{"id":97854,"date":"2026-08-30T13:08:56","date_gmt":"2026-08-30T05:08:56","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/97854.html"},"modified":"2026-08-30T13:08:56","modified_gmt":"2026-08-30T05:08:56","slug":"%e6%89%8b%e6%9c%ba%e4%bb%b7%e6%a0%bc%e5%a4%9a%e5%88%86%e7%b1%bb_%e5%85%a8%e8%bf%9e%e6%8e%a5%e7%a5%9e%e7%bb%8f%e7%bd%91%e7%bb%9c%e6%95%99%e7%a8%8b","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/97854.html","title":{"rendered":"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b"},"content":{"rendered":"

\u3010PyTorch\u3011\u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u5b9e\u73b0\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b（\u9644\u5b8c\u6574\u4ee3\u7801 + \u8bad\u7ec3\u7ed3\u679c）<\/h2>\n

\u672c\u6587\u7528 PyTorch \u642d\u5efa\u4e00\u4e2a\u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc，\u5bf9\u624b\u673a\u4ef7\u683c\u8fdb\u884c\u56db\u5206\u7c7b\u9884\u6d4b\u3002\u4ee3\u7801\u8986\u76d6\u4e86\u6570\u636e\u5904\u7406 \u2192 \u6a21\u578b\u8bad\u7ec3 \u2192 \u6a21\u578b\u8bc4\u4f30\u7684\u5b8c\u6574\u6d41\u7a0b，\u5e76\u628a\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u4fdd\u5b58\u5230\u672c\u5730\u3001\u8bc4\u4f30\u65f6\u518d\u52a0\u8f7d\u56de\u6765\u3002\u5168\u7a0b CPU \u5373\u53ef\u8fd0\u884c，\u9002\u5408\u5165\u95e8\u6559\u5b66\u3002<\/p>\n

\n

\u4e00\u3001\u4efb\u52a1\u80cc\u666f<\/h3>\n

\u7ed9\u5b9a 2000 \u6761\u624b\u673a\u914d\u7f6e\u6570\u636e（\u7535\u6c60\u3001\u5185\u5b58\u3001RAM\u3001\u6444\u50cf\u5934\u3001\u5c4f\u5e55\u7b49 20 \u4e2a\u7279\u5f81），\u9884\u6d4b\u8be5\u624b\u673a\u5c5e\u4e8e\u54ea\u4e00\u4e2a\u4ef7\u683c\u6863\u4f4d（price_range，\u5171 4 \u7c7b：0 \/ 1 \/ 2 \/ 3）\u3002<\/p>\n

\u8fd9\u662f\u4e00\u4e2a\u5178\u578b\u7684\u591a\u5206\u7c7b\u4efb\u52a1\u3002\u672c\u6587\u7528\u4e00\u4e2a 3 \u5c42\u5168\u8fde\u63a5\u7f51\u7edc（\u4e24\u4e2a\u9690\u85cf\u5c42 + \u4e00\u4e2a\u8f93\u51fa\u5c42），\u9690\u85cf\u5c42\u4f7f\u7528 ReLU \u6fc0\u6d3b\u51fd\u6570，\u8f93\u51fa\u5c42\u914d\u5408 CrossEntropyLoss \u5b8c\u6210\u5206\u7c7b\u3002<\/p>\n

\n

\u4e8c\u3001\u6570\u636e\u96c6\u4ecb\u7ecd<\/h3>\n

\u6570\u636e\u96c6\u6765\u81ea Kaggle \u7ecf\u5178\u7684 Mobile Price Classification，\u5171 2000 \u884c \u00d7 21 \u5217，\u5373 20 \u4e2a\u7279\u5f81 + 1 \u4e2a\u76ee\u6807\u53d8\u91cf\u3002<\/p>\n

2.1 \u8fde\u7eed\u6570\u503c\u7279\u5f81（14 \u4e2a）<\/h4>\n\n\u7279\u5f81\u542b\u4e49\u53d6\u503c\u8303\u56f4<\/tr>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
battery_power<\/td>\n\u7535\u6c60\u5bb9\u91cf (mAh)<\/td>\n501 \u2013 1998<\/td>\n<\/tr>\n
clock_speed<\/td>\n\u5904\u7406\u5668\u4e3b\u9891 (GHz)<\/td>\n0.5 \u2013 3.0<\/td>\n<\/tr>\n
fc<\/td>\n\u524d\u7f6e\u6444\u50cf\u5934 (MP)<\/td>\n0 \u2013 19<\/td>\n<\/tr>\n
pc<\/td>\n\u540e\u7f6e\u6444\u50cf\u5934 (MP)<\/td>\n0 \u2013 20<\/td>\n<\/tr>\n
int_memory<\/td>\n\u5185\u5b58 (GB)<\/td>\n2 \u2013 64<\/td>\n<\/tr>\n
ram<\/td>\n\u8fd0\u884c\u5185\u5b58 (MB)<\/td>\n256 \u2013 3998<\/td>\n<\/tr>\n
m_dep<\/td>\n\u673a\u8eab\u539a\u5ea6 (cm)<\/td>\n0.1 \u2013 1.0<\/td>\n<\/tr>\n
mobile_wt<\/td>\n\u91cd\u91cf (g)<\/td>\n80 \u2013 200<\/td>\n<\/tr>\n
n_cores<\/td>\n\u5904\u7406\u5668\u6838\u5fc3\u6570<\/td>\n1 \u2013 8<\/td>\n<\/tr>\n
px_height<\/td>\n\u5c4f\u5e55\u9ad8\u5ea6 (px)<\/td>\n0 \u2013 1960<\/td>\n<\/tr>\n
px_width<\/td>\n\u5c4f\u5e55\u5bbd\u5ea6 (px)<\/td>\n500 \u2013 1998<\/td>\n<\/tr>\n
sc_h<\/td>\n\u5c4f\u5e55\u9ad8\u5ea6 (cm)<\/td>\n5 \u2013 19<\/td>\n<\/tr>\n
sc_w<\/td>\n\u5c4f\u5e55\u5bbd\u5ea6 (cm)<\/td>\n0 \u2013 18<\/td>\n<\/tr>\n
talk_time<\/td>\n\u901a\u8bdd\u65f6\u957f (h)<\/td>\n2 \u2013 20<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

