{"id":88325,"date":"2026-07-31T22:09:10","date_gmt":"2026-07-31T14:09:10","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/88325.html"},"modified":"2026-07-31T22:09:10","modified_gmt":"2026-07-31T14:09:10","slug":"epoch%e3%80%81batch-%e4%b8%8e-dataloader","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/88325.html","title":{"rendered":"Epoch\u3001Batch \u4e0e DataLoader"},"content":{"rendered":"

\u5f88\u591a\u521a\u5f00\u59cb\u9605\u8bfb PyTorch \u63a8\u8350\u7cfb\u7edf\u8bad\u7ec3\u4ee3\u7801\u7684\u540c\u5b66，\u90fd\u4f1a\u5361\u5728\u4e00\u7ec4\u540d\u8bcd：
\nepoch\u3001batch\u3001DataLoader\u3001shuffle\u3001loss.backward()\u3001optimizer.step()\u3001test\u3001HR\u3001NDCG\u3002<\/p>\n

\u5355\u72ec\u770b\u6bcf\u4e2a\u6982\u5ff5\u4e0d\u96be，\u4f46\u653e\u8fdb\u8bad\u7ec3\u5faa\u73af\u91cc，\u5f88\u5bb9\u6613\u5206\u4e0d\u6e05\u6574\u6761\u6d41\u6c34\u7ebf：
\n\u6570\u636e\u4ece\u54ea\u91cc\u8bfb\u53d6？\u5982\u4f55\u5207\u5206\u6210 batch？\u6a21\u578b\u4ec0\u4e48\u65f6\u5019\u66f4\u65b0\u53c2\u6570？\u6d4b\u8bd5\u9636\u6bb5\u4f1a\u4e0d\u4f1a\u5077\u5077\u4fee\u6539\u6743\u91cd？<\/p>\n

\u672c\u6587\u7528\u6700\u7b80\u5355\u7684\u6570\u5b57\u6837\u672c，\u5b8c\u6574\u62c6\u89e3\u6574\u6761\u8bad\u7ec3\u94fe\u8def\u3002
\n\"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\"<\/p>\n

\u2705 \u6838\u5fc3\u7ed3\u8bba<\/h3>\n
  • \u4e00\u4e2a epoch = \u5b8c\u6574\u904d\u5386\u4e00\u904d\u5168\u90e8\u8bad\u7ec3\u6570\u636e\u96c6\u3002<\/li>\n
  • \u5355\u4e2a epoch \u5185\u90e8，\u5faa\u73af\u8bfb\u53d6\u591a\u4e2a batch\u3002<\/li>\n
  • \u8bad\u7ec3\u9636\u6bb5，\u6bcf\u62ff\u5230\u4e00\u4e2a batch，\u6807\u51c6\u6d41\u7a0b：
    \n\u524d\u5411\u4f20\u64ad \u2192 \u8ba1\u7b97\u635f\u5931 \u2192 \u6e05\u7a7a\u68af\u5ea6 \u2192 \u53cd\u5411\u4f20\u64ad \u2192 \u66f4\u65b0\u6a21\u578b\u53c2\u6570\u3002<\/li>\n
  • verbose \u7528\u6765\u63a7\u5236：\u6bcf\u9694\u591a\u5c11\u4e2a epoch，\u6267\u884c\u65e5\u5fd7\u6253\u5370\/\u79bb\u7ebf\u8bc4\u4f30\u3002<\/li>\n
  • \u6d4b\u8bd5\u8bc4\u4f30\u9636\u6bb5：\u4ec5\u505a\u9884\u6d4b\u3001\u8ba1\u7b97\u6307\u6807（HR@K\u3001NDCG@K），\u4e0d\u4f1a\u66f4\u65b0\u6a21\u578b\u53c2\u6570\u3002<\/li>\n
  • shuffle=True \u53ea\u6539\u53d8\u6837\u672c\u8bfb\u53d6\u987a\u5e8f，\u4e0d\u4f1a\u4fee\u6539 Dataset \u5185\u90e8\u539f\u59cb\u6570\u636e\u3002<\/li>\n
  • shuffle=True：\u6bcf\u4e00\u8f6eepoch\u5f00\u59cb，\u91cd\u65b0\u751f\u6210\u968f\u673a\u8bfb\u53d6\u987a\u5e8f\u3002<\/li>\n
  • shuffle=False：\u6240\u6709epoch\u8bfb\u53d6\u6837\u672c\u987a\u5e8f\u5b8c\u5168\u56fa\u5b9a\u3002<\/li>\n

    \u4e00\u53e5\u8bdd\u6982\u62ec\u5206\u5de5：<\/p>\n

    Dataset \u7ba1\u7406\u300c\u6709\u54ea\u4e9b\u6570\u636e\u300d；DataLoader \u7ba1\u7406\u300c\u4ee5\u4ec0\u4e48\u987a\u5e8f\u3001\u591a\u5c11\u6761\u4e00\u7ec4\u53d6\u51fa\u6570\u636e\u300d；\u8bad\u7ec3\u5faa\u73af\u63a7\u5236\u300c\u62ff\u5230batch\u540e\u5982\u4f55\u8fed\u4ee3\u4f18\u5316\u6a21\u578b\u300d\u3002<\/p>\n

