{"id":385,"date":"2022-08-01T21:02:52","date_gmt":"2022-08-01T13:02:52","guid":{"rendered":"http:\/\/www.gislxz.top\/?p=385"},"modified":"2022-08-01T21:05:22","modified_gmt":"2022-08-01T13:05:22","slug":"%e6%b7%b1%e5%ba%a6%e5%ad%a6%e4%b9%a0%e7%ac%94%e8%ae%b0%ef%bc%886%ef%bc%89","status":"publish","type":"post","link":"https:\/\/www.gislxz.com\/index.php\/2022\/08\/01\/%e6%b7%b1%e5%ba%a6%e5%ad%a6%e4%b9%a0%e7%ac%94%e8%ae%b0%ef%bc%886%ef%bc%89\/","title":{"rendered":"\u6df1\u5ea6\u5b66\u4e60\u7b14\u8bb0\uff086\uff09"},"content":{"rendered":"\n<p><a href=\"https:\/\/aistudio.baidu.com\/aistudio\/projectdetail\/1599219\" target=\"_blank\"  rel=\"nofollow\" >paddle\u5b98\u65b9\u6559\u7a0b\u7684\u8fd9\u4e00\u7ae0<\/a>\u5f00\u59cb\u6765\u5904\u7406\u635f\u5931\u51fd\u6570\u7684\u95ee\u9898<\/p>\n\n\n\n<p>\u5728\u4e4b\u524d\u7684\u65b9\u6848\u4e2d\uff0c\u6211\u4eec\u590d\u7528\u4e86\u623f\u4ef7\u9884\u6d4b\u6a21\u578b\u7684\u635f\u5931\u51fd\u6570-\u5747\u65b9\u8bef\u5dee\u3002\u4ece\u9884\u6d4b\u6548\u679c\u6765\u770b\uff0c\u867d\u7136\u635f\u5931\u4e0d\u65ad\u4e0b\u964d\uff0c\u6a21\u578b\u7684\u9884\u6d4b\u503c\u9010\u6e10\u903c\u8fd1\u771f\u5b9e\u503c\uff0c\u4f46\u6a21\u578b\u7684\u6700\u7ec8\u6548\u679c\u4e0d\u591f\u7406\u60f3\u3002\u7a76\u5176\u6839\u672c\uff0c\u4e0d\u540c\u7684\u6df1\u5ea6\u5b66\u4e60\u4efb\u52a1\u9700\u8981\u6709\u5404\u81ea\u9002\u5b9c\u7684\u635f\u5931\u51fd\u6570\u3002\u6211\u4eec\u4ee5\u623f\u4ef7\u9884\u6d4b\u548c\u624b\u5199\u6570\u5b57\u8bc6\u522b\u4e24\u4e2a\u4efb\u52a1\u4e3a\u4f8b\uff0c\u8be6\u7ec6\u5256\u6790\u5176\u4e2d\u7684\u7f18\u7531\u5982\u4e0b\uff1a<\/p>\n\n\n\n<ol class=\"wp-block-list\"><li>\u623f\u4ef7\u9884\u6d4b\u662f\u56de\u5f52\u4efb\u52a1\uff0c\u800c\u624b\u5199\u6570\u5b57\u8bc6\u522b\u662f\u5206\u7c7b\u4efb\u52a1\uff0c\u4f7f\u7528\u5747\u65b9\u8bef\u5dee\u4f5c\u4e3a\u5206\u7c7b\u4efb\u52a1\u7684\u635f\u5931\u51fd\u6570\u5b58\u5728\u903b\u8f91\u548c\u6548\u679c\u4e0a\u7684\u7f3a\u6b20\u3002<\/li><li>\u623f\u4ef7\u53ef\u4ee5\u662f\u5927\u4e8e0\u7684\u4efb\u4f55\u6d6e\u70b9\u6570\uff0c\u800c\u624b\u5199\u6570\u5b57\u8bc6\u522b\u7684\u8f93\u51fa\u53ea\u53ef\u80fd\u662f0~9\u4e4b\u95f4\u768410\u4e2a\u6574\u6570\uff0c\u76f8\u5f53\u4e8e\u4e00\u79cd\u6807\u7b7e\u3002<\/li><li>\u5728\u623f\u4ef7\u9884\u6d4b\u7684\u6848\u4f8b\u4e2d\uff0c\u7531\u4e8e\u623f\u4ef7\u672c\u8eab\u662f\u4e00\u4e2a\u8fde\u7eed\u7684\u5b9e\u6570\u503c\uff0c\u56e0\u6b64\u4ee5\u6a21\u578b\u8f93\u51fa\u7684\u6570\u503c\u548c\u771f\u5b9e\u623f\u4ef7\u5dee\u8ddd\u4f5c\u4e3a\u635f\u5931\u51fd\u6570