2.2 \u4e8c\u503c\u7279\u5f81（6 \u4e2a，\u53d6\u503c 0\/1）<\/h4>\n

blue（\u84dd\u7259）\u3001dual_sim（\u53cc\u5361）\u3001four_g（4G）\u3001three_g（3G）\u3001touch_screen（\u89e6\u6478\u5c4f）\u3001wifi\u3002<\/p>\n

2.3 \u76ee\u6807\u53d8\u91cf\u5206\u5e03<\/h4>\n

price_range \u56db\u4e2a\u7c7b\u522b\u5404 500 \u6761，\u6070\u597d 25% \u5b8c\u5168\u5747\u8861，\u56e0\u6b64\u5efa\u6a21\u65f6\u65e0\u9700\u5904\u7406\u7c7b\u522b\u4e0d\u5e73\u8861，\u76f4\u63a5\u7528\u51c6\u786e\u7387\u8bc4\u4f30\u5373\u53ef\u3002<\/p>\n

\u26a0\ufe0f \u6ce8\u610f：\u5404\u7279\u5f81\u91cf\u7eb2\u5dee\u5f02\u6781\u5927（ram \u6700\u5927 3998，\u800c m_dep \u53ea\u6709 0.1～1.0），\u8bad\u7ec3\u524d\u5fc5\u987b\u505a\u6807\u51c6\u5316，\u5426\u5219\u5927\u6570\u503c\u7279\u5f81\u4f1a\u5728\u68af\u5ea6\u4e0b\u964d\u4e2d\u4e3b\u5bfc\u6743\u91cd\u66f4\u65b0\u3002<\/p>\n

\n

\u4e09\u3001\u73af\u5883\u51c6\u5907<\/h3>\n\n\u4f9d\u8d56\u7248\u672c\u8bf4\u660e<\/tr>\n\n\n\n\n\n
Python<\/td>\n3.8+<\/td>\n\u672c\u6587\u7528 3.12<\/td>\n<\/tr>\n
PyTorch<\/td>\n2.x<\/td>\nCPU \u7248\u5373\u53ef<\/td>\n<\/tr>\n
scikit-learn<\/td>\n1.x<\/td>\n\u5212\u5206\u6570\u636e\u96c6\u3001\u6807\u51c6\u5316\u3001\u8bc4\u4f30<\/td>\n<\/tr>\n
pandas \/ numpy<\/td>\n\u2014<\/td>\n\u6570\u636e\u8bfb\u53d6\u4e0e\u5904\u7406<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

\u5b89\u88c5\u547d\u4ee4：<\/p>\n

pip install<\/span> torch scikit-learn pandas numpy<\/p>\n

\n

\u56db\u3001\u9879\u76ee\u7ed3\u6784<\/h3>\n

day_05\/
\n\u251c\u2500\u2500 data\/
\n\u2502 \u2514\u2500\u2500 \u624b\u673a\u4ef7\u683c\u9884\u6d4b.csv # \u6570\u636e\u96c6
\n\u251c\u2500\u2500 model\/ # \u8bad\u7ec3\u540e\u81ea\u52a8\u751f\u6210
\n\u2502 \u2514\u2500\u2500 phone_price_net.pt # \u4fdd\u5b58\u7684\u6a21\u578b\u6743\u91cd
\n\u2514\u2500\u2500 model.py # \u672c\u6587\u5b8c\u6574\u4ee3\u7801<\/p>\n

\n

\u4e94\u3001\u7f51\u7edc\u7ed3\u6784\u8bbe\u8ba1<\/h3>\n

\u8f93\u5165\u5c42(20) \u2192 \u5168\u8fde\u63a5\u5c42(64, ReLU) \u2192 \u5168\u8fde\u63a5\u5c42(32, ReLU) \u2192 \u8f93\u51fa\u5c42(4)<\/p>\n

    \n
  • \u8f93\u5165\u7ef4\u5ea6：20（\u7279\u5f81\u6570）<\/li>\n
  • \u9690\u85cf\u5c42 1：64 \u4e2a\u795e\u7ecf\u5143，ReLU \u6fc0\u6d3b<\/li>\n
  • \u9690\u85cf\u5c42 2：32 \u4e2a\u795e\u7ecf\u5143，ReLU \u6fc0\u6d3b<\/li>\n
  • \u8f93\u51fa\u7ef4\u5ea6：4（\u5bf9\u5e94 4 \u4e2a\u4ef7\u683c\u6863\u4f4d）<\/li>\n<\/ul>\n

    \u8f93\u51fa\u5c42\u4e0d\u52a0\u6fc0\u6d3b\u51fd\u6570：nn.CrossEntropyLoss \u5185\u90e8\u5df2\u7ecf\u96c6\u6210\u4e86 softmax，\u6240\u4ee5\u8f93\u51fa\u5c42\u76f4\u63a5\u63a5\u7ebf\u6027\u5c42\u5373\u53ef，\u6807\u7b7e\u4f20\u539f\u59cb\u7c7b\u522b\u7f16\u53f7（long \u7c7b\u578b）\u3002<\/p>\n

    \u8d85\u53c2\u6570：\u5b66\u4e60\u7387 1e-3\u3001Adam \u4f18\u5316\u5668\u3001batch_size=64\u3001epochs=50\u3001\u6d4b\u8bd5\u96c6\u6bd4\u4f8b 20%\u3001\u968f\u673a\u79cd\u5b50 42\u3002<\/p>\n

    \n

    \u516d\u3001\u5b8c\u6574\u4ee3\u7801<\/h3>\n

    """
    \n\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b \u2014\u2014 \u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc（PyTorch）<\/p>\n