    \n

    🔢 \u6781\u7b80\u6570\u5b57\u6570\u636e\u96c6\u6f14\u793a<\/h3>\n

    \u5047\u8bbe\u6211\u4eec\u4e00\u5171\u670910\u6761\u8bad\u7ec3\u6837\u672c，\u4f7f\u7528\u7d22\u5f15\u7f16\u53f7\u4ee3\u8868\u6837\u672c：<\/p>\n

    \u539f\u59cb\u6570\u636e\u96c6\u7d22\u5f15：[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]<\/p>\n

    \u53c2\u6570\u8bbe\u7f6e：<\/p>\n

    batch_size =<\/span> 4<\/span><\/p>\n

    \u542b\u4e49：\u4e00\u6b21\u6027\u53d6\u51fa4\u6761\u6837\u672c，\u5c01\u88c5\u4e3a1\u4e2abatch\u3002<\/p>\n

    \u573a\u666f1：shuffle=False（\u4e0d\u6253\u4e71\u987a\u5e8f）<\/h4>\n

    \"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\"<\/p>\n

    \u6570\u636e\u4f1a\u6309\u539f\u751f\u7d22\u5f15\u5207\u5206batch：<\/p>\n

    \u7b2c1\u4e2a batch: [0, 1, 2, 3]
    \n\u7b2c2\u4e2a batch: [4, 5, 6, 7]
    \n\u7b2c3\u4e2a batch: [8, 9]<\/p>\n

    \u603b\u657010\u65e0\u6cd5\u88ab4\u6574\u9664，\u6700\u540e\u4e00\u7ec4\u4e3a\u4e0d\u5b8c\u6574batch\u3002
    \n\u5982\u679c\u5f00\u542f drop_last=True，\u6700\u540e\u8fd9\u4e2a\u4e0d\u5b8c\u6574batch [8, 9] \u4f1a\u76f4\u63a5\u4e22\u5f03\u3002<\/p>\n

    shuffle=False：\u6240\u6709epoch\u987a\u5e8f\u56fa\u5b9a
    \n\u8bfb\u53d6\u7d22\u5f15\u6c38\u8fdc\u4e0d\u53d8，\u8fde\u7eed\u4e24\u8f6eepoch\u6279\u6b21\u5b8c\u5168\u76f8\u540c：<\/p>\n

    epoch 1:
    \nbatch 1: [0, 1, 2, 3]
    \nbatch 2: [4, 5, 6, 7]
    \nbatch 3: [8, 9]<\/p>\n

    epoch 2:
    \nbatch 1: [0, 1, 2, 3]
    \nbatch 2: [4, 5, 6, 7]
    \nbatch 3: [8, 9]<\/p>\n

    \u573a\u666f2：shuffle=True（\u5f00\u542f\u6253\u4e71）<\/h4>\n

    \"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\"<\/p>\n

    \u2757\u91cd\u70b9：\u6253\u4e71\u7684\u662f\u7d22\u5f15\u8bfb\u53d6\u987a\u5e8f，\u539f\u59cbDataset\u5185\u7684\u6570\u636e\u672c\u8eab\u4fdd\u6301\u4e0d\u53d8\u3002
    \n\u539f\u59cb\u6570\u636e\u4f9d\u65e7\u662f [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]<\/p>\n

    \u5173\u952e\u70b9\u603b\u7ed3：
    \nshuffle=True \u2260 \u6c38\u4e45\u4fee\u6539\u6570\u636e；
    \n\u6bcf\u4e00\u8f6eepoch\u542f\u52a8\u65f6，DataLoader\u91cd\u65b0\u89c4\u5212\u672c\u8f6e\u6837\u672c\u8bfb\u53d6\u6b21\u5e8f\u3002<\/p>\n

    \n

    📦 DataLoader \u5230\u5e95\u5982\u4f55\u751f\u6210\u4e00\u4e2a batch？<\/h3>\n

    \u8fd9\u662f\u65b0\u624b\u6700\u5bb9\u6613\u6df7\u6dc6\u7684\u73af\u8282\u3002\u6211\u4eec\u62c6\u6210\u5b8c\u65745\u6b65\u7406\u89e3\u3002<\/p>\n

