{"id":208,"date":"2020-06-24T15:55:43","date_gmt":"2020-06-24T06:55:43","guid":{"rendered":"http:\/\/cedartrees.co.kr\/?p=208"},"modified":"2021-04-03T19:20:01","modified_gmt":"2021-04-03T10:20:01","slug":"pytorch-dataloader-example","status":"publish","type":"post","link":"http:\/\/blog.cedartrees.co.kr\/index.php\/2020\/06\/24\/pytorch-dataloader-example\/","title":{"rendered":"PyTorch DataLoader Example"},"content":{"rendered":"\n

sklearn\uc758 \ubd93\uaf43 \ub370\uc774\ud130\ub97c \ud65c\uc6a9\ud558\uc5ec pytorch\uc640 dataloader\ub97c \ud65c\uc6a9\ud558\uc5ec \ubd84\ub958 \ubb38\uc81c\ub97c \ud480\uc5b4 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

iris \ub370\uc774\ud130\uc14b\uc744 \ubc1b\uc544\uc11c pandas\ub85c \ub370\uc774\ud130\ub97c \ubcc0\ud658\ud569\ub2c8\ub2e4. \ubcc0\ud658 \uacfc\uc815\uc774 \ubc18\ub4dc\uc2dc \ud544\uc694\ud55c \uac83\uc740 \uc544\ub2c8\uc9c0\ub9cc \ub370\uc774\ud130\uc14b\uc744 \ubcc0\uacbd\ud558\uac70\ub098 \ud559\uc2b5\uc6a9 \uceec\ub7fc \uc815\ubcf4\ub97c \uc218\uc815\ud560 \ub54c\uc5d0 \ub3c4\uc6c0\uc774 \ub429\ub2c8\ub2e4.<\/p>\n\n\n\n

import numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\n\nfrom sklearn.datasets import load_iris\niris = load_iris()\n\ndf = pd.DataFrame(iris.data)\ndf.columns = iris.feature_names\ndf['class'] = iris.target<\/pre>\n\n\n\n
\ub2e4\uc74c\uc73c\ub85c PyTorch\ub85c \ub370\uc774\ud130\ub97c import\ud558\uc5ec \ud559\uc2b5\uc6a9 \ub370\uc774\ud130\ub97c \uc0dd\uc131\ud569\ub2c8\ub2e4. \ud559\uc2b5\uc6a9 \ub370\uc774\ud130\ub294 train_data\uc640 valid_data\ub85c \ubd84\ub9ac\ud558\ub418 8:2 \ube44\uc728\ub85c \ubd84\ub9ac\ud569\ub2c8\ub2e4.<\/p>\n\n\n\n
import torch\nimport torch.nn as nn\nimport torch.optim as optim\nimport torch.nn.functional as F\n\n## Prepare Dataset\ndata = torch.from_numpy(df.values).float()\n#data.shape = torch.Size([150, 5])\n\n# \ub370\uc774\ud130\uc14b\uc5d0\uc11c feature \uc815\ubcf4\uc640 label \ub370\uc774\ud130\ub97c \ubd84\ub9ac\ud558\uc5ec x,y \ub370\uc774\ud130\ub97c \uc0dd\uc131\nx = data[:,:4]\ny = data[:,[-1]]\n\n# train, valid \ub370\uc774\ud130\uc14b \ubd84\ub9ac, \ub370\uc774\ud130\ub294 8:2 or 7:3 \uc0dd\uc131\nratio = [.8, .2]\n\ntrain_cnt = int(data.size(0) * ratio[0])\nvalid_cnt = data.size(0) - train_cnt\nprint(train_cnt, valid_cnt) #120, 30\n\n# torch.randperm\uc744 \uc0ac\uc6a9\ud574\uc11c \ub79c\ub364\ud55c int \uc21c\uc5f4\uc744 \uc0dd\uc131, train\/valid \ub370\uc774\ud130\ub85c \ubd84\ub9ac\nindices = torch.randperm(data.size(0))\nx = torch.index_select(x, dim=0, index=indice).split([train_cnt, valid_cnt], dim=0)\ny = torch.index_select(y, dim=0, index=indice).split([train_cnt, valid_cnt], dim=0)<\/pre>\n\n\n\n
pytorch\uc5d0\uc11c \uc81c\uacf5\ud558\ub294 Dataset\uacfc DataLoader\ub97c import\ud569\ub2c8\ub2e4. <\/p>\n\n\n\n
Dataset \ud074\ub798\uc2a4\ub97c \uc0c1\uc18d\ud558\uc5ec IrisDataset \ud074\ub798\uc2a4\ub97c \uc0dd\uc131\ud558\uace0 data, label\uc744 \uc785\ub825\ud569\ub2c8\ub2e4.