\uff08Loss\uff09\u662f\u7b26\u5408\u9053\u7406\u7684\u3002\u4f46\u5bf9\u4e8e\u5206\u7c7b\u95ee\u9898\uff0c\u771f\u5b9e\u7ed3\u679c\u662f\u5206\u7c7b\u6807\u7b7e\uff0c\u800c\u6a21\u578b\u8f93\u51fa\u662f\u5b9e\u6570\u503c\uff0c\u5bfc\u81f4\u4ee5\u4e24\u8005\u76f8\u51cf\u4f5c\u4e3a\u635f\u5931\u4e0d\u5177\u5907\u7269\u7406\u542b\u4e49\u3002<\/li><\/ol>\n\n\n\n<p>\u63a5\u7740\u6559\u7a0b\u8bb2\u4e86softmax\uff0c\u4ea4\u53c9\u71b5\uff0c\u90fd\u662f\u5728\u8bb2\u6570\u5b66\u539f\u7406\uff0c\u5230\u4ee3\u7801\u4e0a\u90fd\u662f\u7528\u5c01\u88c5\u597d\u7684\u51fd\u6570<\/p>\n\n\n\n<p>\u5728\u624b\u5199\u6570\u5b57\u8bc6\u522b\u4efb\u52a1\u4e2d\uff0c\u4ec5\u6539\u52a8\u4e09\u884c\u4ee3\u7801\uff0c\u5c31\u53ef\u4ee5\u5c06\u5728\u73b0\u6709\u6a21\u578b\u7684\u635f\u5931\u51fd\u6570\u66ff\u6362\u6210\u4ea4\u53c9\u71b5\uff08Cross_entropy\uff09\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>\u5728\u8bfb\u53d6\u6570\u636e\u90e8\u5206\uff0c\u5c06\u6807\u7b7e\u7684\u7c7b\u578b\u8bbe\u7f6e\u6210int\uff0c\u4f53\u73b0\u5b83\u662f\u4e00\u4e2a\u6807\u7b7e\u800c\u4e0d\u662f\u5b9e\u6570\u503c\uff08\u98de\u6868\u6846\u67b6\u9ed8\u8ba4\u5c06\u6807\u7b7e\u5904\u7406\u6210int64\uff09\u3002<\/li><li>\u5728\u7f51\u7edc\u5b9a\u4e49\u90e8\u5206\uff0c\u5c06\u8f93\u51fa\u5c42\u6539\u6210\u201c\u8f93\u51fa\u5341\u4e2a\u6807\u7b7e\u7684\u6982\u7387\u201d\u7684\u6a21\u5f0f\u3002<\/li><li>\u5728\u8bad\u7ec3\u8fc7\u7a0b\u90e8\u5206\uff0c\u5c06\u635f\u5931\u51fd\u6570\u4ece\u5747\u65b9\u8bef\u5dee\u6362\u6210\u4ea4\u53c9\u71b5\u3002<\/li><\/ul>\n\n\n\n<p>\u5728\u6570\u636e\u5904\u7406\u90e8\u5206\uff0c\u9700\u8981\u4fee\u6539\u6807\u7b7e\u53d8\u91cfLabel\u7684\u683c\u5f0f\uff0c\u4ee3\u7801\u5982\u4e0b\u6240\u793a\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>\u4ece\uff1alabel = np.reshape(labels[i], [1]).astype('float32')<\/li><li>\u5230\uff1alabel = np.reshape(labels[i], [1]).astype('int64')<\/li><\/ul>\n\n\n\n<pre class=\"wp-block-code\"><code>import os\nimport random\nimport paddle\nimport numpy as np\nfrom PIL import Image\nimport gzip\nimport json\n\n# \u5b9a\u4e49\u6570\u636e\u96c6\u8bfb\u53d6\u5668\ndef load_data(mode='train'):\n\n    # \u6570\u636e\u6587\u4ef6\n    datafile = '.\/work\/mnist.json.gz'\n    print('loading mnist dataset from {} ......'