\/span> <span class=\"token builtin\">range<\/span><span class=\"token punctuation\">(<\/span>epochs<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    <span class=\"token comment\"># \u8bad\u7ec3<\/span><br \/>\n    epoch_loss <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0.0<\/span><br \/>\n    epoch_acc <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0.0<\/span><br \/>\n    n_batches <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><br \/>\n    <span class=\"token keyword\">for<\/span> x_batch<span class=\"token punctuation\">,<\/span> y_batch <span class=\"token keyword\">in<\/span> train_ds<span class=\"token punctuation\">:<\/span><br \/>\n        <span class=\"token keyword\">with<\/span> tf<span class=\"token punctuation\">.<\/span>GradientTape<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token keyword\">as<\/span> tape<span class=\"token punctuation\">:<\/span><br \/>\n            pred <span class=\"token operator\">&#061;<\/span> model<span class=\"token punctuation\">(<\/span>x_batch<span class=\"token punctuation\">,<\/span> training<span class=\"token operator\">&#061;<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">)<\/span><br \/>\n            loss <span class=\"token operator\">&#061;<\/span> loss_fn<span class=\"token punctuation\">(<\/span>y_batch<span class=\"token punctuation\">,<\/span> pred<span class=\"token punctuation\">)<\/span><br \/>\n        grads <span class=\"token operator\">&#061;<\/span> tape<span class=\"token punctuation\">.<\/span>gradient<span class=\"token punctuation\">(<\/span>loss<span class=\"token punctuation\">,<\/span> model<span class=\"token punctuation\">.<\/span>trainable_variables<span class=\"token punctuation\">)<\/span><br \/>\n        optimizer<span class=\"token punctuation\">.<\/span>apply_gradients<span class=\"token punctuation\">(<\/span><span class=\"token builtin\">zip<\/span><span class=\"token punctuation\">(<\/span>grads<span class=\"token punctuation\">,<\/span> model<span class=\"token punctuation\">.<\/span>trainable_variables<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n        epoch_loss <span class=\"token operator\">&#043;&#061;<\/span> loss<span class=\"token punctuation\">.<\/span>numpy<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n        epoch_acc <span class=\"token operator\">&#043;&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>reduce_mean<span class=\"token punctuation\">(<\/span>tf<span class=\"token punctuation\">.<\/span>cast<span class=\"token punctuation\">(<\/span>tf<span class=\"token punctuation\">.<\/span>equal<span class=\"token punctuation\">(<\/span>tf<span class=\"token punctuation\">.<\/span>argmax<span class=\"token punctuation\">(<\/span>pred<span class=\"token punctuation\">,<\/span> axis<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> tf<span class=\"token punctuation\">.<\/span>argmax<span class=\"token punctuation\">(<\/span>y_batch<span class=\"token punctuation\">,<\/span> axis<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> tf<span class=\"token punctuation\">.<\/span>float32<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>numpy<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n        n_batches <span class=\"token operator\">&#043;&#061;<\/span> <span class=\"token number\">1<\/span><br \/>\n        global_step <span class=\"token operator\">&#043;&#061;<\/span> <span class=\"token number\">1<\/span><br \/>\n    train_loss <span class=\"token operator\">&#061;<\/span> epoch_loss <span class=\"token operator\">\/<\/span> n_batches<br \/>\n    train_acc <span class=\"token operator\">&#061;<\/span> epoch_acc <span class=\"token operator\">\/<\/span> n_batches<br \/>\n    <span class=\"token comment\"># \u5199\u5165\u8bad\u7ec3\u6307\u6807<\/span><br \/>\n    <span class=\"token keyword\">with<\/span> train_summary_writer<span class=\"token punctuation\">.<\/span>as_default<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><br \/>\n        tf<span class=\"token punctuation\">.<\/span>summary<span class=\"token punctuation\">.<\/span>scalar<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;loss&#039;<\/span><span class=\"token punctuation\">,<\/span> train_loss<span class=\"token punctuation\">,<\/span> step<span class=\"token operator\">&#061;<\/span>epoch<span class=\"token punctuation\">)<\/span><br \/>\n        tf<span class=\"token punctuation\">.<\/span>summary<span class=\"token punctuation\">.<\/span>scalar<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;accuracy&#039;<\/span><span class=\"token punctuation\">,<\/span> train_acc<span class=\"token punctuation\">,<\/span> step<span class=\"token operator\">&#061;<\/span>epoch<span class=\"token punctuation\">)<\/span><br \/>\n        <span class=\"token comment\"># \u8bb0\u5f55\u7b2c\u4e00\u5c42\u5377\u79ef\u6838\u76f4\u65b9\u56fe<\/span><br \/>\n        conv1_weights <span class=\"token operator\">&#061;<\/span> model<span class=\"token punctuation\">.<\/span>layers<span class=\"token punctuation\">[<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">.<\/span>get_weights<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">[<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">]<\/span><br \/>\n        tf<span class=\"token punctuation\">.<\/span>summary<span class=\"token punctuation\">.<\/span>histogram<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;conv1\/weights&#039;<\/span><span class=\"token punctuation\">,<\/span> conv1_weights<span class=\"token punctuation\">,<\/span> step<span class=\"token operator\">&#061;<\/span>epoch<span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token comment\"># \u9a8c\u8bc1&#xff08;\u7b80\u5355\u793a\u4f8b&#xff0c;\u53d6\u4e00\u4e2abatch&#xff09;<\/span><br \/>\n    x_val<span class=\"token punctuation\">,<\/span> y_val <span class=\"token operator\">&#061;<\/span> <span class=\"token builtin\">next<\/span><span class=\"token punctuation\">(<\/span><span class=\"token builtin\">iter<\/span><span class=\"token punctuation\">(<\/span>val_ds<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span>  <span class=\"token comment\"># \u5b9e\u9645\u5e94\u5355\u72ec\u6784\u5efa\u9a8c\u8bc1\u96c6<\/span><br \/>\n    val_pred <span class=\"token operator\">&#061;<\/span> model<span class=\"token punctuation\">(<\/span>x_val<span class=\"token punctuation\">,<\/span> training<span class=\"token operator\">&#061;<\/span><span class=\"token boolean\">False<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    val_loss <span class=\"token operator\">&#061;<\/span> loss_fn<span class=\"token punctuation\">(<\/span>y_val<span class=\"token punctuation\">,<\/span> val_pred<span class=\"token punctuation\">)<\/span><br \/>\n    val_acc <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>reduce_mean<span class=\"token punctuation\">(<\/span>tf<span class=\"token punctuation\">.