IrisDataset\uc744 DataLoader\uc5d0 \uc785\ub825\ud558\uc5ec \ub370\uc774\ud130\ub97c batch_size \ub9cc\ud07c \ub370\uc774\ud130\ub97c \ubd84\ub9ac\ud558\uc5ec train_loader\uc5d0 \ub123\uc5b4\uc90d\ub2c8\ub2e4. <\/p>\n\n\n\n
iris \ub370\uc774\ud130\uc14b\uc740 \ucd1d 150\uac1c \ub370\uc774\ud130\uc785\ub2c8\ub2e4. \uc774\uac83\uc744 train\/valid \ud615\ud0dc\ub85c 8:2\ub85c \ubd84\ub9ac\ud588\uae30 \ub54c\ubb38\uc5d0 train 120, valid 30\uac1c\uc758 \ub370\uc774\ud130\ub85c \uac01\uac01 \uc0dd\uc131\ub410\uc2b5\ub2c8\ub2e4. \uc774\ub807\uac8c \uc0dd\uc131\ub41c \ub370\uc774\ud130\ub97c \ud55c\ubc88\uc5d0 \ud6c8\ub828\ud558\uc9c0 \uc54a\uace0 \uc77c\uc815 \uac2f\uc218\ub85c \ub370\uc774\ud130\ub97c \ubb36\uc5b4 \uc90d\ub2c8\ub2e4. \uc0ac\uc2e4 \uc18c\uaddc\ubaa8\uc758 \ub370\uc774\ud130 \uc14b\uc5d0\uc11c\ub294 \uc774\ub7ec\ud55c batch \uc791\uc5c5\uc774 \ubd88\ud544\uc694\ud569\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \ub9ce\uc740 \uc218\uc758 \ub370\uc774\ud130\ub97c \ud6c8\ub828\ud558\uae30 \uc704\ud574\uc11c\ub294 \uc774\ub7ec\ud55c \uc791\uc5c5\uc774 \ud544\uc218\uc785\ub2c8\ub2e4. \uc774\ubc88 \uc608\uc81c\uc5d0\uc11c\ub294 30\uac1c \ub2e8\uc704\ub85c \ubb36\uc74c\uc744 \ub9cc\ub4e4\uc5b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n
\ud30c\uc774\ud1a0\uce58\uc5d0\uc11c\ub294 \uc774\ub7ec\ud55c \ubb36\uc74c \uc791\uc5c5\uc744 \ud560 \uc218 \uc788\ub294 DataLoader\ub77c\ub294 \ud3b8\ub9ac\ud55c \ud328\ud0a4\uc9c0\ub97c \uc81c\uacf5\ud569\ub2c8\ub2e4. \uc774\ub7ec\ud55c \uacfc\uc815\uc744 \ud1b5\ud574\uc11c 120\uac1c\uc758 \ub370\uc774\ud130\uac00 30\uac1c\uc2dd 4\ubb36\uc74c\uc73c\ub85c train_loader\uc5d0 \uc800\uc7a5\ub418\uac8c \ub429\ub2c8\ub2e4.<\/p>\n\n\n\n
from torch.utils.data import Dataset, DataLoader\n\n# Dataset \uc0c1\uc18d\nclass IrisDataset(Dataset):\n    \n    def __init__(self, data, labels):\n        super().__init__()\n        self.data = data\n        self.labels = labels\n\n    def __len__(self):\n        return len(self.data)\n    \n    def __getitem__(self, idx):\n        return self.data[idx], self.labels[idx]\n\n# DataLoader\ntrain_loader = DataLoader(dataset=IrisDataset(x[0],y[0]), batch_size=config['batch_size'], shuffle=True)\nvalid_loader = DataLoader(dataset=IrisDataset(x[1],y[1]), batch_size=config['batch_size'], shuffle=False)<\/pre>\n\n\n\n