.format(datafile))\n    data = json.load(gzip.open(datafile))\n    train_set, val_set, eval_set = data\n\n    # \u6570\u636e\u96c6\u76f8\u5173\u53c2\u6570\uff0c\u56fe\u7247\u9ad8\u5ea6IMG_ROWS, \u56fe\u7247\u5bbd\u5ea6IMG_COLS\n    IMG_ROWS = 28\n    IMG_COLS = 28\n\n    if mode == 'train':\n        imgs = train_set&#91;0]\n        labels = train_set&#91;1]\n    elif mode == 'valid':\n        imgs = val_set&#91;0]\n        labels = val_set&#91;1]\n    elif mode == 'eval':\n        imgs = eval_set&#91;0]\n        labels = eval_set&#91;1]\n\n    imgs_length = len(imgs)\n\n    assert len(imgs) == len(labels), \\\n          \"length of train_imgs({}) should be the same as train_labels({})\".format(\n                  len(imgs), len(labels))\n\n    index_list = list(range(imgs_length))\n\n    # \u8bfb\u5165\u6570\u636e\u65f6\u7528\u5230\u7684batchsize\n    BATCHSIZE = 100\n\n    # \u5b9a\u4e49\u6570\u636e\u751f\u6210\u5668\n    def data_generator():\n        if mode == 'train':\n            random.shuffle(index_list)\n        imgs_list = &#91;]\n        labels_list = &#91;]\n        for i in index_list:\n            img = np.reshape(imgs&#91;i], &#91;1, IMG_ROWS, IMG_COLS]).astype('float32')\n            label = np.reshape(labels&#91;i], &#91;1]).astype('int64')\n            imgs_list.append(img) \n            labels_list.append(label)\n            if len(imgs_list) == BATCHSIZE:\n                yield np.array(imgs_list), np.array(labels_list)\n                imgs_list = &#91;]\n                labels_list = &#91;]\n\n        # \u5982\u679c\u5269\u4f59\u6570\u636e\u7684\u6570\u76ee\u5c0f\u4e8eBATCHSIZE\uff0c\n        # \u5219\u5269\u4f59\u6570\u636e\u4e00\u8d77\u6784\u6210\u4e00\u4e2a\u5927\u5c0f\u4e3alen(imgs_list)\u7684mini-batch\n        if len(imgs_list) &gt; 0:\n            yield np.array(imgs_list), np.array(labels_list)\n\n    return data_generator<\/code><\/pre>\n\n\n\n<p>\u5728\u7f51\u7edc\u5b9a\u4e49\u90e8\u5206\uff0c\u9700\u8981\u4fee\u6539\u8f93\u51fa\u5c42\u7ed3\u6784\uff0c\u4ee3\u7801\u5982\u4e0b\u6240\u793a\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>\u4ece\uff1aself.fc = Linear(in_features=980, out_features=1)<\/li><li>\u5230\uff1aself.fc = Linear(in_features=980, out_features=10)<\/li><\/ul>\n\n\n\n<pre class=\"wp-block-code\"><code># \u5b9a\u4e49 SimpleNet \u7f51\u7edc\u7ed3\u6784\nimport paddle\nfrom paddle.nn import Conv2D, MaxPool2D, Linear\nimport paddle.nn.functional as F\n# \u591a\u5c42\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u5b9e\u73b0\nclass MNIST(paddle.nn.Layer):\n     def __init__(self):\n         super(MNIST, self).