    \u81ea\u5b9a\u4e49\u6570\u636e\u96c6\u6a21\u677f：<\/p>\n

    class<\/span> NumberDataset<\/span>(<\/span>torch.<\/span>utils.<\/span>data.<\/span>Dataset)<\/span>:<\/span>
    \n def<\/span> __init__<\/span>(<\/span>self)<\/span>:<\/span>
    \n self.<\/span>samples =<\/span> [<\/span>
    \n {<\/span>"idx"<\/span>:<\/span> 0<\/span>,<\/span> "feature"<\/span>:<\/span> [<\/span>0.0<\/span>,<\/span> 0.5<\/span>]<\/span>,<\/span> "label"<\/span>:<\/span> 0<\/span>}<\/span>,<\/span>
    \n {<\/span>"idx"<\/span>:<\/span> 1<\/span>,<\/span> "feature"<\/span>:<\/span> [<\/span>1.0<\/span>,<\/span> 1.5<\/span>]<\/span>,<\/span> "label"<\/span>:<\/span> 1<\/span>}<\/span>,<\/span>
    \n {<\/span>"idx"<\/span>:<\/span> 2<\/span>,<\/span> "feature"<\/span>:<\/span> [<\/span>2.0<\/span>,<\/span> 2.5<\/span>]<\/span>,<\/span> "label"<\/span>:<\/span> 0<\/span>}<\/span>,<\/span>
    \n # …\u66f4\u591a\u6837\u672c<\/span>
    \n ]<\/span><\/p>\n

    def<\/span> __len__<\/span>(<\/span>self)<\/span>:<\/span>
    \n return<\/span> len<\/span>(<\/span>self.<\/span>samples)<\/span><\/p>\n

    def<\/span> __getitem__<\/span>(<\/span>self,<\/span> index)<\/span>:<\/span>
    \n # \u6839\u636e\u7d22\u5f15\u8fd4\u56de\u3010\u5355\u6761\u6837\u672c\u3011<\/span>
    \n return<\/span> self.<\/span>samples[<\/span>index]<\/span><\/p>\n

    \u8bad\u7ec3\u5faa\u73af\u4ee3\u7801：<\/p>\n

    for<\/span> batch in<\/span> data_loader:<\/span>
    \n .<\/span>.<\/span>.<\/span><\/p>\n

    \u5faa\u73af\u80cc\u540e\u5b8c\u6574\u6d41\u7a0b：<\/p>\n

    \u7b2c1\u6b65：\u751f\u6210\u7d22\u5f15\u987a\u5e8f<\/h4>\n
      \n
    • shuffle=False：\u539f\u751f\u7d22\u5f15 [0,1,2,3,…]<\/li>\n
    • shuffle=True：\u968f\u673a\u6253\u4e71\u7d22\u5f15\u5e8f\u5217<\/li>\n<\/ul>\n

      \u6ce8\u610f：\u6570\u5b57\u53ea\u662f\u6837\u672c\u7d22\u5f15，\u4e0d\u662f\u6837\u672c\u5185\u5bb9\u3002\u540e\u7eed\u4f1a\u8c03\u7528 dataset[index] \u83b7\u53d6\u5355\u6761\u6570\u636e\u3002<\/p>\n

      \u7b2c2\u6b65：\u6309batch_size\u5206\u7ec4\u7d22\u5f15<\/h4>\n

      \u7d22\u5f15\u5e8f\u5217 [3, 7, 1, 9, 0, 6, 2, 8, 4, 5]
      \nbatch_size=4 \u2192 \u5206\u7ec4：
      \n[3,7,1,9]\u3001[0,6,2,8]\u3001[4,5]
      \n\u5f53\u524d\u4ec5\u89c4\u5212\u7d22\u5f15，\u8fd8\u6ca1\u6709\u8bfb\u53d6\u771f\u5b9e\u6837\u672c\u3002<\/p>\n

      \u7b2c3\u6b65：\u9010\u6761\u8c03\u7528 Dataset.__getitem__<\/h4>\n

      \u53d6\u7d22\u5f15\u7ec4 [3,7,1,9]
      \n\u4f9d\u6b21\u6267\u884c：<\/p>\n

      sample_1 =<\/span> dataset[<\/span>3<\/span>]<\/span>
      \nsample_2 =<\/span> dataset[<\/span>7<\/span>]<\/span>
      \nsample_3 =<\/span> dataset[<\/span>1<\/span>]<\/span>
      \nsample_4 =<\/span> dataset[<\/span>9<\/span>]<\/span><\/p>\n

      \u5f97\u52304\u6761\u72ec\u7acb\u6837\u672c\u5b57\u5178：<\/p>\n

      [<\/span>
      \n {<\/span>"idx"<\/span>:<\/span> 3<\/span>,<\/span> "feature"<\/span>:<\/span> [<\/span>3.0<\/span>,<\/span> 3.5<\/span>]<\/span>,<\/span> "label"<\/span>:<\/span> 1<\/span>}<\/span>,<\/span>
      \n {<\/span>"idx"<\/span>:<\/span> 7<\/span>,<\/span> "feature"<\/span>:<\/span> [<\/span>7.0<\/span>,<\/span> 7.5<\/span>]<\/span>,<\/span> "label"<\/span>:<\/span> 1<\/span>}<\/span>,<\/span>
      \n {<\/span>"idx"<\/span>:<\/span> 1<\/span>,<\/span> "feature"<\/span>:<\/span> [<\/span>1.0<\/span>,<\/span> 1.5<\/span>]<\/span>,<\/span> "label"<\/span>:<\/span> 1<\/span>}<\/span>,<\/span>
      \n {<\/span>"idx"<\/span>:<\/span> 9<\/span>,<\/span> "feature"<\/span>:<\/span> [<\/span>9.0<\/span>,<\/span> 9.5<\/span>]<\/span>,<\/span> "label"<\/span>:<\/span> 1<\/span>}<\/span>,<\/span>
      \n]<\/span><\/p>\n