\ucc38\uace0\ub85c data, train, validate, bacth_size, epoch\uc744 \uc774\ud574\ud558\uae30 \uc704\ud574 \uc608\ub97c \ub4e4\uc5b4\ubcf4\uba74…
\uc120\uc0dd\ub2d8\uc774 \ud559\uc0dd\ub4e4\uc758 \ud559\ub825 \uc218\uc900\uc744 \uc54c\uc544\ubcf4\uae30 \uc704\ud574\uc11c 100\ubb38\uc81c\ub97c \ub9cc\ub4e4\uc5c8\uc2b5\ub2c8\ub2e4. \uc120\uc0dd\ub2d8\uc740 \ud559\uc0dd\ub4e4\uc5d0\uac8c 100\ubb38\uc81c \uc911\uc5d0\uc11c 80 \ubb38\uc81c\ub97c \ud480\uc5b4\ubcf4\uba74\uc11c \uc218\ud559\uc801 \uc6d0\ub9ac\ub97c \uc124\uba85\ud569\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \ud55c\ubc88\uc5d0 80\ubb38\uc81c\ub97c \ud480\uae30 \uc5b4\ub824\uc6b0\ub2c8 20\ubb38\uc81c\uc529 1~4\uad50\uc2dc \ub3d9\uc548 \ud480\uc5b4\ubcf4\uac8c \ud569\ub2c8\ub2e4. \ud55c\ubc88\ub9cc \ubb38\uc81c\ub97c \ud480\uc5b4\ubcf4\ub294 \uac83\ubcf4\ub2e4\ub294 \uac19\uc740 \ubb38\uc81c\ub97c \ubc18\ubcf5\ud574\uc11c \ud480\uc5b4\ubcf4\ub294 \uac83\uc774 \ud6a8\uacfc\uc801\uc774\uae30 \ub54c\ubb38\uc5d0 5~8\uad50\uc2dc \ub2e4\uc2dc \ubb38\uc81c\ub97c \ud480\uc5b4\ubd05\ub2c8\ub2e4.<\/span><\/p>\n\n\n\n
\uc774\uc81c \ud559\uc0dd\ub4e4\uc740 80\ubb38\uc81c\ub97c 20\ubb38\uc81c\uc529 \ub098\ub220\uc11c 2\ubc88\uc5d0 \uac78\uccd0 \ud480\uc5b4\ubcf8\uac83\uc774 \ub429\ub2c8\ub2e4. \ub9cc\uc57d \uc2dc\uac04\uc801 \uc5ec\uc720\uac00 \uc788\ub2e4\uba74 2\ubc88\uc774 \uc544\ub2c8\ub77c 3\ubc88, 4\ubc88 \ud480\uc5b4\ubcf8\ub2e4\uba74 \uc544\ub9c8\ub3c4 \ub354 \ud559\uc2b5\uc774 \uc798\ub418\uaca0\uc8e0.<\/p>\n\n\n\n
\uc774\uc81c \ud559\uc0dd\ub4e4\uc774 \uc218\ud559\uc6d0\ub9ac\ub97c \uc798 \uc774\ud574\ud588\ub294\uc9c0 \ud14c\uc2a4\ud2b8\ud574\ubcf4\uae30 \uc704\ud574\uc11c \ub0a8\uaca8\ub454 20\ubb38\uc81c\ub97c \ud480\uc5b4\ubcf4\uac8c \ud569\ub2c8\ub2e4. \uadf8\ub9ac\uace0 20\uac1c\uc758 \ubb38\uc81c\ub97c \uc5bc\ub9c8\ub098 \ub9ce\uc740 \ud559\uc0dd\uc774 \ub9de\ucdc4\ub294\uc9c0\ub97c \uacc4\uc0b0\ud574\ubd05\ub2c8\ub2e4.<\/p>\n\n\n\n