    \u6d41\u7a0b：\u6570\u636e\u5904\u7406 -> \u6a21\u578b\u8bad\u7ec3 -> \u6a21\u578b\u8bc4\u4f30
    \n\u7ed3\u6784：3 \u5c42\u5168\u8fde\u63a5\u7f51\u7edc（20 -> 64 -> 32 -> 4），\u9690\u85cf\u5c42\u7528 ReLU \u6fc0\u6d3b
    \n\u8bf4\u660e：
    \n – \u6570\u636e\u4ece data\/ \u76ee\u5f55\u8bfb\u53d6
    \n – \u8bad\u7ec3\u597d\u7684\u6a21\u578b\u4fdd\u5b58\u5230 model\/ \u76ee\u5f55
    \n – \u8bc4\u4f30\u9636\u6bb5\u4ece model\/ \u76ee\u5f55\u52a0\u8f7d\u6a21\u578b
    \n\u8fd0\u884c：python model.py（\u8def\u5f84\u5df2\u7528 __file__ \u81ea\u9002\u5e94，\u4efb\u610f\u76ee\u5f55\u5747\u53ef\u8fd0\u884c）
    \n"""<\/span><\/p>\n

    import<\/span> os<\/p>\n

    import<\/span> matplotlib
    \nmatplotlib.<\/span>use(<\/span>"Agg"<\/span>)<\/span> # \u65e0\u754c\u9762\u540e\u7aef：\u76f4\u63a5\u4fdd\u5b58\u56fe\u7247，\u4e0d\u5f39\u51fa\u7a97\u53e3<\/span>
    \nimport<\/span> matplotlib.<\/span>pyplot as<\/span> plt
    \nimport<\/span> numpy as<\/span> np
    \nimport<\/span> pandas as<\/span> pd
    \nimport<\/span> torch
    \nimport<\/span> torch.<\/span>nn as<\/span> nn
    \nimport<\/span> torch.<\/span>optim as<\/span> optim
    \nfrom<\/span> sklearn.<\/span>model_selection import<\/span> train_test_split
    \nfrom<\/span> sklearn.<\/span>preprocessing import<\/span> StandardScaler
    \nfrom<\/span> sklearn.<\/span>metrics import<\/span> accuracy_score,<\/span> classification_report
    \nfrom<\/span> torch.<\/span>utils.<\/span>data import<\/span> DataLoader,<\/span> TensorDataset<\/p>\n

    # ================= \u8def\u5f84\u4e0e\u8d85\u53c2\u6570 =================<\/span>
    \nBASE_DIR =<\/span> os.<\/span>path.<\/span>dirname(<\/span>os.<\/span>path.<\/span>abspath(<\/span>__file__)<\/span>)<\/span> # model.py \u6240\u5728\u76ee\u5f55<\/span>
    \nDATA_PATH =<\/span> os.<\/span>path.<\/span>join(<\/span>BASE_DIR,<\/span> "data"<\/span>,<\/span> "\u624b\u673a\u4ef7\u683c\u9884\u6d4b.csv"<\/span>)<\/span>
    \nMODEL_DIR =<\/span> os.<\/span>path.<\/span>join(<\/span>BASE_DIR,<\/span> "model"<\/span>)<\/span>
    \nMODEL_PATH =<\/span> os.<\/span>path.<\/span>join(<\/span>MODEL_DIR,<\/span> "phone_price_net.pt"<\/span>)<\/span><\/p>\n

    INPUT_DIM =<\/span> 20<\/span> # \u8f93\u5165\u7279\u5f81\u6570<\/span>
    \nHIDDEN_1 =<\/span> 64<\/span> # \u9690\u85cf\u5c42 1 \u795e\u7ecf\u5143\u6570<\/span>
    \nHIDDEN_2 =<\/span> 32<\/span> # \u9690\u85cf\u5c42 2 \u795e\u7ecf\u5143\u6570<\/span>
    \nOUTPUT_DIM =<\/span> 4<\/span> # \u7c7b\u522b\u6570（price_range: 0\/1\/2\/3）<\/span><\/p>\n

    LEARNING_RATE =<\/span> 1e-3<\/span>
    \nEPOCHS =<\/span> 50<\/span>
    \nBATCH_SIZE =<\/span> 64<\/span>
    \nTEST_SIZE =<\/span> 0.2<\/span>
    \nRANDOM_SEED =<\/span> 42<\/span><\/p>\n

    # ================= 1. \u7f51\u7edc\u7ed3\u6784 =================<\/span>
    \nclass<\/span> PhonePriceNet<\/span>(<\/span>nn.<\/span>Module)<\/span>:<\/span>
    \n """\u4e09\u5c42\u5168\u8fde\u63a5\u7f51\u7edc：fc1 -> fc2 -> fc3，\u9690\u85cf\u5c42\u7528 ReLU \u6fc0\u6d3b"""<\/span><\/p>\n

    def<\/span> __init__<\/span>(<\/span>self,<\/span> input_dim,<\/span> hidden_1,<\/span> hidden_2,<\/span> output_dim)<\/span>:<\/span>
    \n super<\/span>(<\/span>)<\/span>.<\/span>__init__(<\/span>)<\/span>
    \n self.<\/span>fc1 =<\/span> nn.<\/span>Linear(<\/span>input_dim,<\/span> hidden_1)<\/span> # \u7b2c 1 \u5c42<\/span>
    \n self.<\/span>fc2 =<\/span> nn.<\/span>Linear(<\/span>hidden_1,<\/span> hidden_2)<\/span> # \u7b2c 2 \u5c42<\/span>
    \n self.<\/span>fc3 =<\/span> nn.<\/span>Linear(<\/span>hidden_2,<\/span> output_dim)<\/span> # \u7b2c 3 \u5c42（\u8f93\u51fa\u5c42）<\/span><\/p>\n

    def<\/span> forward<\/span>(<\/span>self,<\/span> x)<\/span>:<\/span>
    \n x =<\/span> torch.<\/span>relu(<\/span>self.<\/span>fc1(<\/span>x)<\/span>)<\/span>
    \n x =<\/span> torch.<\/span>relu(<\/span>self.<\/span>fc2(<\/span>x)<\/span>)<\/span>
    \n x =<\/span> self.<\/span>fc3(<\/span>x)<\/span> # \u8f93\u51fa\u5c42\u4e0d\u52a0\u6fc0\u6d3b，\u4ea4\u7ed9 CrossEntropyLoss<\/span>
    \n return<\/span> x<\/p>\n

    # ================= 2. \u6570\u636e\u5904\u7406 =================<\/span>
    \ndef<\/span> prepare_data<\/span>(<\/span>)<\/span>:<\/span>
    \n """\u8bfb\u53d6 CSV\u3001\u5207\u5206\u7279\u5f81\/\u6807\u7b7e\u3001\u6807\u51c6\u5316\u3001\u5212\u5206\u8bad\u7ec3\/\u6d4b\u8bd5\u96c6\u5e76\u5c01\u88c5\u4e3a DataLoader"""<\/span>
    \n df =<\/span> pd.<\/span>read_csv(<\/span>DATA_PATH)<\/span><\/p>\n