      \u7b2c4\u6b65：collate_fn \u6253\u5305，\u62fc\u63a5\u6210batch\u5f20\u91cf<\/h4>\n

      \"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\"
      \n\u9ed8\u8ba4 default_collate \u8d1f\u8d23\u6253\u5305：\u628a\u591a\u6761\u6837\u672c\u540c\u5b57\u6bb5\u5806\u53e0\u3002
      \n\u6253\u5305\u5b8c\u6210\u540e：<\/p>\n

      {<\/span>
      \n "idx"<\/span>:<\/span> tensor(<\/span>[<\/span>3<\/span>,<\/span> 7<\/span>,<\/span> 1<\/span>,<\/span> 9<\/span>]<\/span>)<\/span>,<\/span>
      \n "feature"<\/span>:<\/span> tensor(<\/span>[<\/span>
      \n [<\/span>3.0<\/span>,<\/span> 3.5<\/span>]<\/span>,<\/span>
      \n [<\/span>7.0<\/span>,<\/span> 7.5<\/span>]<\/span>,<\/span>
      \n [<\/span>1.0<\/span>,<\/span> 1.5<\/span>]<\/span>,<\/span>
      \n [<\/span>9.0<\/span>,<\/span> 9.5<\/span>]<\/span>,<\/span>
      \n ]<\/span>)<\/span>,<\/span>
      \n "label"<\/span>:<\/span> tensor(<\/span>[<\/span>1<\/span>,<\/span> 1<\/span>,<\/span> 1<\/span>,<\/span> 1<\/span>]<\/span>)<\/span>,<\/span>
      \n}<\/span><\/p>\n

      \u8fd9\u5c31\u662f\u8bad\u7ec3\u5faa\u73af\u62ff\u5230\u7684 batch\u3002<\/p>\n

      \u901a\u4fd7\u7406\u89e3：
      \nDataset \u4ea7\u51fa\u4e00\u6761\u6761\u72ec\u7acb\u6837\u672c；collate_fn \u6253\u5305\u5458，\u628a\u591a\u6761\u6837\u672c\u7ec4\u88c5\u6210\u4e00\u4e2abatch\u5f20\u91cf\u3002<\/p>\n

      \u7b2c5\u6b65：\u8bad\u7ec3\u5faa\u73af\u63a5\u6536batch，\u9001\u5165\u6a21\u578b<\/h4>\n

      \u5916\u90e8\u4ee3\u7801\u76f4\u63a5\u4f7f\u7528：<\/p>\n

      for<\/span> batch in<\/span> train_loader:<\/span>
      \n features =<\/span> batch[<\/span>"feature"<\/span>]<\/span>
      \n labels =<\/span> batch[<\/span>"label"<\/span>]<\/span>
      \n scores =<\/span> model(<\/span>features)<\/span><\/p>\n

      \u6574\u6761\u94fe\u8def\u7b80\u5316：
      \n\u751f\u6210\u7d22\u5f15\u987a\u5e8f \u2192 \u7d22\u5f15\u5206\u7ec4 \u2192 __getitem__\u9010\u6761\u8bfb\u53d6\u6837\u672c \u2192 collate_fn\u62fc\u63a5batch \u2192 \u8fd4\u56de\u5faa\u73af<\/p>\n

      \n

      🔁 \u5b8c\u6574\u8bad\u7ec3\u5faa\u73af\u6267\u884c\u6d41\u7a0b<\/h3>\n

      \u793a\u4f8b\u6807\u51c6\u8bad\u7ec3\u4ee3\u7801：<\/p>\n

      epochs =<\/span> 3<\/span>
      \nbatch_size =<\/span> 4<\/span>
      \nverbose =<\/span> 2<\/span><\/p>\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 # \u5185\u5c42\u5faa\u73af\u904d\u5386\u6240\u6709batch<\/span>
      \n for<\/span> batch in<\/span> train_loader:<\/span>
      \n features =<\/span> batch[<\/span>"feature"<\/span>]<\/span>
      \n labels =<\/span> batch[<\/span>"label"<\/span>]<\/span>
      \n scores =<\/span> model(<\/span>features)<\/span>
      \n loss =<\/span> loss_fn(<\/span>scores,<\/span> labels)<\/span><\/p>\n

      optimizer.<\/span>zero_grad(<\/span>)<\/span>
      \n loss.<\/span>backward(<\/span>)<\/span>
      \n optimizer.<\/span>step(<\/span>)<\/span><\/p>\n