<\/span>cast<span class=\"token punctuation\">(<\/span>tf<span class=\"token punctuation\">.<\/span>equal<span class=\"token punctuation\">(<\/span>tf<span class=\"token punctuation\">.<\/span>argmax<span class=\"token punctuation\">(<\/span>val_pred<span class=\"token punctuation\">,<\/span> axis<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> tf<span class=\"token punctuation\">.<\/span>argmax<span class=\"token punctuation\">(<\/span>y_val<span class=\"token punctuation\">,<\/span> axis<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> tf<span class=\"token punctuation\">.<\/span>float32<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token keyword\">with<\/span> val_summary_writer<span class=\"token punctuation\">.<\/span>as_default<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><br \/>\n        tf<span class=\"token punctuation\">.<\/span>summary<span class=\"token punctuation\">.<\/span>scalar<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;loss&#039;<\/span><span class=\"token punctuation\">,<\/span> val_loss<span class=\"token punctuation\">,<\/span> step<span class=\"token operator\">&#061;<\/span>epoch<span class=\"token punctuation\">)<\/span><br \/>\n        tf<span class=\"token punctuation\">.<\/span>summary<span class=\"token punctuation\">.<\/span>scalar<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;accuracy&#039;<\/span><span class=\"token punctuation\">,<\/span> val_acc<span class=\"token punctuation\">,<\/span> step<span class=\"token operator\">&#061;<\/span>epoch<span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token keyword\">print<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string-interpolation\"><span class=\"token string\">f&#034;Epoch <\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>epoch<span class=\"token operator\">&#043;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">: train_loss&#061;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>train_loss<span class=\"token punctuation\">:<\/span><span class=\"token format-spec\">.4f<\/span><span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">, val_acc&#061;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>val_acc<span class=\"token punctuation\">.<\/span>numpy<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><span class=\"token format-spec\">.4f<\/span><span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">&#034;<\/span><\/span><span class=\"token punctuation\">)<\/span><\/p>\n<h4>4.3 \u53ef\u89c6\u5316\u8bcd\u5411\u91cf&#xff08;\u5d4c\u5165\u6295\u5f71\u4eea&#xff09;<\/h4>\n<p><span class=\"token comment\"># \u4f7f\u7528\u7b2c30\u8bfe\u7684\u8bcd\u5411\u91cf\u793a\u4f8b<\/span><br \/>\n<span class=\"token keyword\">import<\/span> tensorflow <span class=\"token keyword\">as<\/span> tf<br \/>\n<span class=\"token keyword\">from<\/span> tensorflow<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>preprocessing<span class=\"token punctuation\">.<\/span>text <span class=\"token keyword\">import<\/span> Tokenizer<br \/>\n<span class=\"token keyword\">from<\/span> tensorflow<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>preprocessing<span class=\"token punctuation\">.<\/span>sequence <span class=\"token keyword\">import<\/span> pad_sequences<\/p>\n<p><span class=\"token comment\"># \u7b80\u5355\u8bed\u6599<\/span><br \/>\nsentences <span class=\"token operator\">&#061;<\/span> <span class=\"token punctuation\">[<\/span><span class=\"token string\">&#034;cat sat on mat&#034;<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token string\">&#034;dog sat on log&#034;<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token string\">&#034;cat and dog&#034;<\/span><span class=\"token punctuation\">]<\/span><br \/>\ntokenizer <span class=\"token operator\">&#061;<\/span> Tokenizer<span class=\"token punctuation\">(<\/span>num_words<span class=\"token operator\">&#061;<\/span><span class=\"token number\">10<\/span><span class=\"token punctuation\">)<\/span><br \/>\ntokenizer<span class=\"token punctuation\">.<\/span>fit_on_texts<span class=\"token punctuation\">(<\/span>sentences<span class=\"token punctuation\">)<\/span><br \/>\nsequences <span class=\"token operator\">&#061;<\/span> tokenizer<span class=\"token punctuation\">.<\/span>texts_to_sequences<span class=\"token punctuation\">(<\/span>sentences<span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token comment\"># \u8bad\u7ec3\u4e00\u4e2a\u7b80\u5355\u7684\u8bcd\u5d4c\u5165\u6a21\u578b<\/span><br \/>\nvocab_size <span class=\"token operator\">&#061;<\/span> <span class=\"token builtin\">len<\/span><span class=\"token punctuation\">(<\/span>tokenizer<span class=\"token punctuation\">.<\/span>word_index<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">&#043;<\/span> <span class=\"token number\">1<\/span><br \/>\nembedding_dim <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">8<\/span><br \/>\nmodel_emb <span class=\"token operator\">&#061;<\/span> models<span class=\"token punctuation\">.<\/span>Sequential<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">[<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Embedding<span class=\"token punctuation\">(<\/span>vocab_size<span class=\"token punctuation\">,<\/span> embedding_dim<span class=\"token punctuation\">,<\/span> input_length<span class=\"token operator\">&#061;<\/span><span class=\"token number\">4<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>GlobalAveragePooling1D<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    layers<span class=\"token punctuation\">.<\/span>Dense<span class=\"token punctuation\">(<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;sigmoid&#039;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><br \/>\nmodel_emb<span class=\"token punctuation\">.<\/span><span class=\"token builtin\">compile<\/span><span class=\"token punctuation\">(<\/span>optimizer<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;adam&#039;<\/span><span class=\"token punctuation\">,<\/span> loss<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;binary_crossentropy&#039;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token comment\"># \u968f\u673a\u6807\u7b7e<\/span><br \/>\nlabels <span class=\"token operator\">&#061;<\/span> np<span class=\"token punctuation\">.<\/span>random<span class=\"token punctuation\">.<\/span>randint<span class=\"token punctuation\">(<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">2<\/span><span class=\"token punctuation\">,<\/span> size<span class=\"token operator\">&#061;<\/span><span class=\"token builtin\">len<\/span><span class=\"token punctuation\">(<\/span>sentences<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\nmodel_emb<span class=\"token punctuation\">.