    X =<\/span> df.<\/span>iloc[<\/span>:<\/span>,<\/span> :<\/span>–<\/span>1<\/span>]<\/span>.<\/span>values.<\/span>astype(<\/span>np.<\/span>float32)<\/span> # \u524d 20 \u5217：\u7279\u5f81<\/span>
    \n y =<\/span> df.<\/span>iloc[<\/span>:<\/span>,<\/span> –<\/span>1<\/span>]<\/span>.<\/span>values.<\/span>astype(<\/span>np.<\/span>int64)<\/span> # \u6700\u540e 1 \u5217：\u6807\u7b7e<\/span><\/p>\n

    # \u5206\u5c42\u5212\u5206：\u4fdd\u8bc1\u8bad\u7ec3\/\u6d4b\u8bd5\u96c6\u4e2d\u5404\u7c7b\u522b\u6bd4\u4f8b\u4e00\u81f4<\/span>
    \n X_train,<\/span> X_test,<\/span> y_train,<\/span> y_test =<\/span> train_test_split(<\/span>
    \n X,<\/span> y,<\/span> test_size=<\/span>TEST_SIZE,<\/span> random_state=<\/span>RANDOM_SEED,<\/span> stratify=<\/span>y
    \n )<\/span><\/p>\n

    # \u6807\u51c6\u5316：\u53ea\u7528\u8bad\u7ec3\u96c6\u8ba1\u7b97\u5747\u503c\/\u6807\u51c6\u5dee，\u518d\u4f5c\u7528\u5230\u6d4b\u8bd5\u96c6（\u907f\u514d\u6570\u636e\u6cc4\u9732）<\/span>
    \n scaler =<\/span> StandardScaler(<\/span>)<\/span>
    \n X_train =<\/span> scaler.<\/span>fit_transform(<\/span>X_train)<\/span>
    \n X_test =<\/span> scaler.<\/span>transform(<\/span>X_test)<\/span><\/p>\n

    # \u8f6c\u6210 Tensor \u5e76\u5c01\u88c5\u4e3a DataLoader<\/span>
    \n train_loader =<\/span> DataLoader(<\/span>
    \n TensorDataset(<\/span>torch.<\/span>from_numpy(<\/span>X_train)<\/span>,<\/span> torch.<\/span>from_numpy(<\/span>y_train)<\/span>)<\/span>,<\/span>
    \n batch_size=<\/span>BATCH_SIZE,<\/span> shuffle=<\/span>True<\/span>,<\/span>
    \n )<\/span>
    \n test_loader =<\/span> DataLoader(<\/span>
    \n TensorDataset(<\/span>torch.<\/span>from_numpy(<\/span>X_test)<\/span>,<\/span> torch.<\/span>from_numpy(<\/span>y_test)<\/span>)<\/span>,<\/span>
    \n batch_size=<\/span>BATCH_SIZE,<\/span> shuffle=<\/span>False<\/span>,<\/span>
    \n )<\/span>
    \n return<\/span> train_loader,<\/span> test_loader<\/p>\n

    # ================= 3. \u6a21\u578b\u8bad\u7ec3 =================<\/span>
    \ndef<\/span> train_model<\/span>(<\/span>train_loader)<\/span>:<\/span>
    \n """\u8bad\u7ec3\u6a21\u578b，\u5e76\u628a\u8bad\u7ec3\u597d\u7684\u6743\u91cd\u4fdd\u5b58\u5230 model\/ \u76ee\u5f55"""<\/span>
    \n torch.<\/span>manual_seed(<\/span>RANDOM_SEED)<\/span> # \u56fa\u5b9a\u968f\u673a\u79cd\u5b50，\u4fdd\u8bc1\u7ed3\u679c\u53ef\u590d\u73b0<\/span><\/p>\n

    model =<\/span> PhonePriceNet(<\/span>INPUT_DIM,<\/span> HIDDEN_1,<\/span> HIDDEN_2,<\/span> OUTPUT_DIM)<\/span>
    \n criterion =<\/span> nn.<\/span>CrossEntropyLoss(<\/span>)<\/span> # \u591a\u5206\u7c7b\u635f\u5931<\/span>
    \n optimizer =<\/span> optim.<\/span>Adam(<\/span>model.<\/span>parameters(<\/span>)<\/span>,<\/span> lr=<\/span>LEARNING_RATE)<\/span><\/p>\n

    loss_history =<\/span> [<\/span>]<\/span> # \u8bb0\u5f55\u6bcf\u4e2a epoch \u7684\u5e73\u5747\u635f\u5931<\/span>
    \n for<\/span> epoch in<\/span> range<\/span>(<\/span>1<\/span>,<\/span> EPOCHS +<\/span> 1<\/span>)<\/span>:<\/span>
    \n model.<\/span>train(<\/span>)<\/span>
    \n total_loss =<\/span> 0.0<\/span>
    \n for<\/span> xb,<\/span> yb in<\/span> train_loader:<\/span>
    \n optimizer.<\/span>zero_grad(<\/span>)<\/span> # \u6e05\u7a7a\u68af\u5ea6<\/span>
    \n pred =<\/span> model(<\/span>xb)<\/span> # \u524d\u5411\u4f20\u64ad<\/span>
    \n loss =<\/span> criterion(<\/span>pred,<\/span> yb)<\/span> # \u8ba1\u7b97\u635f\u5931<\/span>
    \n loss.<\/span>backward(<\/span>)<\/span> # \u53cd\u5411\u4f20\u64ad<\/span>
    \n optimizer.<\/span>step(<\/span>)<\/span> # \u66f4\u65b0\u53c2\u6570<\/span>
    \n total_loss +=<\/span> loss.<\/span>item(<\/span>)<\/span> *<\/span> xb.<\/span>size(<\/span>0<\/span>)<\/span>
    \n avg_loss =<\/span> total_loss \/<\/span> len<\/span>(<\/span>train_loader.<\/span>dataset)<\/span>
    \n loss_history.<\/span>append(<\/span>avg_loss)<\/span> # \u8bb0\u5f55\u672c\u8f6e\u635f\u5931<\/span><\/p>\n

    if<\/span> epoch %<\/span> 10<\/span> ==<\/span> 0<\/span>:<\/span>
    \n print<\/span>(<\/span>f" Epoch <\/span>{<\/span>epoch:<\/span>3d<\/span>}<\/span><\/span>\/<\/span>{<\/span>EPOCHS}<\/span><\/span> loss = <\/span>{<\/span>avg_loss:<\/span>.4f<\/span>}<\/span><\/span>"<\/span><\/span>)<\/span><\/p>\n