      # \u95f4\u9694verbose\u4e2aepoch\u6267\u884c\u6d4b\u8bd5<\/span>
      \n if<\/span> epoch %<\/span> verbose ==<\/span> 0<\/span>:<\/span>
      \n evaluate(<\/span>model,<\/span> test_loader)<\/span><\/p>\n

      \u6d41\u7a0b\u7ffb\u8bd1：<\/p>\n

        \n
      • Epoch 1：\u5b8c\u6574\u904d\u5386\u8bad\u7ec3\u96c6，\u6240\u6709batch\u66f4\u65b0\u53c2\u6570；1%2\u22600，\u4e0d\u6d4b\u8bd5<\/li>\n
      • Epoch 2：\u5b8c\u6574\u904d\u5386\u8bad\u7ec3\u96c6，\u6240\u6709batch\u66f4\u65b0\u53c2\u6570；2%2=0，\u6267\u884c\u6d4b\u8bd5\u8bc4\u4f30<\/li>\n
      • Epoch 3：\u5b8c\u6574\u904d\u5386\u8bad\u7ec3\u96c6，\u6240\u6709batch\u66f4\u65b0\u53c2\u6570；3%2\u22600，\u4e0d\u6d4b\u8bd5\"\u5728\u8fd9\u91cc\u63d2\u5165\u56fe\u7247\u63cf\u8ff0\"<\/li>\n<\/ul>\n

        \u5355\u4e2abatch\u5185\u90e8\u6807\u51c6\u8bad\u7ec3\u95ed\u73af<\/h4>\n

        model.<\/span>train(<\/span>)<\/span> # \u5207\u6362\u8bad\u7ec3\u6a21\u5f0f(Dropout\/BatchNorm\u751f\u6548)<\/span>
        \nscores =<\/span> model(<\/span>features)<\/span># \u524d\u5411\u4f20\u64ad，\u5f97\u5230\u9884\u6d4b\u503c<\/span>
        \nloss =<\/span> loss_fn(<\/span>scores,<\/span> labels)<\/span> # \u8ba1\u7b97\u635f\u5931<\/span>
        \noptimizer.<\/span>zero_grad(<\/span>)<\/span> # \u6e05\u7a7a\u4e0a\u4e00\u8f6e\u68af\u5ea6<\/span>
        \nloss.<\/span>backward(<\/span>)<\/span> # \u53cd\u5411\u4f20\u64ad，\u8ba1\u7b97\u53c2\u6570\u68af\u5ea6<\/span>
        \noptimizer.<\/span>step(<\/span>)<\/span> # \u4f7f\u7528\u68af\u5ea6\u66f4\u65b0\u6a21\u578b\u6743\u91cd<\/span><\/p>\n

        \n

        \u2696\ufe0f \u8bad\u7ec3\u9636\u6bb5 VS \u6d4b\u8bd5\u8bc4\u4f30\u9636\u6bb5<\/h3>\n\n\u9636\u6bb5\u662f\u5426\u8ba1\u7b97\u68af\u5ea6\u662f\u5426\u53cd\u5411\u4f20\u64ad\u662f\u5426\u66f4\u65b0\u53c2\u6570\u6838\u5fc3\u76ee\u7684<\/tr>\n\n\n\n
        \u8bad\u7ec3\u9636\u6bb5<\/td>\n\u2705 \u662f<\/td>\n\u2705 \u662f<\/td>\n\u2705 \u662f<\/td>\n\u8fed\u4ee3\u4f18\u5316\u6a21\u578b<\/td>\n<\/tr>\n
        \u6d4b\u8bd5\u8bc4\u4f30<\/td>\n\u274c \u5426<\/td>\n\u274c \u5426<\/td>\n\u274c \u5426<\/td>\n\u89c2\u6d4b\u6a21\u578b\u6cdb\u5316\u6548\u679c<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

        \u8bad\u7ec3\u4ee3\u7801\u6a21\u677f<\/h4>\n

        model.<\/span>train(<\/span>)<\/span>
        \nfor<\/span> batch in<\/span> train_loader:<\/span>
        \n scores =<\/span> model(<\/span>batch[<\/span>"feature"<\/span>]<\/span>)<\/span>
        \n loss =<\/span> loss_fn(<\/span>scores,<\/span> batch[<\/span>"label"<\/span>]<\/span>)<\/span>
        \n optimizer.<\/span>zero_grad(<\/span>)<\/span>
        \n loss.<\/span>backward(<\/span>)<\/span>
        \n optimizer.<\/span>step(<\/span>)<\/span><\/p>\n