<\/span>fit<span class=\"token punctuation\">(<\/span>pad_sequences<span class=\"token punctuation\">(<\/span>sequences<span class=\"token punctuation\">,<\/span> maxlen<span class=\"token operator\">&#061;<\/span><span class=\"token number\">4<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> labels<span class=\"token punctuation\">,<\/span> epochs<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">,<\/span> verbose<span class=\"token operator\">&#061;<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token comment\"># \u63d0\u53d6\u8bcd\u5d4c\u5165\u77e9\u9635<\/span><br \/>\nembedding_matrix <span class=\"token operator\">&#061;<\/span> model_emb<span class=\"token punctuation\">.<\/span>layers<span class=\"token punctuation\">[<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">.<\/span>get_weights<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">[<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">]<\/span>  <span class=\"token comment\"># (vocab_size, 8)<\/span><\/p>\n<p><span class=\"token comment\"># \u521b\u5efa\u65e5\u5fd7\u76ee\u5f55\u548c\u5143\u6570\u636e\u6587\u4ef6<\/span><br \/>\nlog_dir <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#034;logs\/embeddings\/&#034;<\/span><br \/>\nos<span class=\"token punctuation\">.<\/span>makedirs<span class=\"token punctuation\">(<\/span>log_dir<span class=\"token punctuation\">,<\/span> exist_ok<span class=\"token operator\">&#061;<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token comment\"># \u4fdd\u5b58\u5143\u6570\u636e&#xff08;\u8bcd\u4e0e\u6807\u7b7e&#xff09;<\/span><br \/>\n<span class=\"token keyword\">with<\/span> <span class=\"token builtin\">open<\/span><span class=\"token punctuation\">(<\/span>os<span class=\"token punctuation\">.<\/span>path<span class=\"token punctuation\">.<\/span>join<span class=\"token punctuation\">(<\/span>log_dir<span class=\"token punctuation\">,<\/span> <span class=\"token string\">&#039;metadata.tsv&#039;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token string\">&#039;w&#039;<\/span><span class=\"token punctuation\">,<\/span> encoding<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;utf-8&#039;<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token keyword\">as<\/span> f<span class=\"token punctuation\">:<\/span><br \/>\n    f<span class=\"token punctuation\">.<\/span>write<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;Word\\\\tIndex\\\\n&#034;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token keyword\">for<\/span> word<span class=\"token punctuation\">,<\/span> idx <span class=\"token keyword\">in<\/span> tokenizer<span class=\"token punctuation\">.<\/span>word_index<span class=\"token punctuation\">.<\/span>items<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><br \/>\n        f<span class=\"token punctuation\">.<\/span>write<span class=\"token punctuation\">(<\/span><span class=\"token string-interpolation\"><span class=\"token string\">f&#034;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>word<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">\\\\t<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>idx<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">\\\\n&#034;<\/span><\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token comment\"># \u4fdd\u5b58\u5d4c\u5165\u5411\u91cf&#xff08;TSV\u683c\u5f0f&#xff09;<\/span><br \/>\nnp<span class=\"token punctuation\">.<\/span>savetxt<span class=\"token punctuation\">(<\/span>os<span class=\"token punctuation\">.<\/span>path<span class=\"token punctuation\">.<\/span>join<span class=\"token punctuation\">(<\/span>log_dir<span class=\"token punctuation\">,<\/span> <span class=\"token string\">&#039;vectors.tsv&#039;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> embedding_matrix<span class=\"token punctuation\">,<\/span> delimiter<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;\\\\t&#039;<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token comment\"># \u914d\u7f6e\u5d4c\u5165\u6295\u5f71\u4eea&#xff08;\u9700\u5728TensorBoard\u4e2d\u52a0\u8f7d&#xff09;<\/span><br \/>\n<span class=\"token comment\"># \u542f\u52a8TensorBoard\u540e&#xff0c;\u8fdb\u5165&#034;Projector&#034;\u9875\u9762&#xff0c;\u70b9\u51fb&#034;Load&#034;&#xff0c;\u9009\u62e9vectors.tsv\u548cmetadata.tsv\u5373\u53ef<\/span><\/p>\n<h4>4.4 \u8d85\u53c2\u6570\u8c03\u4f18&#xff08;HParams&#xff09;<\/h4>\n<p><span class=\"token keyword\">from<\/span> tensorboard<span class=\"token punctuation\">.<\/span>plugins<span class=\"token punctuation\">.<\/span>hparams <span class=\"token keyword\">import<\/span> api <span class=\"token keyword\">as<\/span> hp<\/p>\n<p><span class=\"token comment\"># \u5b9a\u4e49\u8d85\u53c2\u6570\u7a7a\u95f4<\/span><br \/>\nHP_LEARNING_RATE <span class=\"token operator\">&#061;<\/span> hp<span class=\"token punctuation\">.<\/span>HParam<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;learning_rate&#039;<\/span><span class=\"token punctuation\">,<\/span> hp<span class=\"token punctuation\">.<\/span>RealInterval<span class=\"token punctuation\">(<\/span><span class=\"token number\">1e-4<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">1e-2<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\nHP_NUM_UNITS <span class=\"token operator\">&#061;<\/span> hp<span class=\"token punctuation\">.<\/span>HParam<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;num_units&#039;<\/span><span class=\"token punctuation\">,<\/span> hp<span class=\"token punctuation\">.<\/span>IntInterval<span class=\"token punctuation\">(<\/span><span class=\"token number\">32<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">128<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\nHP_DROPOUT <span class=\"token operator\">&#061;<\/span> hp<span class=\"token punctuation\">.<\/span>HParam<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;dropout&#039;<\/span><span class=\"token punctuation\">,<\/span> hp<span class=\"token punctuation\">.<\/span>RealInterval<span class=\"token punctuation\">(<\/span><span class=\"token number\">0.1<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">0.5<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token comment\"># \u65e5\u5fd7\u76ee\u5f55<\/span><br \/>\nhparams_log_dir <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#039;logs\/hparams\/&#039;<\/span><\/p>\n<p><span class=\"token keyword\">def<\/span> <span class=\"token function\">train_with_hparams<\/span><span class=\"token punctuation\">(<\/span>hparams<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    model <span class=\"token operator\">&#061;<\/span> models<span class=\"token punctuation\">.<\/span>Sequential<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">[<\/span><br \/>\n        layers<span class=\"token punctuation\">.