    # \u4fdd\u5b58\u6a21\u578b<\/span>
    \n os.<\/span>makedirs(<\/span>MODEL_DIR,<\/span> exist_ok=<\/span>True<\/span>)<\/span>
    \n torch.<\/span>save(<\/span>model.<\/span>state_dict(<\/span>)<\/span>,<\/span> MODEL_PATH)<\/span>
    \n print<\/span>(<\/span>f" \u6a21\u578b\u5df2\u4fdd\u5b58 -> <\/span>{<\/span>MODEL_PATH}<\/span><\/span>"<\/span><\/span>)<\/span><\/p>\n

    # \u7ed8\u5236\u5e76\u4fdd\u5b58 loss \u66f2\u7ebf<\/span>
    \n plot_loss_curve(<\/span>loss_history)<\/span>
    \n return<\/span> model<\/p>\n

    # ================= 3.5 \u7ed8\u5236\u635f\u5931\u66f2\u7ebf =================<\/span>
    \ndef<\/span> plot_loss_curve<\/span>(<\/span>loss_history)<\/span>:<\/span>
    \n """\u7ed8\u5236\u8bad\u7ec3\u635f\u5931\u66f2\u7ebf\u5e76\u4fdd\u5b58\u4e3a PNG（model \u76ee\u5f55\u4e0b loss_curve.png）"""<\/span>
    \n plt.<\/span>figure(<\/span>figsize=<\/span>(<\/span>8<\/span>,<\/span> 5<\/span>)<\/span>)<\/span>
    \n plt.<\/span>plot(<\/span>range<\/span>(<\/span>1<\/span>,<\/span> len<\/span>(<\/span>loss_history)<\/span> +<\/span> 1<\/span>)<\/span>,<\/span> loss_history,<\/span>
    \n marker=<\/span>"o"<\/span>,<\/span> markersize=<\/span>3<\/span>,<\/span> linewidth=<\/span>1.5<\/span>,<\/span> color=<\/span>"#1f77b4"<\/span>)<\/span>
    \n plt.<\/span>xlabel(<\/span>"Epoch"<\/span>)<\/span>
    \n plt.<\/span>ylabel(<\/span>"Loss"<\/span>)<\/span>
    \n plt.<\/span>title(<\/span>"Training Loss Curve"<\/span>)<\/span>
    \n plt.<\/span>grid(<\/span>True<\/span>,<\/span> linestyle=<\/span>"–"<\/span>,<\/span> alpha=<\/span>0.5<\/span>)<\/span>
    \n plt.<\/span>tight_layout(<\/span>)<\/span>
    \n loss_path =<\/span> os.<\/span>path.<\/span>join(<\/span>MODEL_DIR,<\/span> "loss_curve.png"<\/span>)<\/span>
    \n plt.<\/span>savefig(<\/span>loss_path,<\/span> dpi=<\/span>150<\/span>)<\/span>
    \n plt.<\/span>close(<\/span>)<\/span>
    \n print<\/span>(<\/span>f" loss \u66f2\u7ebf\u5df2\u4fdd\u5b58 -> <\/span>{<\/span>loss_path}<\/span><\/span>"<\/span><\/span>)<\/span><\/p>\n

    # ================= 4. \u6a21\u578b\u8bc4\u4f30 =================<\/span>
    \ndef<\/span> evaluate_model<\/span>(<\/span>test_loader)<\/span>:<\/span>
    \n """\u4ece model\/ \u76ee\u5f55\u52a0\u8f7d\u8bad\u7ec3\u597d\u7684\u6a21\u578b，\u5728\u6d4b\u8bd5\u96c6\u4e0a\u8bc4\u4f30"""<\/span>
    \n # \u5148\u91cd\u5efa\u540c\u7ed3\u6784\u7684\u7f51\u7edc，\u518d\u52a0\u8f7d\u4fdd\u5b58\u7684\u6743\u91cd<\/span>
    \n model =<\/span> PhonePriceNet(<\/span>INPUT_DIM,<\/span> HIDDEN_1,<\/span> HIDDEN_2,<\/span> OUTPUT_DIM)<\/span>
    \n model.<\/span>load_state_dict(<\/span>torch.<\/span>load(<\/span>MODEL_PATH,<\/span> weights_only=<\/span>True<\/span>)<\/span>)<\/span>
    \n model.<\/span>eval<\/span>(<\/span>)<\/span><\/p>\n

    y_true,<\/span> y_pred =<\/span> [<\/span>]<\/span>,<\/span> [<\/span>]<\/span>
    \n with<\/span> torch.<\/span>no_grad(<\/span>)<\/span>:<\/span> # \u8bc4\u4f30\u9636\u6bb5\u4e0d\u8ba1\u7b97\u68af\u5ea6<\/span>
    \n for<\/span> xb,<\/span> yb in<\/span> test_loader:<\/span>
    \n pred =<\/span> model(<\/span>xb)<\/span>
    \n y_pred.<\/span>extend(<\/span>torch.<\/span>argmax(<\/span>pred,<\/span> dim=<\/span>1<\/span>)<\/span>.<\/span>tolist(<\/span>)<\/span>)<\/span>
    \n y_true.<\/span>extend(<\/span>yb.<\/span>tolist(<\/span>)<\/span>)<\/span><\/p>\n

    acc =<\/span> accuracy_score(<\/span>y_true,<\/span> y_pred)<\/span>
    \n print<\/span>(<\/span>f" \u6d4b\u8bd5\u96c6\u51c6\u786e\u7387 = <\/span>{<\/span>acc *<\/span> 100<\/span>:<\/span>.2f<\/span>}<\/span><\/span>%"<\/span><\/span>)<\/span>
    \n print<\/span>(<\/span>"\\\\n \u5206\u7c7b\u62a5\u544a："<\/span>)<\/span>
    \n print<\/span>(<\/span>classification_report(<\/span>y_true,<\/span> y_pred)<\/span>)<\/span>
    \n return<\/span> acc<\/p>\n