        \u6d4b\u8bd5\u8bc4\u4f30\u6a21\u677f<\/h4>\n

        model.<\/span>eval<\/span>(<\/span>)<\/span>
        \nwith<\/span> torch.<\/span>no_grad(<\/span>)<\/span>:<\/span> # \u5173\u95ed\u68af\u5ea6\u8ba1\u7b97，\u8282\u7701\u663e\u5b58<\/span>
        \n for<\/span> batch in<\/span> test_loader:<\/span>
        \n scores =<\/span> model(<\/span>batch[<\/span>"feature"<\/span>]<\/span>)<\/span>
        \n # \u6839\u636e\u9884\u6d4b\u5206\u6570\u8ba1\u7b97HR\u3001NDCG\u6307\u6807<\/span><\/p>\n

        torch.no_grad()：\u544a\u77e5PyTorch\u4ec5\u63a8\u7406，\u4e0d\u9700\u8981\u6784\u5efa\u68af\u5ea6\u8ba1\u7b97\u56fe\u3002<\/p>\n

        \n

        📊 \u63a8\u8350\u7cfb\u7edf\u6307\u6807：HR@K \u548c NDCG@K<\/h3>\n

        \u63a8\u8350\u4efb\u52a1\u79bb\u7ebf\u8bc4\u4f30\u6700\u5e38\u7528\u4e24\u4e2a\u6307\u6807：<\/p>\n

        HR@K（Hit Ratio，\u547d\u4e2d\u7387）<\/h4>\n

        \u542b\u4e49：\u7ed9\u7528\u6237\u63a8\u8350Top-K\u7269\u54c1，\u771f\u5b9e\u4ea4\u4e92\u7269\u54c1\u662f\u5426\u51fa\u73b0\u5728\u63a8\u8350\u5217\u8868\u5185\u3002<\/p>\n

          \n
        • \u547d\u4e2d：HR=1<\/li>\n
        • \u672a\u547d\u4e2d：HR=0<\/li>\n<\/ul>\n

          \u793a\u4f8b：
          \n\u63a8\u8350Top5\u5217\u8868：[item_8, item_2, item_6, item_1, item_9]
          \n\u7528\u6237\u771f\u5b9e\u559c\u7231\u7269\u54c1：item_6
          \nitem_6 \u5728\u5217\u8868\u4e2d \u2192 HR@5 = 1\u3002<\/p>\n

          NDCG@K<\/h4>\n

          HR\u53ea\u5173\u5fc3\u300c\u6709\u6ca1\u6709\u547d\u4e2d\u300d；NDCG\u989d\u5916\u5173\u6ce8\u547d\u4e2d\u7269\u54c1\u7684\u6392\u5e8f\u4f4d\u7f6e\u3002
          \n\u6848\u4f8b：\u771f\u5b9e\u7269\u54c1\u4e3a item_6<\/p>\n

            \n
          • \u63a8\u8350A：[item_6, item_2, item_8, item_1, item_9]（\u547d\u4e2d\u6392\u5728\u7b2c1\u4f4d）<\/li>\n
          • \u63a8\u8350B：[item_8, item_2, item_1, item_9, item_6]（\u547d\u4e2d\u6392\u5728\u7b2c5\u4f4d）<\/li>\n<\/ul>\n

            \u4e24\u8005HR\u90fd\u7b49\u4e8e1，\u4f46\u63a8\u8350A\u7684NDCG\u66f4\u9ad8\u3002
            \n\u7b80\u5355\u7406\u89e3：<\/p>\n

            HR：\u662f\u5426\u731c\u4e2d；NDCG：\u731c\u4e2d\u4e4b\u540e，\u6392\u5f97\u591f\u4e0d\u591f\u9760\u524d\u3002<\/p>\n

            \n

            🛠\ufe0f \u53ef\u76f4\u63a5\u8fd0\u884c\u7684\u6f14\u793a\u4ee3\u7801<\/h3>\n

            \u590d\u5236\u8fd0\u884c，\u76f4\u89c2\u89c2\u5bdf __getitem__ \u7684\u8c03\u7528\u65f6\u673a：<\/p>\n

            import<\/span> torch
            \nfrom<\/span> torch.<\/span>utils.<\/span>data import<\/span> Dataset,<\/span> DataLoader<\/p>\n

            class<\/span> NumberDataset<\/span>(<\/span>Dataset)<\/span>:<\/span>
            \n def<\/span> __init__<\/span>(<\/span>self)<\/span>:<\/span>
            \n self.<\/span>samples =<\/span> [<\/span>]<\/span>
            \n for<\/span> idx in<\/span> range<\/span>(<\/span>10<\/span>)<\/span>:<\/span>
            \n self.<\/span>samples.<\/span>append(<\/span>{<\/span>
            \n "idx"<\/span>:<\/span> idx,<\/span>
            \n "feature"<\/span>:<\/span> torch.<\/span>tensor(<\/span>[<\/span>float<\/span>(<\/span>idx)<\/span>,<\/span> float<\/span>(<\/span>idx)<\/span> +<\/span> 0.5<\/span>]<\/span>)<\/span>,<\/span>
            \n "label"<\/span>:<\/span> torch.<\/span>tensor(<\/span>idx %<\/span> 2<\/span>,<\/span> dtype=<\/span>torch.<\/span>float32)<\/span>,<\/span>
            \n }<\/span>)<\/span><\/p>\n