<\/span>Dense<span class=\"token punctuation\">(<\/span>hparams<span class=\"token punctuation\">[<\/span>HP_NUM_UNITS<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;relu&#039;<\/span><span class=\"token punctuation\">,<\/span> input_shape<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">(<\/span><span class=\"token number\">784<\/span><span class=\"token punctuation\">,<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n        layers<span class=\"token punctuation\">.<\/span>Dropout<span class=\"token punctuation\">(<\/span>hparams<span class=\"token punctuation\">[<\/span>HP_DROPOUT<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n        layers<span class=\"token punctuation\">.<\/span>Dense<span class=\"token punctuation\">(<\/span><span class=\"token number\">10<\/span><span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;softmax&#039;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    model<span class=\"token punctuation\">.<\/span><span class=\"token builtin\">compile<\/span><span class=\"token punctuation\">(<\/span>optimizer<span class=\"token operator\">&#061;<\/span>tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>optimizers<span class=\"token punctuation\">.<\/span>Adam<span class=\"token punctuation\">(<\/span>hparams<span class=\"token punctuation\">[<\/span>HP_LEARNING_RATE<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                  loss<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;sparse_categorical_crossentropy&#039;<\/span><span class=\"token punctuation\">,<\/span> metrics<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">[<\/span><span class=\"token string\">&#039;accuracy&#039;<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token comment\"># \u4f7f\u7528HParams\u56de\u8c03<\/span><br \/>\n    callback <span class=\"token operator\">&#061;<\/span> hp<span class=\"token punctuation\">.<\/span>KerasCallback<span class=\"token punctuation\">(<\/span>hparams_log_dir<span class=\"token punctuation\">,<\/span> hparams<span class=\"token punctuation\">)<\/span><br \/>\n    model<span class=\"token punctuation\">.<\/span>fit<span class=\"token punctuation\">(<\/span>x_train<span class=\"token punctuation\">.<\/span>reshape<span class=\"token punctuation\">(<\/span><span class=\"token operator\">&#8211;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">,<\/span><span class=\"token number\">784<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> y_train<span class=\"token punctuation\">,<\/span> epochs<span class=\"token operator\">&#061;<\/span><span class=\"token number\">3<\/span><span class=\"token punctuation\">,<\/span> batch_size<span class=\"token operator\">&#061;<\/span><span class=\"token number\">128<\/span><span class=\"token punctuation\">,<\/span> callbacks<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">[<\/span>callback<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">,<\/span> verbose<span class=\"token operator\">&#061;<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    _<span class=\"token punctuation\">,<\/span> accuracy <span class=\"token operator\">&#061;<\/span> model<span class=\"token punctuation\">.<\/span>evaluate<span class=\"token punctuation\">(<\/span>x_test<span class=\"token punctuation\">.<\/span>reshape<span class=\"token punctuation\">(<\/span><span class=\"token operator\">&#8211;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">,<\/span><span class=\"token number\">784<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> y_test<span class=\"token punctuation\">,<\/span> verbose<span class=\"token operator\">&#061;<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token keyword\">return<\/span> accuracy<\/p>\n<p><span class=\"token comment\"># \u7f51\u683c\u641c\u7d22<\/span><br \/>\nsession_num <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">0<\/span><br \/>\n<span class=\"token keyword\">for<\/span> lr <span class=\"token keyword\">in<\/span> <span class=\"token punctuation\">[<\/span><span class=\"token number\">0.001<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">0.0005<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    <span class=\"token keyword\">for<\/span> units <span class=\"token keyword\">in<\/span> <span class=\"token punctuation\">[<\/span><span class=\"token number\">64<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">128<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">:<\/span><br \/>\n        <span class=\"token keyword\">for<\/span> dropout <span class=\"token keyword\">in<\/span> <span class=\"token punctuation\">[<\/span><span class=\"token number\">0.2<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">0.4<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">:<\/span><br \/>\n            hparams <span class=\"token operator\">&#061;<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n                HP_LEARNING_RATE<span class=\"token punctuation\">:<\/span> lr<span class=\"token punctuation\">,<\/span><br \/>\n                HP_NUM_UNITS<span class=\"token punctuation\">:<\/span> units<span class=\"token punctuation\">,<\/span><br \/>\n                HP_DROPOUT<span class=\"token punctuation\">:<\/span> dropout<span class=\"token punctuation\">,<\/span><br \/>\n            <span class=\"token punctuation\">}<\/span><br \/>\n            accuracy <span class=\"token operator\">&#061;<\/span> train_with_hparams<span class=\"token punctuation\">(<\/span>hparams<span class=\"token punctuation\">)<\/span><br \/>\n            <span class=\"token keyword\">print<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string-interpolation\"><span class=\"token string\">f&#034;Session <\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>session_num<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">: lr&#061;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>lr<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">, units&#061;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>units<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">, dropout&#061;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>dropout<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">, accuracy&#061;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>accuracy<span class=\"token punctuation\">:<\/span><span class=\"token format-spec\">.4f<\/span><span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">&#034;<\/span><\/span><span class=\"token punctuation\">)<\/span><br \/>\n            session_num <span class=\"token operator\">&#043;&#061;<\/span> <span class=\"token number\">1<\/span><\/p>\n<p><span class=\"token comment\"># \u542f\u52a8TensorBoard &#8211;logdir logs\/hparams \u540e&#xff0c;\u70b9\u51fbHParams\u6807\u7b7e\u67e5\u770b\u5bf9\u6bd4\u56fe<\/span><\/p>\n<h4>4.5 \u6027\u80fd\u5206\u6790&#xff08;Profile&#xff09;<\/h4>\n<p><span class=\"token comment\"># \u4f7f\u7528Keras\u56de\u8c03\u7684\u65b9\u5f0f\u542f\u7528profile<\/span><br \/>\ntensorboard_callback <span class=\"token operator\">&#061;<\/span> callbacks<span class=\"token punctuation\">.