    # ================= \u4e3b\u51fd\u6570 =================<\/span>
    \ndef<\/span> main<\/span>(<\/span>)<\/span>:<\/span>
    \n print<\/span>(<\/span>"="<\/span> *<\/span> 56<\/span>)<\/span>
    \n print<\/span>(<\/span>"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b \u2014\u2014 \u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc"<\/span>)<\/span>
    \n print<\/span>(<\/span>"="<\/span> *<\/span> 56<\/span>)<\/span><\/p>\n

    print<\/span>(<\/span>"\\\\n[1\/3] \u6570\u636e\u5904\u7406 …"<\/span>)<\/span>
    \n train_loader,<\/span> test_loader =<\/span> prepare_data(<\/span>)<\/span>
    \n print<\/span>(<\/span>f" \u8bad\u7ec3\u6837\u672c <\/span>{<\/span>len<\/span>(<\/span>train_loader.<\/span>dataset)<\/span>}<\/span><\/span> \u6761，\u6d4b\u8bd5\u6837\u672c <\/span>{<\/span>len<\/span>(<\/span>test_loader.<\/span>dataset)<\/span>}<\/span><\/span> \u6761"<\/span><\/span>)<\/span><\/p>\n

    print<\/span>(<\/span>"\\\\n[2\/3] \u6a21\u578b\u8bad\u7ec3 …"<\/span>)<\/span>
    \n train_model(<\/span>train_loader)<\/span><\/p>\n

    print<\/span>(<\/span>"\\\\n[3\/3] \u6a21\u578b\u8bc4\u4f30 …"<\/span>)<\/span>
    \n evaluate_model(<\/span>test_loader)<\/span><\/p>\n

    print<\/span>(<\/span>"\\\\n\u5b8c\u6210！"<\/span>)<\/span><\/p>\n

    if<\/span> __name__ ==<\/span> "__main__"<\/span>:<\/span>
    \n main(<\/span>)<\/span><\/p>\n

    \n

    \u4e03\u3001\u4ee3\u7801\u8be6\u89e3<\/h3>\n

    7.1 \u6570\u636e\u5904\u7406 prepare_data()<\/h4>\n
      \n
    • df.iloc[:, :-1] \u53d6\u524d 20 \u5217\u4f5c\u7279\u5f81，df.iloc[:, -1] \u53d6\u6700\u540e\u4e00\u5217\u4f5c\u6807\u7b7e；<\/li>\n
    • \u6807\u7b7e\u8f6c\u6210 np.int64（\u5206\u7c7b\u6807\u7b7e\u5fc5\u987b\u662f\u6574\u6570，\u914d\u5408 CrossEntropyLoss）；<\/li>\n
    • train_test_split(…, stratify=y) \u5206\u5c42\u62bd\u6837，\u4fdd\u8bc1\u8bad\u7ec3\/\u6d4b\u8bd5\u96c6\u91cc 4 \u4e2a\u7c7b\u522b\u7684\u6bd4\u4f8b\u90fd\u4e0e\u539f\u59cb\u6570\u636e\u4e00\u81f4；<\/li>\n
    • StandardScaler \u6807\u51c6\u5316：\u53ea\u7528\u8bad\u7ec3\u96c6 fit，\u518d transform \u5230\u6d4b\u8bd5\u96c6，\u907f\u514d\u6570\u636e\u6cc4\u9732；<\/li>\n
    • \u6700\u540e\u7528 TensorDataset + DataLoader \u5c01\u88c5\u6210\u53ef\u6279\u91cf\u8fed\u4ee3\u7684\u52a0\u8f7d\u5668\u3002<\/li>\n<\/ul>\n

      7.2 \u6a21\u578b\u8bad\u7ec3 train_model()<\/h4>\n

      \u6807\u51c6\u4e94\u6b65\u8bad\u7ec3\u5faa\u73af：<\/p>\n

      optimizer.<\/span>zero_grad(<\/span>)<\/span> # 1. \u6e05\u7a7a\u68af\u5ea6<\/span>
      \npred =<\/span> model(<\/span>xb)<\/span> # 2. \u524d\u5411\u4f20\u64ad<\/span>
      \nloss =<\/span> criterion(<\/span>pred,<\/span> yb)<\/span> # 3. \u8ba1\u7b97\u635f\u5931<\/span>
      \nloss.<\/span>backward(<\/span>)<\/span> # 4. \u53cd\u5411\u4f20\u64ad<\/span>
      \noptimizer.<\/span>step(<\/span>)<\/span> # 5. \u66f4\u65b0\u53c2\u6570<\/span><\/p>\n

      \u635f\u5931\u6309\u6837\u672c\u6570\u52a0\u6743\u5e73\u5747（loss.item() * xb.size(0)），\u907f\u514d\u6700\u540e\u4e00\u4e2a batch \u6837\u672c\u4e0d\u8db3\u5bfc\u81f4\u7684\u504f\u5dee\u3002\u8bad\u7ec3\u5b8c\u6210\u540e\u7528 torch.save \u628a state_dict() \u5b58\u5230 model\/ \u76ee\u5f55\u3002<\/p>\n

      7.3 \u6a21\u578b\u8bc4\u4f30 evaluate_model()<\/h4>\n
        \n
      • \u91cd\u5efa\u540c\u7ed3\u6784\u7f51\u7edc，\u518d\u7528 load_state_dict \u52a0\u8f7d\u4fdd\u5b58\u7684\u6743\u91cd \u2014\u2014 \u8fd9\u5c31\u662f"\u6a21\u578b\u4fdd\u5b58\/\u52a0\u8f7d"\u7684\u6807\u51c6\u7528\u6cd5；<\/li>\n
      • model.eval() \u5173\u95ed Dropout\/BN \u7b49\u8bad\u7ec3\u4e13\u7528\u884c\u4e3a；<\/li>\n
      • torch.no_grad() \u5173\u95ed\u68af\u5ea6\u8ba1\u7b97，\u8282\u7701\u663e\u5b58\u3001\u52a0\u901f\u63a8\u7406；<\/li>\n
      • torch.argmax(pred, dim=1) \u53d6\u51fa\u6bcf\u884c\u5f97\u5206\u6700\u9ad8\u7684\u7c7b\u522b\u7d22\u5f15\u4f5c\u4e3a\u9884\u6d4b\u7ed3\u679c；<\/li>\n
      • \u8f93\u51fa\u51c6\u786e\u7387 + classification_report（\u5404\u7c7b\u522b\u7684\u7cbe\u786e\u7387\u3001\u53ec\u56de\u7387\u3001F1）\u3002<\/li>\n<\/ul>\n

        7.4 \u4e3b\u51fd\u6570 main()<\/h4>\n

        \u7528 if __name__ == "__main__" \u4fdd\u62a4\u5165\u53e3，\u4f9d\u6b21\u8c03\u7528\u4e09\u4e2a\u51fd\u6570，\u5f62\u6210\u6e05\u6670\u7684\u300c\u4e09\u6bb5\u5f0f\u300d\u6d41\u7a0b\u3002<\/p>\n