            def<\/span> __len__<\/span>(<\/span>self)<\/span>:<\/span>
            \n return<\/span> len<\/span>(<\/span>self.<\/span>samples)<\/span><\/p>\n

            def<\/span> __getitem__<\/span>(<\/span>self,<\/span> index)<\/span>:<\/span>
            \n print<\/span>(<\/span>f" __getitem__ \u88ab\u8c03\u7528，index=<\/span>{<\/span>index}<\/span><\/span>"<\/span><\/span>)<\/span>
            \n return<\/span> self.<\/span>samples[<\/span>index]<\/span><\/p>\n

            dataset =<\/span> NumberDataset(<\/span>)<\/span>
            \nloader =<\/span> DataLoader(<\/span>dataset,<\/span> batch_size=<\/span>4<\/span>,<\/span> shuffle=<\/span>False<\/span>)<\/span><\/p>\n

            for<\/span> batch_id,<\/span> batch in<\/span> enumerate<\/span>(<\/span>loader,<\/span> start=<\/span>1<\/span>)<\/span>:<\/span>
            \n print<\/span>(<\/span>f"\\\\n\u7b2c <\/span>{<\/span>batch_id}<\/span><\/span> \u4e2a batch"<\/span><\/span>)<\/span>
            \n print<\/span>(<\/span>"idx:"<\/span>,<\/span> batch[<\/span>"idx"<\/span>]<\/span>)<\/span>
            \n print<\/span>(<\/span>"feature:"<\/span>,<\/span> batch[<\/span>"feature"<\/span>]<\/span>)<\/span>
            \n print<\/span>(<\/span>"label:"<\/span>,<\/span> batch[<\/span>"label"<\/span>]<\/span>)<\/span><\/p>\n

            \u8fd0\u884c\u8f93\u51fa\u53ef\u4ee5\u9a8c\u8bc1：<\/p>\n

          • DataLoader \u9010\u4e2a\u8c03\u7528 __getitem__ \u83b7\u53d6\u5355\u6761\u6837\u672c<\/li>\n
          • \u6536\u96c6\u8db3\u591f\u6837\u672c\u540e，collate_fn\u81ea\u52a8\u5408\u5e76\u4e3abatch\u5f20\u91cf<\/li>\n
          • \u8bad\u7ec3\u5faa\u73af\u62ff\u5230\u7684\u662f\u5c01\u88c5\u5b8c\u6210\u7684\u6279\u91cf\u6570\u636e，\u800c\u975e\u5355\u6837\u672c<\/li>\n
            \n

            💡 \u6536\u5c3e\u603b\u7ed3<\/h3>\n

            \u770b\u5230\u63a8\u8350\u7cfb\u7edf\u7ecf\u5178\u53cc\u5c42\u5faa\u73af\u4ee3\u7801：<\/p>\n

            for<\/span> epoch in<\/span> range<\/span>(<\/span>epochs)<\/span>:<\/span>
            \n for<\/span> batch in<\/span> train_loader:<\/span>
            \n .<\/span>.<\/span>.<\/span><\/p>\n

            \u8111\u6d77\u4e2d\u81ea\u52a8\u7ffb\u8bd1：<\/p>\n

            \u5916\u5c42epoch\u5faa\u73af：\u63a7\u5236\u4e00\u5171\u5b8c\u6574\u8bad\u7ec3\u591a\u5c11\u8f6e；
            \n\u5185\u5c42batch\u5faa\u73af：\u4e00\u8f6e\u8bad\u7ec3\u4e2d\u5206\u6279\u8bfb\u53d6\u6570\u636e；
            \n\u6bcf\u4e00\u4e2a\u8bad\u7ec3batch\u90fd\u4f1a\u66f4\u65b0\u6a21\u578b\u53c2\u6570；
            \n\u5230\u8fbe\u6307\u5b9aepoch\u95f4\u9694，\u8fdb\u5165\u8bc4\u4f30\u6a21\u5f0f，\u53ea\u9884\u6d4b\u3001\u8ba1\u7b97\u6307\u6807，\u4e0d\u66f4\u65b0\u6743\u91cd\u3002<\/p>\n

            \u6253\u901a\u8fd9\u5957\u6d41\u7a0b\u540e，\u518d\u9605\u8bfb BPR\u3001NeuMF\u3001LightGCN \u7b49\u63a8\u8350\u6a21\u578b\u8bad\u7ec3\u4ee3\u7801，\u7406\u89e3\u95e8\u69db\u4f1a\u5927\u5e45\u964d\u4f4e\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