<\/span>TensorBoard<span class=\"token punctuation\">(<\/span><br \/>\n    log_dir<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;logs\/profile&#039;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    profile_batch<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;10,20&#039;<\/span>  <span class=\"token comment\"># \u5bf9\u7b2c10-20\u4e2abatch\u8fdb\u884cprofile<\/span><br \/>\n<span class=\"token punctuation\">)<\/span><br \/>\nmodel<span class=\"token punctuation\">.<\/span>fit<span class=\"token punctuation\">(<\/span>x_train<span class=\"token punctuation\">,<\/span> y_train<span class=\"token punctuation\">,<\/span> epochs<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">,<\/span> batch_size<span class=\"token operator\">&#061;<\/span><span class=\"token number\">128<\/span><span class=\"token punctuation\">,<\/span> callbacks<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">[<\/span>tensorboard_callback<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">,<\/span> verbose<span class=\"token operator\">&#061;<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token comment\"># \u5728TensorBoard\u4e2d\u67e5\u770bProfile\u6807\u7b7e\u9875<\/span><\/p>\n<p><span class=\"token comment\"># \u6216\u5728\u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d\u4f7f\u7528tf.profiler<\/span><br \/>\n<span class=\"token keyword\">import<\/span> tensorflow <span class=\"token keyword\">as<\/span> tf<br \/>\ntf<span class=\"token punctuation\">.<\/span>profiler<span class=\"token punctuation\">.<\/span>experimental<span class=\"token punctuation\">.<\/span>start<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;logs\/profile&#039;<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token comment\"># \u8bad\u7ec3\u82e5\u5e72step<\/span><br \/>\n<span class=\"token keyword\">for<\/span> step<span class=\"token punctuation\">,<\/span> <span class=\"token punctuation\">(<\/span>x_batch<span class=\"token punctuation\">,<\/span> y_batch<span class=\"token punctuation\">)<\/span> <span class=\"token keyword\">in<\/span> <span class=\"token builtin\">enumerate<\/span><span class=\"token punctuation\">(<\/span>train_ds<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    <span class=\"token keyword\">if<\/span> step <span class=\"token operator\">&gt;&#061;<\/span> <span class=\"token number\">10<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token keyword\">break<\/span><br \/>\n    <span class=\"token comment\"># \u8bad\u7ec3\u4ee3\u7801&#8230;<\/span><br \/>\ntf<span class=\"token punctuation\">.<\/span>profiler<span class=\"token punctuation\">.<\/span>experimental<span class=\"token punctuation\">.<\/span>stop<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<h3>5. \u6848\u4f8b\u5b9e\u64cd\u6f14\u7ec3<\/h3>\n<p>\u6848\u4f8b&#xff1a;\u4f7f\u7528TensorBoard\u76d1\u63a7ResNet\u5728CIFAR-10\u4e0a\u7684\u8bad\u7ec3\u8fc7\u7a0b&#xff0c;\u5e76\u901a\u8fc7HParams\u5bf9\u6bd4\u4e0d\u540c\u521d\u59cb\u5b66\u4e60\u7387<\/p>\n<h4>5.1 \u6570\u636e\u52a0\u8f7d\u4e0e\u6a21\u578b\u6784\u5efa<\/h4>\n<p><span class=\"token punctuation\">(<\/span>x_train<span class=\"token punctuation\">,<\/span> y_train<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token punctuation\">(<\/span>x_test<span class=\"token punctuation\">,<\/span> y_test<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">&#061;<\/span> datasets<span class=\"token punctuation\">.<\/span>cifar10<span class=\"token punctuation\">.<\/span>load_data<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\nx_train <span class=\"token operator\">&#061;<\/span> x_train<span class=\"token punctuation\">.<\/span>astype<span class=\"token punctuation\">(<\/span>np<span class=\"token punctuation\">.<\/span>float32<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">\/<\/span> <span class=\"token number\">255.0<\/span><br \/>\nx_test <span class=\"token operator\">&#061;<\/span> x_test<span class=\"token punctuation\">.<\/span>astype<span class=\"token punctuation\">(<\/span>np<span class=\"token punctuation\">.<\/span>float32<span class=\"token punctuation\">)<\/span> <span class=\"token operator\">\/<\/span> <span class=\"token number\">255.0<\/span><br \/>\ny_train <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>utils<span class=\"token punctuation\">.<\/span>to_categorical<span class=\"token punctuation\">(<\/span>y_train<span class=\"token punctuation\">,<\/span> <span class=\"token number\">10<\/span><span class=\"token punctuation\">)<\/span><br \/>\ny_test <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>utils<span class=\"token punctuation\">.<\/span>to_categorical<span class=\"token punctuation\">(<\/span>y_test<span class=\"token punctuation\">,<\/span> <span class=\"token number\">10<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p><span class=\"token keyword\">def<\/span> <span class=\"token function\">build_resnet20<\/span><span class=\"token punctuation\">(<\/span>input_shape<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">(<\/span><span class=\"token number\">32<\/span><span class=\"token punctuation\">,<\/span><span class=\"token number\">32<\/span><span class=\"token punctuation\">,<\/span><span class=\"token number\">3<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> num_classes<span class=\"token operator\">&#061;<\/span><span class=\"token number\">10<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    inputs <span class=\"token operator\">&#061;<\/span> layers<span class=\"token punctuation\">.<\/span>Input<span class=\"token punctuation\">(<\/span>shape<span class=\"token operator\">&#061;<\/span>input_shape<span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token comment\"># \u7b80\u5316\u7248ResNet<\/span><br \/>\n    x <span class=\"token operator\">&#061;<\/span> layers<span class=\"token punctuation\">.<\/span>Conv2D<span class=\"token punctuation\">(<\/span><span class=\"token number\">16<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">3<\/span><span class=\"token punctuation\">,<\/span> padding<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;same&#039;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">(<\/span>inputs<span class=\"token punctuation\">)<\/span><br \/>\n    x <span class=\"token operator\">&#061;<\/span> layers<span class=\"token punctuation\">.<\/span>BatchNormalization<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">(<\/span>x<span class=\"token punctuation\">)<\/span><br \/>\n    x <span class=\"token operator\">&#061;<\/span> layers<span class=\"token punctuation\">.<\/span>ReLU<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">(<\/span>x<span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token comment\"># \u6dfb\u52a0\u6b8b\u5dee\u5757&#xff08;\u7701\u7565\u8be6\u7ec6\u5b9e\u73b0&#xff09;<\/span><br \/>\n    x <span class=\"token operator\">&#061;<\/span> layers<span class=\"token punctuation\">.