\uc774\ub7ec\ud55c \uacfc\uc815\uc740 \ubcf4\ud1b5\uc758 \ud559\uc2b5\uc5d0\uc11c \ub9e4\uc6b0 \uc77c\ubc18\uc801\uc778 \ubc29\ubc95\uc785\ub2c8\ub2e4. \uc774\uc81c \uc0dd\uac01\ud574\ubcf4\uba74 100\ubb38\uc81c\uac00 data, 80\ubb38\uc81c\uac00 train_data, 20\ubb38\uc81c\uac00 valid_data, 80\ubb38\uc81c\ub97c 20\ubb38\uc81c\uc529 \ub098\ub220\uc11c 4\ubb36\uc74c\uc744 \ub9cc\ub4dc\ub294 \uacfc\uc815 batch, \uac19\uc740 \ubb38\uc81c\ub97c \ucd1d 2\ud68c \ud480\uc5b4\ubd04 epoch \uc774\uac83\uc774 \uc9c0\uae08\uae4c\uc9c0\uc758 \uacfc\uc815\uc5d0\uc11c \uc0ac\uc6a9\ud588\ub358 \uc6a9\uc5b4\ub97c \uc815\uc758\ud55c \uac83\uc785\ub2c8\ub2e4.<\/span><\/p>\n\n\n\n
\uc989, train_loader\ub294 120\uac1c\uc758 \ub370\uc774\ud130\uac00 30\uac1c\uc529 4\ubb36\uc74c\uc73c\ub85c \ub418\uc5b4 \uc788\ub294 \uac83\uc774 \ub429\ub2c8\ub2e4. valid_loader\ub294 30\uac1c\uc758 \ub370\uc774\ud130\uac00 30\uac1c\uc529 1\ubb36\uc74c\uc774 \ub418\uaca0\ub124\uc694.<\/p>\n\n\n\n
\uc790, \uc774\uc81c \ubaa8\ub378\uc744 \uac04\ub2e8\ud788 \uad6c\uc131\ud569\ub2c8\ub2e4. \ud559\uc2b5\uc744 \uc704\ud55c \ubaa8\ub378\uc774\ub77c\uae30 \ubcf4\ub2e4\ub294 \uac04\ub2e8\ud788 \ud14c\uc2a4\ud2b8\ud558\uae30 \uc704\ud55c \uac83\uc784\uc73c\ub85c \uac04\ub2e8\ud55c \ubaa8\ub378\uc744 \ub9cc\ub4e4\uc5b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n
\uc608\uce21 \ub370\uc774\ud130\ub294 \ubd93\uaf43\uc758 \uaf43\ubc1b\uce68\uc758 \uae38\uc774\uc640 \ub108\ube44, \uaf43\uc78e\uc758 \uae38\uc774\uc640 \ub108\ube44\uc5d0 \ub530\ub77c 3\uc885\ub958 \uc911 \ud558\ub098\ub85c \uc608\uce21\ud558\ub294 \uac83\uc784\uc73c\ub85c \ucd5c\uc885 \uc544\uc6c3\ud48b\uc758 \ud615\ud0dc\ub294  3\uc785\ub2c8\ub2e4. \uadf8\ub9ac\uace0 \ud574\ub2f9 \ub370\uc774\ud130\ub97c \ud655\ub960 \uac12\uc73c\ub85c \ub098\ud0c0\ub0b4\uae30 \uc704\ud558\uc5ec softmax_classification\uc744 \ud65c\uc6a9\ud569\ub2c8\ub2e4.<\/p>\n\n\n\n
# model \uc0dd\uc131\nmodel = nn.Sequential(\n    nn.Linear(4,3)\n)\n\noptimizer = optim.Adam(model.parameters())\n\nfrom copy import deepcopy\nlowest_loss = np.inf\nbest_model = None\nlowest_epoch = np.inf<\/pre>\n\n\n\n
copy \ud328\ud0a4\uc9c0\ub85c\ubd80\ud130 deepcopy\ub97c import\ud569\ub2c8\ub2e4. \uc774\uac83\uc740 \uc774\ubc88\uc5d0 \ub370\uc774\ud130\ub97c \ub9cc\ub4dc\ub294 \uacfc\uc815\uacfc \uc9c1\uc811\uc801\uc778 \uad00\ub828\uc774 \uc5c6\uae30 \ub54c\ubb38\uc5d0 \uac04\ub2e8\ud788\ub9cc \uc124\uba85\ud558\uba74 \uac1d\uccb4\uc758 \ubaa8\ub4e0 \ub0b4\uc6a9\uc744 \ubcf5\uc0ac\ud574\uc11c \uc0c8\ub85c\uc6b4 \ud558\ub098\uc758 \uac1d\uccb4\ub97c \ub9cc\ub4dc\ub294 \uac83\uc744 deep copy\ub77c\uace0 \ud569\ub2c8\ub2e4. \ubc18\ub300\uc758 \uac1c\ub150\uc740 shallow copy \uc785\ub2c8\ub2e4.<\/p>\n\n\n\n