__init__()\n         \n         # \u5b9a\u4e49\u5377\u79ef\u5c42\uff0c\u8f93\u51fa\u7279\u5f81\u901a\u9053out_channels\u8bbe\u7f6e\u4e3a20\uff0c\u5377\u79ef\u6838\u7684\u5927\u5c0fkernel_size\u4e3a5\uff0c\u5377\u79ef\u6b65\u957fstride=1\uff0cpadding=2\n         self.conv1 = Conv2D(in_channels=1, out_channels=20, kernel_size=5, stride=1, padding=2)\n         # \u5b9a\u4e49\u6c60\u5316\u5c42\uff0c\u6c60\u5316\u6838\u7684\u5927\u5c0fkernel_size\u4e3a2\uff0c\u6c60\u5316\u6b65\u957f\u4e3a2\n         self.max_pool1 = MaxPool2D(kernel_size=2, stride=2)\n         # \u5b9a\u4e49\u5377\u79ef\u5c42\uff0c\u8f93\u51fa\u7279\u5f81\u901a\u9053out_channels\u8bbe\u7f6e\u4e3a20\uff0c\u5377\u79ef\u6838\u7684\u5927\u5c0fkernel_size\u4e3a5\uff0c\u5377\u79ef\u6b65\u957fstride=1\uff0cpadding=2\n         self.conv2 = Conv2D(in_channels=20, out_channels=20, kernel_size=5, stride=1, padding=2)\n         # \u5b9a\u4e49\u6c60\u5316\u5c42\uff0c\u6c60\u5316\u6838\u7684\u5927\u5c0fkernel_size\u4e3a2\uff0c\u6c60\u5316\u6b65\u957f\u4e3a2\n         self.max_pool2 = MaxPool2D(kernel_size=2, stride=2)\n         # \u5b9a\u4e49\u4e00\u5c42\u5168\u8fde\u63a5\u5c42\uff0c\u8f93\u51fa\u7ef4\u5ea6\u662f10\n         self.fc = Linear(in_features=980, out_features=10)\n         \n    # \u5b9a\u4e49\u7f51\u7edc\u524d\u5411\u8ba1\u7b97\u8fc7\u7a0b\uff0c\u5377\u79ef\u540e\u7d27\u63a5\u7740\u4f7f\u7528\u6c60\u5316\u5c42\uff0c\u6700\u540e\u4f7f\u7528\u5168\u8fde\u63a5\u5c42\u8ba1\u7b97\u6700\u7ec8\u8f93\u51fa\n    # \u5377\u79ef\u5c42\u6fc0\u6d3b\u51fd\u6570\u4f7f\u7528Relu\uff0c\u5168\u8fde\u63a5\u5c42\u6fc0\u6d3b\u51fd\u6570\u4f7f\u7528softmax\n     def forward(self, inputs):\n         x = self.conv1(inputs)\n         x = F.relu(x)\n         x = self.max_pool1(x)\n         x = self.conv2(x)\n         x = F.relu(x)\n         x = self.max_pool2(x)\n         x = paddle.reshape(x, &#91;x.shape&#91;0], 980])\n         x = self.fc(x)\n         x = F.softmax(x)\n         return x<\/code><\/pre>\n\n\n\n<p>\u4fee\u6539\u8ba1\u7b97\u635f\u5931\u7684\u51fd\u6570\uff0c\u4ece\u5747\u65b9\u8bef\u5dee\uff08\u5e38\u7528\u4e8e\u56de\u5f52\u95ee\u9898\uff09\u5230\u4ea4\u53c9\u71b5\u8bef\u5dee\uff08\u5e38\u7528\u4e8e\u5206\u7c7b\u95ee\u9898\uff09\uff0c\u4ee3\u7801\u5982\u4e0b\u6240\u793a\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>\u4ece\uff1aloss = paddle.nn.functional.square_error_cost(predict, label)<\/li><li>\u5230\uff1aloss = paddle.nn.functional.cross_entropy(predict, label)<\/li><\/ul>\n\n\n\n<pre class=\"wp-block-code\"><code>#\u4ec5\u4fee\u6539\u8ba1\u7b97\u635f\u5931\u7684\u51fd\u6570\uff0c\u4ece\u5747\u65b9\u8bef\u5dee\uff08\u5e38\u7528\u4e8e\u56de\u5f52\u95ee\u9898\uff09\u5230\u4ea4\u53c9\u71b5\u8bef\u5dee\uff08\u5e38\u7528\u4e8e\u5206\u7c7b\u95ee\u9898\uff09\ndef