<\/span>GlobalAveragePooling2D<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">(<\/span>x<span class=\"token punctuation\">)<\/span><br \/>\n    outputs <span class=\"token operator\">&#061;<\/span> layers<span class=\"token punctuation\">.<\/span>Dense<span class=\"token punctuation\">(<\/span>num_classes<span class=\"token punctuation\">,<\/span> activation<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;softmax&#039;<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">(<\/span>x<span class=\"token punctuation\">)<\/span><br \/>\n    model <span class=\"token operator\">&#061;<\/span> tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>Model<span class=\"token punctuation\">(<\/span>inputs<span class=\"token punctuation\">,<\/span> outputs<span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token keyword\">return<\/span> model<\/p>\n<p><span class=\"token comment\"># \u65e5\u5fd7\u76ee\u5f55<\/span><br \/>\nlog_dir <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#039;logs\/resnet_cifar\/&#039;<\/span><\/p>\n<h4>5.2 \u8d85\u53c2\u6570\u641c\u7d22\u4e0e\u8bb0\u5f55<\/h4>\n<p><span class=\"token comment\"># \u4f7f\u7528HParams\u8bb0\u5f55\u4e0d\u540c\u5b66\u4e60\u7387\u7684\u7ed3\u679c<\/span><br \/>\n<span class=\"token keyword\">from<\/span> tensorboard<span class=\"token punctuation\">.<\/span>plugins<span class=\"token punctuation\">.<\/span>hparams <span class=\"token keyword\">import<\/span> api <span class=\"token keyword\">as<\/span> hp<\/p>\n<p>HP_LR <span class=\"token operator\">&#061;<\/span> hp<span class=\"token punctuation\">.<\/span>HParam<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#039;learning_rate&#039;<\/span><span class=\"token punctuation\">,<\/span> hp<span class=\"token punctuation\">.<\/span>RealInterval<span class=\"token punctuation\">(<\/span><span class=\"token number\">1e-4<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">1e-2<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><br \/>\nhparams_log_dir <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#039;logs\/hparam_resnet\/&#039;<\/span><\/p>\n<p><span class=\"token keyword\">def<\/span> <span class=\"token function\">train_and_log<\/span><span class=\"token punctuation\">(<\/span>lr<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    model <span class=\"token operator\">&#061;<\/span> build_resnet20<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    model<span class=\"token punctuation\">.<\/span><span class=\"token builtin\">compile<\/span><span class=\"token punctuation\">(<\/span>optimizer<span class=\"token operator\">&#061;<\/span>tf<span class=\"token punctuation\">.<\/span>keras<span class=\"token punctuation\">.<\/span>optimizers<span class=\"token punctuation\">.<\/span>Adam<span class=\"token punctuation\">(<\/span>lr<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                  loss<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#039;categorical_crossentropy&#039;<\/span><span class=\"token punctuation\">,<\/span> metrics<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">[<\/span><span class=\"token string\">&#039;accuracy&#039;<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    tensorboard_cb <span class=\"token operator\">&#061;<\/span> callbacks<span class=\"token punctuation\">.<\/span>TensorBoard<span class=\"token punctuation\">(<\/span>log_dir<span class=\"token operator\">&#061;<\/span>hparams_log_dir <span class=\"token operator\">&#043;<\/span> <span class=\"token string-interpolation\"><span class=\"token string\">f&#039;lr_<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>lr<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">&#039;<\/span><\/span><span class=\"token punctuation\">,<\/span> histogram_freq<span class=\"token operator\">&#061;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    hparams_cb <span class=\"token operator\">&#061;<\/span> hp<span class=\"token punctuation\">.<\/span>KerasCallback<span class=\"token punctuation\">(<\/span>hparams_log_dir<span class=\"token punctuation\">,<\/span> <span class=\"token punctuation\">{<\/span>HP_LR<span class=\"token punctuation\">:<\/span> lr<span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    history <span class=\"token operator\">&#061;<\/span> model<span class=\"token punctuation\">.<\/span>fit<span class=\"token punctuation\">(<\/span>x_train<span class=\"token punctuation\">,<\/span> y_train<span class=\"token punctuation\">,<\/span> epochs<span class=\"token operator\">&#061;<\/span><span class=\"token number\">10<\/span><span class=\"token punctuation\">,<\/span> batch_size<span class=\"token operator\">&#061;<\/span><span class=\"token number\">128<\/span><span class=\"token punctuation\">,<\/span> validation_split<span class=\"token operator\">&#061;<\/span><span class=\"token number\">0.1<\/span><span class=\"token punctuation\">,<\/span><br \/>\n                        callbacks<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">[<\/span>tensorboard_cb<span class=\"token punctuation\">,<\/span> hparams_cb<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">,<\/span> verbose<span class=\"token operator\">&#061;<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">)<\/span><br \/>\n    val_acc <span class=\"token operator\">&#061;<\/span> history<span class=\"token punctuation\">.<\/span>history<span class=\"token punctuation\">[<\/span><span class=\"token string\">&#039;val_accuracy&#039;<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">[<\/span><span class=\"token operator\">&#8211;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">]<\/span><br \/>\n    <span class=\"token keyword\">return<\/span> val_acc<\/p>\n<p><span class=\"token keyword\">for<\/span> lr <span class=\"token keyword\">in<\/span> <span class=\"token punctuation\">[<\/span><span class=\"token number\">0.001<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">0.0005<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token number\">0.0001<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    acc <span class=\"token operator\">&#061;<\/span> train_and_log<span class=\"token punctuation\">(<\/span>lr<span class=\"token punctuation\">)<\/span><br \/>\n    <span class=\"token keyword\">print<\/span><span class=\"token punctuation\">(<\/span><span class=\"token string-interpolation\"><span class=\"token string\">f&#034;LR&#061;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>lr<span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">, val_acc&#061;<\/span><span class=\"token interpolation\"><span class=\"token punctuation\">{<\/span>acc<span class=\"token punctuation\">:<\/span><span class=\"token format-spec\">.4f<\/span><span class=\"token punctuation\">}<\/span><\/span><span class=\"token