\uc774\uc81c \ud559\uc2b5\uc744 \uc2dc\uc791\ud569\ub2c8\ub2e4. \uc774 \ubaa8\ub378\uc740 2\uac1c\uc758 for loop\uc73c\ub85c \ub418\uc5b4 \uc788\uc2b5\ub2c8\ub2e4. \uac00\uc7a5 \uba3c\uc800 \ub098\uc624\ub294 for loop\uc740 epoch\uc5d0 \ub300\ud55c \uc815\uc758\ub85c train data\ub97c \ucd1d \uba87\ubc88 \ud559\uc2b5\ud558\ub294\uac00\uc5d0 \ub300\ud55c \uc815\uc758\uc785\ub2c8\ub2e4.  \ub2e4\uc74c\uc5d0 \ub098\uc624\ub294 \ub610 \ud558\ub098\uc758 for loop\uc740 \ud559\uc2b5 \ub370\uc774\ud130\ub97c \uba87\uac1c\ub85c \ub098\ub220\uc11c \ud559\uc2b5\ud560 \uac83\uc778\uac00 \uc989, batch\uc5d0 \ub300\ud55c \ubb38\uc81c\uc785\ub2c8\ub2e4.<\/p>\n\n\n\n
1\ubc88 \ud559\uc2b5\uc774 \ub05d\ub098\uba74 \ud559\uc2b5\uc758 loss\ub97c \uacc4\uc0b0\ud574\ubd05\ub2c8\ub2e4. loss\ub294 \uc815\ub2f5\uacfc\uc758 \ucc28\uc774\ub97c \uc758\ubbf8\ud558\ub294 \uac83\uc73c\ub85c \uc791\uc73c\uba74 \uc791\uc744 \uc218\ub85d \ud559\uc2b5\uc774 \uc798\ub410\ub2e4\ub294 \uc758\ubbf8\uc785\ub2c8\ub2e4. \ud55c\ubc88 \ud559\uc2b5\uc774 \ub05d\ub098\uba74 valid data\ub97c \uc2e4\ud589\ud574\ubd05\ub2c8\ub2e4. \uadf8\ub9ac\uace0 valid\uc5d0\uc11c \ub098\uc628 loss\uc640 train\uc5d0\uc11c \ub098\uc628 loss\ub97c \ube44\uad50\ud574\ubcf4\uace0 valid\uc758 loss\uac00 \ub354 \uc88b\uc744 \ub54c\uc5d0 \ud574\ub2f9 \ud559\uc2b5\uc5d0 \uc0ac\uc6a9\ud55c \ubaa8\ub378\uc744 deepcopy\ud574\uc11c \uc800\uc7a5\ud569\ub2c8\ub2e4.<\/p>\n\n\n\n
\uadf8 \uc774\uc720\ub294 \ubb34\uc870\uac74 \ud559\uc2b5\uc744 \uc624\ub798 \ud55c\ub2e4\uace0 \ud574\uc11c \uc88b\uc740 \uacb0\uacfc\uac00 \ub098\uc624\ub294 \uac83\uc774 \uc544\ub2c8\uace0 \uc5b4\ub290 \uc21c\uac04\uc5d0 \ud559\uc2b5\uc774 \uc815\uccb4\ub418\uac70\ub098 \uacfc\uc801\ud569 \ub418\ub294 \uc77c\uc774 \uc788\uae30 \ub54c\ubb38\uc5d0 \uac00\uc7a5 \uc88b\uc740 \ubaa8\ub378\uc744 \uc800\uc7a5\ud558\ub294 \uac83\uc785\ub2c8\ub2e4.<\/p>\n\n\n\n