train(model):\n    model.train()\n    #\u8c03\u7528\u52a0\u8f7d\u6570\u636e\u7684\u51fd\u6570\n    train_loader = load_data('train')\n    opt = paddle.optimizer.SGD(learning_rate=0.01, parameters=model.parameters())\n    EPOCH_NUM = 3\n    for epoch_id in range(EPOCH_NUM):\n        for batch_id, data in enumerate(train_loader()):\n            #\u51c6\u5907\u6570\u636e\n            images, labels = data\n            images = paddle.to_tensor(images)\n            labels = paddle.to_tensor(labels)\n            #\u524d\u5411\u8ba1\u7b97\u7684\u8fc7\u7a0b\n            predicts = model(images)\n            \n            #\u8ba1\u7b97\u635f\u5931\uff0c\u4f7f\u7528\u4ea4\u53c9\u71b5\u635f\u5931\u51fd\u6570\uff0c\u53d6\u4e00\u4e2a\u6279\u6b21\u6837\u672c\u635f\u5931\u7684\u5e73\u5747\u503c\n            loss = F.cross_entropy(predicts, labels)\n            avg_loss = paddle.mean(loss)\n            \n            #\u6bcf\u8bad\u7ec3\u4e86200\u6279\u6b21\u7684\u6570\u636e\uff0c\u6253\u5370\u4e0b\u5f53\u524dLoss\u7684\u60c5\u51b5\n            if batch_id % 200 == 0:\n                print(\"epoch: {}, batch: {}, loss is: {}\".format(epoch_id, batch_id, avg_loss.numpy()))\n            \n            #\u540e\u5411\u4f20\u64ad\uff0c\u66f4\u65b0\u53c2\u6570\u7684\u8fc7\u7a0b\n            avg_loss.backward()\n            # \u6700\u5c0f\u5316loss,\u66f4\u65b0\u53c2\u6570\n            opt.step()\n            # \u6e05\u9664\u68af\u5ea6\n            opt.clear_grad()\n   \n    #\u4fdd\u5b58\u6a21\u578b\u53c2\u6570\n    paddle.save(model.state_dict(), 'mnist.pdparams')\n    \n#\u521b\u5efa\u6a21\u578b    \nmodel = MNIST()\n#\u542f\u52a8\u8bad\u7ec3\u8fc7\u7a0b\ntrain(model)<\/code><\/pre>\n\n\n\n<p>\u7531\u4e8e\u6211\u4eec\u4fee\u6539\u4e86\u6a21\u578b\u7684\u8f93\u51fa\u683c\u5f0f\uff0c\u56e0\u6b64\u4f7f\u7528\u6a21\u578b\u505a\u9884\u6d4b\u65f6\u7684\u4ee3\u7801\u4e5f\u9700\u8981\u505a\u76f8\u5e94\u7684\u8c03\u6574\u3002\u4ece\u6a21\u578b\u8f93\u51fa10\u4e2a\u6807\u7b7e\u7684\u6982\u7387\u4e2d\u9009\u62e9\u6700\u5927\u7684\uff0c\u5c06\u5176\u6807\u7b7e\u7f16\u53f7\u8f93\u51fa\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code># \u8bfb\u53d6\u4e00\u5f20\u672c\u5730\u7684\u6837\u4f8b\u56fe\u7247\uff0c\u8f6c\u53d8\u6210\u6a21\u578b\u8f93\u5165\u7684\u683c\u5f0f\ndef load_image(img_path):\n    # \u4eceimg_path\u4e2d\u8bfb\u53d6\u56fe\u50cf\uff0c\u5e76\u8f6c\u4e3a\u7070\u5ea6\u56fe\n    im = Image.open(img_path).convert('L')\n    im = im.resize((28, 28), Image.ANTIALIAS)\n    im = np.array(im).reshape(1, 1, 28, 28).astype(np.float32)\n    # \u56fe\u50cf\u5f52\u4e00\u5316\n    im = 1.0 - im \/ 255.