            \u5f88\u591a\u521a\u5f00\u59cb\u9605\u8bfb PyTorch \u63a8\u8350\u7cfb\u7edf\u8bad\u7ec3\u4ee3\u7801\u7684\u540c\u5b66，\u90fd\u4f1a\u5361\u5728\u4e00\u7ec4\u540d\u8bcd：
            \nepoch\u3001batch\u3001DataLoader\u3001shuffle\u3001loss.backward()\u3001optimizer.step()\u3001test\u3001HR\u3001NDCG\u3002
            \n\u5355\u72ec\u770b\u6bcf\u4e2a\u6982\u5ff5\u4e0d\u96be，\u4f46\u653e\u8fdb\u8bad\u7ec3\u5faa\u73af\u91cc，\u5f88\u5bb9\u6613\u5206\u4e0d\u6e05\u6574\u6761\u6d41\u6c34\u7ebf：
            \n\u6570\u636e\u4ece\u54ea\u91cc\u8bfb\u53d6？\u5982\u4f55\u5207\u5206\u6210 batch？\u6a21\u578b\u4ec0\u4e48\u65f6\u5019\u66f4\u65b0\u53c2\u6570？\u6d4b\u8bd5\u9636\u6bb5\u4f1a<\/p>\n","protected":false},"author":2,"featured_media":88320,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[7698,190,86],"topic":[],"class_list":["post-88325","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-batch","tag-190","tag-86"],"yoast_head":"\nEpoch\u3001Batch \u4e0e DataLoader - \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\/88325.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Epoch\u3001Batch \u4e0e DataLoader - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"\u5f88\u591a\u521a\u5f00\u59cb\u9605\u8bfb PyTorch \u63a8\u8350\u7cfb\u7edf\u8bad\u7ec3\u4ee3\u7801\u7684\u540c\u5b66，\u90fd\u4f1a\u5361\u5728\u4e00\u7ec4\u540d\u8bcd： epoch\u3001batch\u3001DataLoader\u3001shuffle\u3001loss.backward()\u3001optimizer.step()\u3001test\u3001HR\u3001NDCG\u3002 \u5355\u72ec\u770b\u6bcf\u4e2a\u6982\u5ff5\u4e0d\u96be，\u4f46\u653e\u8fdb\u8bad\u7ec3\u5faa\u73af\u91cc，\u5f88\u5bb9\u6613\u5206\u4e0d\u6e05\u6574\u6761\u6d41\u6c34\u7ebf： \u6570\u636e\u4ece\u54ea\u91cc\u8bfb\u53d6？\u5982\u4f55\u5207\u5206\u6210 batch？\u6a21\u578b\u4ec0\u4e48\u65f6\u5019\u66f4\u65b0\u53c2\u6570？\u6d4b\u8bd5\u9636\u6bb5\u4f1a\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/88325.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-07-31T14:09:10+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/07\/20260731140902-6a6cac7eb8b84.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=\"4 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/88325.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/88325.html\",\"name\":\"Epoch\u3001Batch \u4e0e DataLoader - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-07-31T14:09:10+00:00\",\"dateModified\":\"2026-07-31T14:09:10+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/88325.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/88325.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/88325.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Epoch\u3001Batch \u4e0e DataLoader\"}]},{\"@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":"Epoch\u3001Batch \u4e0e DataLoader - \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\/88325.html","og_locale":"zh_CN","og_type":"article","og_title":"Epoch\u3001Batch \u4e0e DataLoader - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"\u5f88\u591a\u521a\u5f00\u59cb\u9605\u8bfb PyTorch \u63a8\u8350\u7cfb\u7edf\u8bad\u7ec3\u4ee3\u7801\u7684\u540c\u5b66，\u90fd\u4f1a\u5361\u5728\u4e00\u7ec4\u540d\u8bcd： epoch\u3001batch\u3001DataLoader\u3001shuffle\u3001loss.backward()\u3001optimizer.step()\u3001test\u3001HR\u3001NDCG\u3002 \u5355\u72ec\u770b\u6bcf\u4e2a\u6982\u5ff5\u4e0d\u96be，\u4f46\u653e\u8fdb\u8bad\u7ec3\u5faa\u73af\u91cc，\u5f88\u5bb9\u6613\u5206\u4e0d\u6e05\u6574\u6761\u6d41\u6c34\u7ebf： \u6570\u636e\u4ece\u54ea\u91cc\u8bfb\u53d6？\u5982\u4f55\u5207\u5206\u6210 batch？\u6a21\u578b\u4ec0\u4e48\u65f6\u5019\u66f4\u65b0\u53c2\u6570？\u6d4b\u8bd5\u9636\u6bb5\u4f1a","og_url":"https:\/\/www.wsisp.com\/helps\/88325.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-07-31T14:09:10+00:00","og_image":[{"url":"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/07\/20260731140902-6a6cac7eb8b84.png"}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"4 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/88325.html","url":"https:\/\/www.wsisp.com\/helps\/88325.html","name":"Epoch\u3001Batch \u4e0e DataLoader - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-07-31T14:09:10+00:00","dateModified":"2026-07-31T14:09:10+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/88325.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/88325.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/88325.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"Epoch\u3001Batch \u4e0e DataLoader"}]},{"@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\/88325","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=88325"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/88325\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media\/88320"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=88325"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=88325"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=88325"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=88325"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}