train_history, valid_history = [], []\n\nfor i in range(config['n_epochs']+1):\n    model.train()\n    \n    train_loss, valid_loss = 0, 0\n    y_hat = []\n    \n    # train_batch start\n    for x_i, y_i in train_loader:\n        y_hat_i = model(x_i)\n        loss = F.cross_entropy(y_hat_i, y_i.long().squeeze())\n        \n        optimizer.zero_grad()\n        loss.backward()\n\n        optimizer.step()        \n        train_loss += float(loss) # This is very important to prevent memory leak.\n\n    train_loss = train_loss \/ len(train_loader)\n    \n    model.eval()\n    with torch.no_grad():\n        valid_loss = 0\n        \n        for x_i, y_i in valid_loader:\n            y_hat_i = model(x_i)\n            loss = F.cross_entropy(y_hat_i, y_i.long().squeeze())\n            \n            valid_loss += float(loss)\n            \n            y_hat += [y_hat_i]\n            \n    valid_loss = valid_loss \/ len(valid_loader)\n    \n    train_history.append(train_loss)\n    valid_history.append(valid_loss)\n    \n    if i % config['print_interval'] == 0:\n        print('Epoch %d: train loss=%.4e  valid_loss=%.4e  lowest_loss=%.4e' % (i, train_loss, valid_loss, lowest_loss))\n        \n    if valid_loss <= lowest_loss:\n        lowest_loss = valid_loss\n        lowest_epoch = i\n        best_model = deepcopy(model.state_dict())\n        \n    model.load_state_dict(best_model)<\/pre>\n\n\n\n
\uc774\uc81c \ud559\uc2b5\uc774 \uc798\ub410\ub294\uc9c0 \uc544\ub798\uc640 \uac19\uc740 \ubc29\ubc95\uc73c\ub85c train_loss\uc640 valid_loss\ub97c \ud45c\uc2dc\ud574\ubd05\ub2c8\ub2e4.<\/p>\n\n\n\n
import matplotlib.pyplot as plt\n\nfig, loss_ax = plt.subplots()\n\nloss_ax.plot(train_history, 'y', label='train loss')\nloss_ax.plot(valid_history, 'r', label='val loss')\n\nloss_ax.set_xlabel('epoch')\nloss_ax.set_ylabel('loss')\n\nloss_ax.legend(loc='upper left')\n\nplt.show()<\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
\uc0ac\uc2e4 \uc774 \uc608\uc81c\ub294 torch\uc758 Dataset\uacfc DataLoader\ub97c \uc0ac\uc6a9\ud558\ub294 \ubc29\ubc95\uc5d0 \ub300\ud55c \uc608\uc81c\uc600\ub294\ub370 \uc774\uac83\uc800\uac83 \uc124\uba85\ud558\ub2e4 \ubcf4\ub2c8 \uae00\uc774 \uae38\uc5b4\uc84c\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n