\n    return im\n\n# \u5b9a\u4e49\u9884\u6d4b\u8fc7\u7a0b\nmodel = MNIST()\nparams_file_path = 'mnist.pdparams'\nimg_path = 'work\/example_0.jpg'\n# \u52a0\u8f7d\u6a21\u578b\u53c2\u6570\nparam_dict = paddle.load(params_file_path)\nmodel.load_dict(param_dict)\n# \u704c\u5165\u6570\u636e\nmodel.eval()\ntensor_img = load_image(img_path)\n#\u6a21\u578b\u53cd\u998810\u4e2a\u5206\u7c7b\u6807\u7b7e\u7684\u5bf9\u5e94\u6982\u7387\nresults = model(paddle.to_tensor(tensor_img))\n#\u53d6\u6982\u7387\u6700\u5927\u7684\u6807\u7b7e\u4f5c\u4e3a\u9884\u6d4b\u8f93\u51fa\nlab = np.argsort(results.numpy())\nprint(\"\u672c\u6b21\u9884\u6d4b\u7684\u6570\u5b57\u662f: \", lab&#91;0]&#91;-1])<\/code><\/pre>\n\n\n\n<p>\u4e4b\u540e\u6211\u4eec\u5bf9pytorch\u7684\u4ee3\u7801\u4e5f\u8fdb\u884c\u540c\u6837\u7684\u4fee\u6539,\u6ce8\u610ftorch\u7684\u4ea4\u53c9\u71b5\u51fd\u6570\u5bf9label\u8981\u6c42\u662f1\u7ef4\u7684\uff0c\u800c\u6570\u636e\u8fed\u4ee3\u5668\u8f93\u51fa\u7684\u662f\u4e8c\u7ef4\uff0c\u5373\u30101\uff0cbatchsize\u3011\uff0c\u9700\u8981\u964d\u4e00\u4e0b\u7ef4\uff0c\u53e6\u5916\u8bad\u7ec3\u4e86\u4e00\u4e0b\u53d1\u73b0loss\u57fa\u672c\u6ca1\u52a8\uff0c\u628alr\u63d0\u9ad8\u5230\u52300.01\u660e\u663e\u6539\u5584\u3002<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>from torch.nn import Conv2d,MaxPool2d,Linear\n#\u6570\u636e\u5904\u7406\u90e8\u5206\u4e4b\u540e\u7684\u4ee3\u7801\uff0c\u6570\u636e\u8bfb\u53d6\u7684\u90e8\u5206\u8c03\u7528load_data\u51fd\u6570\n# \u5b9a\u4e49\u7f51\u7edc\u7ed3\u6784\uff0c\u540c\u4e0a\u4e00\u8282\u6240\u4f7f\u7528\u7684\u7f51\u7edc\u7ed3\u6784\nclass Mnist(nn.Module):\n    def __init__(self):\n        super(Mnist,self).__init__()\n        self.conv1 = Conv2d(in_channels=1,out_channels=20,kernel_size=5,stride=1,padding=2)\n        self.max_pool1 = MaxPool2d(kernel_size=2,stride=2)\n        self.conv2 = Conv2d(in_channels=20,out_channels=20,kernel_size=5,stride=1,padding=2)\n        self.max_pool2 = MaxPool2d(kernel_size=2,stride=2)\n        self.fc = Linear(in_features=980,out_features=10)\n    def forward(self,x):\n        x = self.conv1(x)\n        x = torch.relu(x)\n        x = self.max_pool1(x)\n        x = self.conv2(x)\n        x = torch.relu(x)\n        x = self.max_pool2(x)\n        x = torch.reshape(x,&#91;x.shape&#91;0],-1])\n        x = self.fc(x)\n        x = F.softmax(x)\n        return x\n\n# \u8bad\u7ec3\u914d\u7f6e\uff0c\u5e76\u542f\u52a8\u8bad\u7ec3\u8fc7\u7a0b\nmodel = Mnist()\nmodel.train(mode=True)\n#\u8c03\u7528\u52a0\u8f7d\u6570\u636e\u7684\u51fd\u6570\ntrain_loader = load_data('train')\noptimizer = optim.SGD(model.parameters(),lr= 0.01)\n\nEPOCH_NUM = 10\nfor epoch_id in range(EPOCH_NUM):\n    for batch_id, data in enumerate(train_loader()):\n        #\u51c6\u5907\u6570\u636e\uff0c\u53d8\u5f97\u66f4\u52a0\u7b80\u6d01\n        image_data, label_data = data\n        image = torch.tensor(image_data)\n        label = torch.tensor(label_data)\n        image = torch.reshape(image,&#91;image.shape&#91;0],1,28,28]) \n        #\u524d\u5411\u8ba1\u7b97\u7684\u8fc7\u7a0b\n        predict = model(image)\n            \n        #\u8ba1\u7b97\u635f\u5931\uff0c\u53d6\u4e00\u4e2a\u6279\u6b21\u6837\u672c\u635f\u5931\u7684\u5e73\u5747\u503c\n        loss = F.cross_entropy(predict, label.squeeze(dim=1))\n        avg_loss = torch.mean(loss)\n            \n        #\u6bcf\u8bad\u7ec3\u4e86200\u6279\u6b21\u7684\u6570\u636e\uff0c\u6253\u5370\u4e0b\u5f53\u524dLoss\u7684\u60c5\u51b5\n        if batch_id % 200 == 0:\n            print(\"epoch: {}, batch: {}, loss is: {}\".format(epoch_id, batch_id, avg_loss.detach().numpy()))\n            \n        #\u540e\u5411\u4f20\u64ad\uff0c\u66f4\u65b0\u53c2\u6570\u7684\u8fc7\u7a0b\n        avg_loss.backward()\n        optimizer.step()\n        model.zero_grad()<\/code><\/pre>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"280\" height=\"240\" src=\"http:\/\/www.gislxz.com\/wp-content\/uploads\/2022\/08\/121.gif\" alt=\"\" class=\"wp-image-388\"\/><\/figure>\n","protected":false},"excerpt":{"rendered":"<p>softmax\uff0c\u635f\u5931\u51fd\u6570\uff08\u4ea4\u53c9\u71b5\uff09<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[21],"tags":[],"class_list":["post-385","post","type-post","status-publish","format-standard","hentry","category-21"],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/posts\/385","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/comments?post=385"}],"version-history":[{"count":2,"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/posts\/385\/revisions"}],"predecessor-version":[{"id":391,"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/posts\/385\/revisions\/391"}],"wp:attachment":[{"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/media?parent=385"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/categories?post=385"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.gislxz.com\/index.php\/wp-json\/wp\/v2\/tags?post=385"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}