        \n

        \u516b\u3001\u8fd0\u884c\u7ed3\u679c\u4e0e\u5206\u6790<\/h3>\n

        \u8bad\u7ec3\u6837\u672c 1600 \u6761，\u6d4b\u8bd5\u6837\u672c 400 \u6761<\/p>\n

        Epoch 10\/50 loss = 0.2512
        \nEpoch 20\/50 loss = 0.0849
        \nEpoch 30\/50 loss = 0.0402
        \nEpoch 40\/50 loss = 0.0216
        \nEpoch 50\/50 loss = 0.0126<\/p>\n

        \u6d4b\u8bd5\u96c6\u51c6\u786e\u7387 = 93.50%<\/p>\n

        \u635f\u5931\u4ece 0.2512 \u4e00\u8def\u4e0b\u964d\u5230 0.0126，\u6536\u655b\u826f\u597d\u3002\u6d4b\u8bd5\u96c6\u51c6\u786e\u7387 93.50%，\u5404\u7c7b\u522b F1 \u90fd\u5728 0.92 \u4ee5\u4e0a，\u6a21\u578b\u6cdb\u5316\u80fd\u529b\u5f3a\u3002<\/p>\n

        \u8bad\u7ec3\u635f\u5931\u66f2\u7ebf：
        \n\"\u8bf7\u6dfb\u52a0\u56fe\u7247\u63cf\u8ff0\"<\/p>\n

        \u5206\u7c7b\u62a5\u544a：<\/p>\n\n\u7c7b\u522bprecisionrecallf1-score\u6837\u672c\u6570<\/tr>\n\n\n\n\n\n\n
        0<\/td>\n0.94<\/td>\n0.95<\/td>\n0.95<\/td>\n100<\/td>\n<\/tr>\n
        1<\/td>\n0.93<\/td>\n0.91<\/td>\n0.92<\/td>\n100<\/td>\n<\/tr>\n
        2<\/td>\n0.94<\/td>\n0.91<\/td>\n0.92<\/td>\n100<\/td>\n<\/tr>\n
        3<\/td>\n0.93<\/td>\n0.97<\/td>\n0.95<\/td>\n100<\/td>\n<\/tr>\n
        macro avg<\/td>\n0.94<\/td>\n0.94<\/td>\n0.93<\/td>\n400<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

        \u56db\u4e2a\u7c7b\u522b\u8868\u73b0\u5747\u8861，\u6ca1\u6709\u51fa\u73b0\u67d0\u4e00\u7c7b\u88ab"\u504f\u7231"\u6216"\u5ffd\u7565"\u7684\u60c5\u51b5\u3002<\/p>\n

        \n

        \u4e5d\u3001\u5173\u952e\u77e5\u8bc6\u70b9\u603b\u7ed3<\/h3>\n
      • \u5206\u7c7b vs \u56de\u5f52：\u591a\u5206\u7c7b\u4efb\u52a1\u8981\u7528 CrossEntropyLoss + \u8f93\u51fa\u7ef4\u5ea6\u7b49\u4e8e\u7c7b\u522b\u6570，\u6807\u7b7e\u7528\u6574\u6570；\u4e0d\u8981\u7528 MSELoss（\u90a3\u662f\u56de\u5f52）\u3002<\/li>\n
      • \u8f93\u51fa\u5c42\u6fc0\u6d3b：CrossEntropyLoss \u5185\u7f6e softmax，\u8f93\u51fa\u5c42\u65e0\u9700\u624b\u52a8\u52a0\u6fc0\u6d3b\u51fd\u6570\u3002<\/li>\n
      • \u6807\u51c6\u5316：\u7279\u5f81\u91cf\u7eb2\u5dee\u5f02\u5927\u65f6\u5fc5\u987b\u6807\u51c6\u5316，\u4e14 fit \u53ea\u80fd\u7528\u8bad\u7ec3\u96c6\u3002<\/li>\n
      • \u5206\u5c42\u62bd\u6837：stratify=y \u4fdd\u8bc1\u7c7b\u522b\u5747\u8861\u5212\u5206\u3002<\/li>\n
      • \u6a21\u578b\u4fdd\u5b58\/\u52a0\u8f7d：torch.save(model.state_dict(), path) \u4fdd\u5b58，load_state_dict(torch.load(path)) \u52a0\u8f7d，\u52a0\u8f7d\u524d\u8981\u5148\u91cd\u5efa\u540c\u7ed3\u6784\u7f51\u7edc\u3002<\/li>\n
      • \u8bc4\u4f30\u6a21\u5f0f：\u63a8\u7406\u65f6 model.eval() + torch.no_grad()\u3002<\/li>\n
      • \u8def\u5f84\u81ea\u9002\u5e94：\u7528 os.path.dirname(os.path.abspath(__file__)) \u5b9a\u4f4d\u6587\u4ef6\u76ee\u5f55，\u907f\u514d\u76f8\u5bf9\u8def\u5f84\u4f9d\u8d56\u8fd0\u884c\u4f4d\u7f6e\u3002<\/li>\n
        \n

        \u5341\u3001\u7ed3\u8bed<\/h3>\n

        \u672c\u6587\u7528\u7ea6 160 \u884c\u4ee3\u7801\u5b8c\u6210\u4e86\u4e00\u4e2a\u624b\u673a\u4ef7\u683c\u56db\u5206\u7c7b\u4efb\u52a1，\u5b8c\u6574\u8986\u76d6\u4e86\u6570\u636e\u9884\u5904\u7406\u3001\u7f51\u7edc\u642d\u5efa\u3001\u8bad\u7ec3\u3001\u4fdd\u5b58\u3001\u52a0\u8f7d\u4e0e\u8bc4\u4f30\u3002\u4ee3\u7801\u7ed3\u6784\u6e05\u6670\u3001\u6ce8\u91ca\u9f50\u5168，\u53ef\u76f4\u63a5\u590d\u5236\u8fd0\u884c，\u4e5f\u53ef\u4ee5\u5728\u6b64\u57fa\u7840\u4e0a\u4fee\u6539\u9690\u85cf\u5c42\u7ef4\u5ea6\u3001\u4f18\u5316\u5668\u3001\u5b66\u4e60\u7387\u7b49\u8d85\u53c2\u6570\u505a\u5bf9\u6bd4\u5b9e\u9a8c\u3002<\/p>\n