\uc5ec\uae30\uc11c \uc911\uc694\ud55c \uac83\uc740 Dataset\uc744 \ub9cc\ub4e4\uace0 DataLoader\ub97c \ud1b5\ud574\uc11c \ud559\uc2b5\uc5d0 \uc0ac\uc6a9\ud558\ub294 \ubc29\ubc95\uc5d0 \ub300\ud55c \ub0b4\uc6a9\uc774 \uc911\uc694\ud558\ub2c8 \uc608\uc81c \ucf54\ub4dc\ub97c \ud65c\uc6a9\ud574\uc11c \uc9c1\uc811 \ud14c\uc2a4\ud2b8\ud574\ubcf4\uc2dc\uae30 \ubc14\ub78d\ub2c8\ub2e4.<\/p>\n","protected":false},"excerpt":{"rendered":"
sklearn\uc758 \ubd93\uaf43 \ub370\uc774\ud130\ub97c \ud65c\uc6a9\ud558\uc5ec pytorch\uc640 dataloader\ub97c \ud65c\uc6a9\ud558\uc5ec \ubd84\ub958 \ubb38\uc81c\ub97c \ud480\uc5b4 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4. iris \ub370\uc774\ud130\uc14b\uc744 \ubc1b\uc544\uc11c pandas\ub85c \ub370\uc774\ud130\ub97c \ubcc0\ud658\ud569\ub2c8\ub2e4. \ubcc0\ud658 \uacfc\uc815\uc774 \ubc18\ub4dc\uc2dc \ud544\uc694\ud55c \uac83\uc740 \uc544\ub2c8\uc9c0\ub9cc \ub370\uc774\ud130\uc14b\uc744 \ubcc0\uacbd\ud558\uac70\ub098 \ud559\uc2b5\uc6a9 \uceec\ub7fc \uc815\ubcf4\ub97c \uc218\uc815\ud560 \ub54c\uc5d0 \ub3c4\uc6c0\uc774 \ub429\ub2c8\ub2e4. \ub2e4\uc74c\uc73c\ub85c PyTorch\ub85c \ub370\uc774\ud130\ub97c import\ud558\uc5ec \ud559\uc2b5\uc6a9 \ub370\uc774\ud130\ub97c \uc0dd\uc131\ud569\ub2c8\ub2e4. \ud559\uc2b5\uc6a9 \ub370\uc774\ud130\ub294 train_data\uc640 valid_data\ub85c \ubd84\ub9ac\ud558\ub418 8:2 \ube44\uc728\ub85c \ubd84\ub9ac\ud569\ub2c8\ub2e4. pytorch\uc5d0\uc11c \uc81c\uacf5\ud558\ub294 Dataset\uacfc DataLoader\ub97c import\ud569\ub2c8\ub2e4. Dataset \ud074\ub798\uc2a4\ub97c \uc0c1\uc18d\ud558\uc5ec … <\/p>\n
\ub354 \ubcf4\uae30 “PyTorch DataLoader Example”<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[24,21,27],"tags":[128,6],"_links":{"self":[{"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/posts\/208"}],"collection":[{"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/comments?post=208"}],"version-history":[{"count":12,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/posts\/208\/revisions"}],"predecessor-version":[{"id":741,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/posts\/208\/revisions\/741"}],"wp:attachment":[{"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/media?parent=208"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/categories?post=208"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/tags?post=208"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}