string\">&#034;<\/span><\/span><span class=\"token punctuation\">)<\/span><\/p>\n<h4>5.3 \u7ed3\u679c\u5206\u6790<\/h4>\n<p>\u542f\u52a8TensorBoard&#xff0c;\u5728HParams\u9875\u9762\u4e2d\u53ef\u4ee5\u5bf9\u6bd4\u4e0d\u540c\u5b66\u4e60\u7387\u7684\u635f\u5931\/\u51c6\u786e\u7387\u66f2\u7ebf&#xff0c;\u4ee5\u53ca\u5e73\u884c\u5750\u6807\u56fe&#xff0c;\u76f4\u89c2\u9009\u62e9\u6700\u4f73\u8d85\u53c2\u6570\u3002<\/p>\n<h3>6. \u5e38\u89c1\u5751\u70b9\u4e0e\u6392\u9519\u603b\u7ed3<\/h3>\n<h4>6.1 \u65e5\u5fd7\u5199\u5165\u5751\u70b9<\/h4>\n<ul>\n<li>\n<p>\u57511&#xff1a;\u6bcf\u6b21\u8bad\u7ec3\u4f7f\u7528\u76f8\u540c\u7684\u65e5\u5fd7\u76ee\u5f55&#xff0c;\u5bfc\u81f4TensorBoard\u4e2d\u66f2\u7ebf\u6df7\u4e71&#xff08;\u591a\u6761\u66f2\u7ebf\u91cd\u53e0&#xff09;\u3002<\/p>\n<ul>\n<li>\u89e3\u51b3&#xff1a;\u4f7f\u7528\u65f6\u95f4\u6233\u6216\u8fd0\u884cID\u521b\u5efa\u5b50\u76ee\u5f55&#xff1a;log_dir &#061; f&#039;logs\/run_{datetime.datetime.now().strftime(&#034;%Y%m%d-%H%M%S&#034;)}&#039;\u3002<\/li>\n<\/ul>\n<\/li>\n<li>\n<p>\u57512&#xff1a;\u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d\u5fd8\u8bb0\u5237\u65b0summary writer&#xff0c;\u5bfc\u81f4\u65e5\u5fd7\u6ca1\u6709\u53ca\u65f6\u5199\u5165\u78c1\u76d8\u3002<\/p>\n<ul>\n<li>\u89e3\u51b3&#xff1a;\u4f7f\u7528tf.summary.flush()&#xff0c;\u6216writer\u9ed8\u8ba4\u6bcf120\u79d2\u81ea\u52a8\u5237\u65b0\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<h4>6.2 \u76f4\u65b9\u56fe\u7406\u89e3\u8bef\u533a<\/h4>\n<ul>\n<li>\n<p>\u57513&#xff1a;\u76f4\u65b9\u56fe\u53ea\u663e\u793a\u6743\u91cd\u7684\u5206\u5e03&#xff0c;\u4f46\u65e0\u6cd5\u76f4\u63a5\u770b\u51fa\u68af\u5ea6\u6d88\u5931\u3002\u9700\u7ed3\u5408\u68af\u5ea6\u7684\u76f4\u65b9\u56fe&#xff08;\u9700\u624b\u52a8\u8bb0\u5f55\u68af\u5ea6&#xff09;\u3002<\/p>\n<ul>\n<li>\u5efa\u8bae&#xff1a;\u5728\u81ea\u5b9a\u4e49\u5faa\u73af\u4e2d\u8bb0\u5f55tf.summary.histogram(&#039;gradients\/layer&#039;, grad)\u3002<\/li>\n<\/ul>\n<\/li>\n<li>\n<p>\u57514&#xff1a;\u76f4\u65b9\u56fe\u6a2a\u8f74\u53d6\u503c\u8303\u56f4\u8fc7\u5927&#xff0c;\u4e0d\u6613\u89c2\u5bdf\u7ec6\u8282\u3002\u53ef\u70b9\u51fb\u56fe\u8868\u9009\u62e9\u5bf9\u6570\u5750\u6807\u6216\u8c03\u6574\u8303\u56f4\u3002<\/p>\n<\/li>\n<\/ul>\n<h4>6.3 \u8ba1\u7b97\u56fe\u53ef\u89c6\u5316<\/h4>\n<ul>\n<li>\n<p>\u57515&#xff1a;\u6a21\u578b\u56fe\u8fc7\u4e8e\u590d\u6742&#xff0c;\u96be\u4ee5\u9605\u8bfb\u3002\u53ef\u4ee5\u4f7f\u7528tf.keras.utils.plot_model\u751f\u6210\u9759\u6001\u56fe&#xff0c;\u6216\u4f7f\u7528TensorBoard\u7684\u56fe\u9762\u677f\u6298\u53e0\u8282\u70b9\u3002<\/p>\n<\/li>\n<li>\n<p>\u57516&#xff1a;write_graph&#061;True\u5bfc\u81f4\u65e5\u5fd7\u6587\u4ef6\u8fc7\u5927&#xff0c;\u5c24\u5176\u662f\u5728\u4fdd\u5b58\u591a\u4e2acheckpoint\u65f6\u3002<\/p>\n<ul>\n<li>\u89e3\u51b3&#xff1a;\u53ea\u5728\u9700\u8981\u67e5\u770b\u56fe\u65f6\u5f00\u542f&#xff0c;\u5e73\u65f6\u5173\u95ed\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<h4>6.4 \u5d4c\u5165\u6295\u5f71\u4eea<\/h4>\n<ul>\n<li>\n<p>\u57517&#xff1a;\u5d4c\u5165\u6295\u5f71\u4eea\u8981\u6c42\u5411\u91cf\u7ef4\u5ea6\u8f83\u4f4e&#xff08;&lt;5000\u7ef4&#xff09;&#xff0c;\u5426\u5219t-SNE\/PCA\u8ba1\u7b97\u5f88\u6162\u3002\u5efa\u8bae\u5148\u964d\u7ef4\u518d\u8bb0\u5f55\u3002<\/p>\n<\/li>\n<li>\n<p>\u57518&#xff1a;\u5143\u6570\u636e\u6587\u4ef6&#xff08;TSV&#xff09;\u7684\u7f16\u7801\u95ee\u9898&#xff0c;\u4e2d\u6587\u53ef\u80fd\u4e71\u7801\u3002\u4f7f\u7528UTF-8\u7f16\u7801&#xff0c;\u4e14\u5728TensorBoard\u4e2d\u9700\u652f\u6301\u4e2d\u6587\u3002<\/p>\n<\/li>\n<\/ul>\n<h4>6.5 Profile<\/h4>\n<ul>\n<li>\u57519&#xff1a;Profile\u5728CPU\u4e0a\u610f\u4e49\u4e0d\u5927&#xff0c;\u5efa\u8bae\u5728GPU\u4e0a\u8fd0\u884c\u65f6\u4f7f\u7528\u3002<\/li>\n<li>\u575110&#xff1a;profile_batch\u8303\u56f4\u8fc7\u5927\u4f1a\u5bfc\u81f4\u5927\u91cf\u989d\u5916\u5f00\u9500&#xff0c;\u4ec5\u9700\u5206\u6790\u5c11\u91cfbatch\u5373\u53ef\u3002<\/li>\n<\/ul>\n<h3>7. \u77e5\u8bc6\u70b9\u603b\u7ed3 &#043; \u8bfe\u540e\u4f5c\u4e1a<\/h3>\n<h4>7.1 \u6838\u5fc3\u77e5\u8bc6\u70b9\u68b3\u7406<\/h4>\n<ul>\n<li>\u6807\u91cf\u4eea\u8868\u677f&#xff1a;\u8bb0\u5f55\u635f\u5931\u3001\u51c6\u786e\u7387\u7b49&#xff0c;Keras\u56de\u8c03\u81ea\u52a8&#xff1b;\u81ea\u5b9a\u4e49\u5faa\u73af\u7528tf.summary.scalar\u3002<\/li>\n<li>\u8ba1\u7b97\u56fe&#xff1a;write_graph\u5f00\u542f\u540e&#xff0c;\u53ef\u67e5\u770b\u6a21\u578b\u7ed3\u6784\u3002<\/li>\n<li>\u76f4\u65b9\u56fe&#xff1a;\u8bb0\u5f55\u6743\u91cd\u3001\u68af\u5ea6\u5206\u5e03&#xff0c;\u8bca\u65ad\u68af\u5ea6\u6d88\u5931\/\u7206\u70b8\u3002<\/li>\n<li>\u5d4c\u5165\u6295\u5f71\u4eea&#xff1a;\u53ef\u89c6\u5316\u9ad8\u7ef4\u5411\u91cf&#xff0c;\u9700\u63d0\u4f9b\u5143\u6570\u636e\u548c\u5411\u91cf\u6587\u4ef6\u3002<\/li>\n<li>HParams&#xff1a;\u5bf9\u6bd4\u8d85\u53c2\u6570&#xff0c;\u8f85\u52a9\u8c03\u4f18\u3002<\/li>\n<li>Profile&#xff1a;\u5206\u6790\u6027\u80fd\u74f6\u9888\u3002<\/li>\n<\/ul>\n<h4>7.2 \u57fa\u7840\u4f5c\u4e1a<\/h4>\n<li>\u4f7f\u7528Keras\u7684TensorBoard\u56de\u8c03\u8bad\u7ec3\u4e00\u4e2a\u7b80\u5355\u7684DNN\u6a21\u578b&#xff0c;\u8bb0\u5f55\u635f\u5931\u548c\u51c6\u786e\u7387&#xff0c;\u542f\u52a8TensorBoard\u67e5\u770b\u66f2\u7ebf\u3002<\/li>\n<li>\u5728\u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d&#xff0c;\u624b\u52a8\u8bb0\u5f55\u6bcf\u5c42\u7684\u6743\u91cd\u76f4\u65b9\u56fe\u548c\u68af\u5ea6\u76f4\u65b9\u56fe&#xff0c;\u89c2\u5bdf\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u7684\u53d8\u5316\u3002<\/li>\n<li>\u4f7f\u7528\u5d4c\u5165\u6295\u5f71\u4eea\u53ef\u89c6\u5316MNIST\u6d4b\u8bd5\u96c6\u7684\u6700\u540e\u4e00\u4e2a\u5168\u8fde\u63a5\u5c42\u8f93\u51fa&#xff08;256\u7ef4&#xff09;&#xff0c;\u964d\u7ef4\u52302D\u62163D&#xff0c;\u89c2\u5bdf\u4e0d\u540c\u7c7b\u522b\u7684\u805a\u7c7b\u60c5\u51b5\u3002<\/li>\n<h4>7.3 \u8fdb\u9636\u5b9e\u64cd\u4f5c\u4e1a<\/h4>\n<p>\u4efb\u52a1&#xff1a;\u4f7f\u7528HParams\u548cTensorBoard\u5bf9\u6bd4\u4e0d\u540c\u7f51\u7edc\u6df1\u5ea6\u7684\u6027\u80fd<\/p>\n<ul>\n<li>\u6784\u5efa3\u4e2a\u4e0d\u540c\u6df1\u5ea6\u7684\u5168\u8fde\u63a5\u7f51\u7edc&#xff08;2\u5c42\u30014\u5c42\u30016\u5c42&#xff09;&#xff0c;\u56fa\u5b9a\u5176\u4ed6\u8d85\u53c2\u6570\u3002<\/li>\n<li>\u4f7f\u7528HParams\u8bb0\u5f55\u6df1\u5ea6&#xff08;\u4f5c\u4e3a\u5206\u7c7b\u8d85\u53c2\u6570&#xff09;&#xff0c;\u5e76\u8bb0\u5f55\u6bcf\u4e2aepoch\u7684\u9a8c\u8bc1\u635f\u5931\u3002<\/li>\n<li>\u5728TensorBoard\u4e2d\u751f\u6210\u5e73\u884c\u5750\u6807\u56fe&#xff0c;\u5206\u6790\u6df1\u5ea6\u4e0e\u6700\u7ec8\u51c6\u786e\u7387\u7684\u5173\u7cfb\u3002<\/li>\n<li>\u540c\u65f6\u8bb0\u5f55\u8bad\u7ec3\u65f6\u95f4\u548c\u53c2\u6570\u91cf&#xff0c;\u63a2\u8ba8\u6df1\u5ea6\u4e0e\u8bad\u7ec3\u901f\u5ea6\u7684\u6743\u8861\u3002<\/li>\n<\/ul>\n<h4>7.4 \u601d\u8003\u62d3\u5c55\u9898<\/h4>\n<li>\n<p>\u5728\u81ea\u5b9a\u4e49\u8bad\u7ec3\u5faa\u73af\u4e2d&#xff0c;\u5982\u679c\u8981\u8bb0\u5f55\u6bcf\u4e2abatch\u7684\u635f\u5931&#xff0c;\u800c\u4e0d\u662f\u6bcf\u4e2aepoch&#xff0c;\u5e94\u5982\u4f55\u5b9e\u73b0&#xff1f;\u8fd9\u79cd\u7ec6\u7c92\u5ea6\u8bb0\u5f55\u6709\u4f55\u4f18\u7f3a\u70b9&#xff1f;<\/p>\n<\/li>\n<li>\n<p>\u5d4c\u5165\u6295\u5f71\u4eea\u4e2d\u7684t-SNE\u964d\u7ef4\u7ed3\u679c\u5177\u6709\u968f\u673a\u6027&#xff0c;\u5982\u4f55\u8ba9\u591a\u6b21\u8fd0\u884c\u7ed3\u679c\u53ef\u6bd4\u8f83&#xff1f;\u5c1d\u8bd5\u56fa\u5b9a\u968f\u673a\u79cd\u5b50\u3002<\/p>\n<\/li>\n<li>\n<p>\u5047\u8bbe\u4f60\u770b\u5230\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u6743\u91cd\u7684\u76f4\u65b9\u56fe\u9010\u6e10\u5411\u96f6\u6536\u7f29&#xff0c;\u8fd9\u4f1a\u662f\u4ec0\u4e48\u95ee\u9898&#xff1f;\u5982\u4f55\u89e3\u51b3&#xff1f;<\/p>\n<\/li>\n<hr \/>\n<p>\u4e0b\u4e00\u8bfe\u9884\u544a&#xff1a;\u81ea\u5b9a\u4e49\u5c42\u4e0e\u81ea\u5b9a\u4e49\u635f\u5931\u51fd\u6570\u2014\u2014\u6211\u4eec\u5c06\u5b66\u4e60\u5982\u4f55\u6269\u5c55TensorFlow&#xff0c;\u7f16\u5199\u6ee1\u8db3\u7279\u6b8a\u9700\u6c42\u7684\u5c42\u548c\u635f\u5931\u51fd\u6570&#xff0c;\u63d0\u5347\u6a21\u578b\u7684\u7075\u6d3b\u6027\u3002<\/p>\n<hr \/>\n<h2>&#x1f517;\u300aTensorFlow2.x: \u6df1\u5ea6\u5b66\u4e60\u5165\u95e8\u5230\u9ad8\u9636\u5b9e\u6218\u6559\u7a0b\u300b\u7cfb\u5217\u8bfe\u7a0b\u5bfc\u822a<\/h2>\n<p>\u53bb\u8ba2\u9605<\/p>\n<p>\u7b2c\u4e00\u90e8\u5206&#xff1a;\u57fa\u7840\u5165\u95e8&#xff08;1-10 \u8bfe&#xff09; \u7b2c\u4e8c\u90e8\u5206&#xff1a;\u795e\u7ecf\u7f51\u7edc\u6838\u5fc3&#xff08;11-25 \u8bfe&#xff09; \u7b2c\u4e09\u90e8\u5206&#xff1a;\u8fdb\u9636\u7f51\u7edc\u4e0e\u6846\u67b6\u9ad8\u9636&#xff08;26-40 \u8bfe&#xff09; \u7b2c\u56db\u90e8\u5206&#xff1a;\u4f01\u4e1a\u5b9e\u6218\u4e0e\u9879\u76ee\u843d\u5730&#xff08;41-50 \u8bfe&#xff09;<\/p>\n<p>&#x1f31f; \u611f\u8c22\u60a8\u8010\u5fc3\u9605\u8bfb\u5230\u8fd9\u91cc&#xff01; &#x1f4a1; \u5982\u679c\u672c\u6587\u5bf9\u60a8\u6709\u6240\u542f\u53d1\u6b22\u8fce&#xff1a; &#x1f44d; \u70b9\u8d5e&#x1f4cc; \u6536\u85cf &#x1f4e4; \u5206\u4eab\u7ed9\u66f4\u591a\u9700\u8981\u7684\u4f19\u4f34\u3002 &#x1f5e3;\ufe0f \u671f\u5f85\u5728\u8bc4\u8bba\u533a\u770b\u5230\u60a8\u7684\u60f3\u6cd5, \u5171\u540c\u8fdb\u6b65\u3002 &#x1f514; \u5173\u6ce8\u6211&#xff0c;\u6301\u7eed\u83b7\u53d6\u66f4\u591a\u5e72\u8d27\u5185\u5bb9&#xff5e; &#x1f917; \u6211\u4eec\u4e0b\u7bc7\u6587\u7ae0\u89c1&#xff5e;<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u76ee\u5f551. \u8bfe\u524d\u5bfc\u8bfb1.1 \u672c\u8282\u8bfe\u5b66\u4e60\u76ee\u68071.2 \u77e5\u8bc6\u91cd\u96be\u70b91.3 \u5b66\u4e60\u524d\u7f6e\u6761\u4ef61.4 \u5b66\u5b8c\u53ef\u638c\u63e1\u80fd\u529b1.5 \u884c\u4e1a\u5e94\u7528\u573a\u666f2. \u6838\u5fc3\u7406\u8bba\u7cbe\u8bb22.1 TensorBoard \u67b6\u6784\u4e0e\u65e5\u5fd72.2 \u6807\u91cf\u4eea\u8868\u677f&#xff08;Scalar&#xff09;2.3 \u8ba1\u7b97\u56fe\u53ef\u89c6\u53162.4 \u76f4\u65b9\u56fe\u4eea\u8868\u677f&#xff08;Histogram&#xff09;2.5 \u5d4c\u5165\u6295\u5f71\u4eea&#xff08;Embedding Projector&#xff09;2.6 \u8d85\u53c2\u6570\u8c03\u4f18&#xff08;HParams&amp;#xf<\/p>\n","protected":false},"author":2,"featured_media":103554,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[81,3677,50],"topic":[],"class_list":["post-103555","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-python","tag-tensorflow","tag-50"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v20.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/103555.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u6587\u7ae0\u76ee\u5f551. \u8bfe\u524d\u5bfc\u8bfb1.1 \u672c\u8282\u8bfe\u5b66\u4e60\u76ee\u68071.2 \u77e5\u8bc6\u91cd\u96be\u70b91.3 \u5b66\u4e60\u524d\u7f6e\u6761\u4ef61.4 \u5b66\u5b8c\u53ef\u638c\u63e1\u80fd\u529b1.5 \u884c\u4e1a\u5e94\u7528\u573a\u666f2. \u6838\u5fc3\u7406\u8bba\u7cbe\u8bb22.1 TensorBoard \u67b6\u6784\u4e0e\u65e5\u5fd72.2 \u6807\u91cf\u4eea\u8868\u677f&#xff08;Scalar&#xff09;2.3 \u8ba1\u7b97\u56fe\u53ef\u89c6\u53162.4 \u76f4\u65b9\u56fe\u4eea\u8868\u677f&#xff08;Histogram&#xff09;2.5 \u5d4c\u5165\u6295\u5f71\u4eea&#xff08;Embedding Projector&#xff09;2.6 \u8d85\u53c2\u6570\u8c03\u4f18&#xff08;HParams&amp;#xf\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/103555.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-10T15:09:10+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260910150908-6aa2c81445d73.jpg\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/103555.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/103555.html\",\"name\":\"\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-09-10T15:09:10+00:00\",\"dateModified\":\"2026-09-10T15:09:10+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/103555.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/103555.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/103555.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/103555.html","og_locale":"zh_CN","og_type":"article","og_title":"\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u6587\u7ae0\u76ee\u5f551. \u8bfe\u524d\u5bfc\u8bfb1.1 \u672c\u8282\u8bfe\u5b66\u4e60\u76ee\u68071.2 \u77e5\u8bc6\u91cd\u96be\u70b91.3 \u5b66\u4e60\u524d\u7f6e\u6761\u4ef61.4 \u5b66\u5b8c\u53ef\u638c\u63e1\u80fd\u529b1.5 \u884c\u4e1a\u5e94\u7528\u573a\u666f2. \u6838\u5fc3\u7406\u8bba\u7cbe\u8bb22.1 TensorBoard \u67b6\u6784\u4e0e\u65e5\u5fd72.2 \u6807\u91cf\u4eea\u8868\u677f&#xff08;Scalar&#xff09;2.3 \u8ba1\u7b97\u56fe\u53ef\u89c6\u53162.4 \u76f4\u65b9\u56fe\u4eea\u8868\u677f&#xff08;Histogram&#xff09;2.5 \u5d4c\u5165\u6295\u5f71\u4eea&#xff08;Embedding Projector&#xff09;2.6 \u8d85\u53c2\u6570\u8c03\u4f18&#xff08;HParams&amp;#xf","og_url":"https:\/\/www.wsisp.com\/helps\/103555.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-09-10T15:09:10+00:00","og_image":[{"url":"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260910150908-6aa2c81445d73.jpg"}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"8 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/103555.html","url":"https:\/\/www.wsisp.com\/helps\/103555.html","name":"\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-09-10T15:09:10+00:00","dateModified":"2026-09-10T15:09:10+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/103555.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/103555.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/103555.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"\u7b2c38\u8bfe\uff1aTensorFlow\uff5c\u53ef\u89c6\u5316\u5de5\u5177TensorBoard\u5168\u7528\u6cd5\u3010\u65e5\u5fd7\u5199\u5165\u3001\u6307\u6807\u76d1\u63a7\u3001\u7f51\u7edc\u53ef\u89c6\u5316\u3011"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/103555","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=103555"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/103555\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media\/103554"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=103555"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=103555"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=103555"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=103555"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}