        \u5982\u679c\u5bf9\u4f60\u6709\u5e2e\u52a9，\u6b22\u8fce\u70b9\u8d5e\u3001\u6536\u85cf\u3001\u8bc4\u8bba\u4ea4\u6d41 😊<\/p>\n","protected":false},"excerpt":{"rendered":"

        \u3010PyTorch\u3011\u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u5b9e\u73b0\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b（\u9644\u5b8c\u6574\u4ee3\u7801 \u8bad\u7ec3\u7ed3\u679c）\u672c\u6587\u7528 PyTorch \u642d\u5efa\u4e00\u4e2a\u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc，\u5bf9\u624b\u673a\u4ef7\u683c\u8fdb\u884c\u56db\u5206\u7c7b\u9884\u6d4b\u3002\u4ee3\u7801\u8986\u76d6\u4e86\u6570\u636e\u5904\u7406 \u2192 \u6a21\u578b\u8bad\u7ec3 \u2192 \u6a21\u578b\u8bc4\u4f30\u7684\u5b8c\u6574\u6d41\u7a0b，\u5e76\u628a\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u4fdd\u5b58\u5230\u672c\u5730\u3001\u8bc4\u4f30\u65f6\u518d\u52a0\u8f7d\u56de\u6765\u3002\u5168\u7a0b CPU \u5373\u53ef\u8fd0\u884c，\u9002\u5408\u5165\u95e8\u6559\u5b66\u3002\u4e00\u3001\u4efb\u52a1\u80cc\u666f
        \n\u7ed9\u5b9a 2000 \u6761\u624b\u673a\u914d\u7f6e\u6570\u636e（\u7535\u6c60\u3001\u5185\u5b58\u3001RAM\u3001\u6444\u50cf\u5934\u3001\u5c4f\u5e55\u7b49 20 \u4e2a\u7279\u5f81&<\/p>\n","protected":false},"author":2,"featured_media":97853,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[10187,2395,62,405,86],"topic":[],"class_list":["post-97854","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-codingplan","tag-2395","tag-62","tag-405","tag-86"],"yoast_head":"\n\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b - \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\/97854.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u3010PyTorch\u3011\u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u5b9e\u73b0\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b（\u9644\u5b8c\u6574\u4ee3\u7801 \u8bad\u7ec3\u7ed3\u679c）\u672c\u6587\u7528 PyTorch \u642d\u5efa\u4e00\u4e2a\u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc，\u5bf9\u624b\u673a\u4ef7\u683c\u8fdb\u884c\u56db\u5206\u7c7b\u9884\u6d4b\u3002\u4ee3\u7801\u8986\u76d6\u4e86\u6570\u636e\u5904\u7406 \u2192 \u6a21\u578b\u8bad\u7ec3 \u2192 \u6a21\u578b\u8bc4\u4f30\u7684\u5b8c\u6574\u6d41\u7a0b，\u5e76\u628a\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u4fdd\u5b58\u5230\u672c\u5730\u3001\u8bc4\u4f30\u65f6\u518d\u52a0\u8f7d\u56de\u6765\u3002\u5168\u7a0b CPU \u5373\u53ef\u8fd0\u884c，\u9002\u5408\u5165\u95e8\u6559\u5b66\u3002\u4e00\u3001\u4efb\u52a1\u80cc\u666f \u7ed9\u5b9a 2000 \u6761\u624b\u673a\u914d\u7f6e\u6570\u636e（\u7535\u6c60\u3001\u5185\u5b58\u3001RAM\u3001\u6444\u50cf\u5934\u3001\u5c4f\u5e55\u7b49 20 \u4e2a\u7279\u5f81&\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/97854.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-08-30T05:08:56+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/08\/20260830050855-6a93bae716155.png\" \/>\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=\"6 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/97854.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/97854.html\",\"name\":\"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-08-30T05:08:56+00:00\",\"dateModified\":\"2026-08-30T05:08:56+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/97854.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/97854.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/97854.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b\"}]},{\"@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":"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b - \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\/97854.html","og_locale":"zh_CN","og_type":"article","og_title":"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u3010PyTorch\u3011\u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u5b9e\u73b0\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b（\u9644\u5b8c\u6574\u4ee3\u7801 \u8bad\u7ec3\u7ed3\u679c）\u672c\u6587\u7528 PyTorch \u642d\u5efa\u4e00\u4e2a\u4e09\u5c42\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc，\u5bf9\u624b\u673a\u4ef7\u683c\u8fdb\u884c\u56db\u5206\u7c7b\u9884\u6d4b\u3002\u4ee3\u7801\u8986\u76d6\u4e86\u6570\u636e\u5904\u7406 \u2192 \u6a21\u578b\u8bad\u7ec3 \u2192 \u6a21\u578b\u8bc4\u4f30\u7684\u5b8c\u6574\u6d41\u7a0b，\u5e76\u628a\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u4fdd\u5b58\u5230\u672c\u5730\u3001\u8bc4\u4f30\u65f6\u518d\u52a0\u8f7d\u56de\u6765\u3002\u5168\u7a0b CPU \u5373\u53ef\u8fd0\u884c，\u9002\u5408\u5165\u95e8\u6559\u5b66\u3002\u4e00\u3001\u4efb\u52a1\u80cc\u666f \u7ed9\u5b9a 2000 \u6761\u624b\u673a\u914d\u7f6e\u6570\u636e（\u7535\u6c60\u3001\u5185\u5b58\u3001RAM\u3001\u6444\u50cf\u5934\u3001\u5c4f\u5e55\u7b49 20 \u4e2a\u7279\u5f81&","og_url":"https:\/\/www.wsisp.com\/helps\/97854.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-08-30T05:08:56+00:00","og_image":[{"url":"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/08\/20260830050855-6a93bae716155.png"}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"6 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/97854.html","url":"https:\/\/www.wsisp.com\/helps\/97854.html","name":"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-08-30T05:08:56+00:00","dateModified":"2026-08-30T05:08:56+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/97854.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/97854.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/97854.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"\u624b\u673a\u4ef7\u683c\u591a\u5206\u7c7b_\u5168\u8fde\u63a5\u795e\u7ecf\u7f51\u7edc\u6559\u7a0b"}]},{"@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\/97854","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=97854"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/97854\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media\/97853"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=97854"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=97854"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=97854"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=97854"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}