\ub525\ub7ec\ub2dd\uc740 \ucd5c\uadfc \uba87 \ub144\uac04 \ube60\ub974\uac8c \ubc1c\uc804\ud574\uc654\uc73c\uba70, \uc774\ub7ec\ud55c \ubc1c\uc804\uc5d0\ub294 \uac15\ub825\ud55c \ud558\ub4dc\uc6e8\uc5b4\uc758 \ub3c4\uc6c0\uc774 \uaf2d \ud544\uc694\ud569\ub2c8\ub2e4. \ud2b9\ud788 CPU\uc640 GPU\ub294 \ub525\ub7ec\ub2dd \ubaa8\ub378\uc758 \ud6c8\ub828 \ubc0f \ucd94\ub860 \uc131\ub2a5\uc5d0\uc11c \uc911\uc694\ud55c \uc5ed\ud560\uc744 \ud569\ub2c8\ub2e4. \ubcf8 \uac15\uc88c\uc5d0\uc11c\ub294 CPU\uc640 GPU\uc758 \uad6c\uc870, \uc791\ub3d9 \uc6d0\ub9ac, \uadf8\ub9ac\uace0 \uc774\ub97c \ud65c\uc6a9\ud55c \ud30c\uc774\ud1a0\uce58 \uc608\uc81c \ucf54\ub4dc\ub97c \ud1b5\ud574 \uc5b4\ub5bb\uac8c \ud6a8\uc728\uc801\uc73c\ub85c \ub525\ub7ec\ub2dd \ubaa8\ub378\uc744 \ud6c8\ub828\uc2dc\ud0ac \uc218 \uc788\ub294\uc9c0 \uc54c\uc544\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
CPU(Central Processing Unit)\ub294 \ucef4\ud4e8\ud130\uc758 \uc911\uc559 \ucc98\ub9ac \uc7a5\uce58\ub85c, \ubcf5\uc7a1\ud55c \uacc4\uc0b0\uc744 \uc218\ud589\ud558\uc5ec \ub2e4\uc591\ud55c \uc791\uc5c5\uc744 \ucc98\ub9ac\ud558\ub294 \ub2a5\ub825\uc774 \ub6f0\uc5b4\ub0a9\ub2c8\ub2e4. \ubc18\uba74 GPU(Graphics Processing Unit)\ub294 \ub300\ub7c9\uc758 \ub370\uc774\ud130 \ubcd1\ub82c \ucc98\ub9ac\ub97c \uc218\ud589\ud558\ub294 \ub370 \ucd5c\uc801\ud654\ub41c \ud558\ub4dc\uc6e8\uc5b4\uc785\ub2c8\ub2e4. \uc774 \ub450 \ud504\ub85c\uc138\uc11c\ub294 \uac01\uac01 \ub2e4\uc74c\uacfc \uac19\uc740 \ud2b9\uc131\uc744 \uac00\uc9c0\uace0 \uc788\uc2b5\ub2c8\ub2e4:<\/p>\n
\ub525\ub7ec\ub2dd \ubaa8\ub378 \ud6c8\ub828\uc5d0\uc11c\ub294 \uc218\ucc9c \uac1c\uc758 \ud30c\ub77c\ubbf8\ud130\ub97c \ucd5c\uc801\ud654\ud574\uc57c \ud558\uba70, \uc774 \uacfc\uc815\uc740 \uc218\ub9ce\uc740 \ud589\ub82c \uc5f0\uc0b0\uc73c\ub85c \uc774\ub8e8\uc5b4\uc9d1\ub2c8\ub2e4. \uc774\ub54c GPU\ub294 \ub300\ub7c9\uc758 \ub370\uc774\ud130\ub97c \ub3d9\uc2dc\uc5d0 \ucc98\ub9ac\ud560 \uc218 \uc788\ub294 \ubcd1\ub82c \ucc98\ub9ac \ub2a5\ub825\uc744 \ubc1c\ud718\ud558\uc5ec \ud6c8\ub828 \uc2dc\uac04\uc744 \ub2e8\ucd95\uc2dc\ucf1c \uc90d\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, GPU\ub97c \uc774\uc6a9\ud55c \ud6c8\ub828\uc740 CPU\uc5d0 \ube44\ud574 \uc218\uc2ed \ubc30\uc5d0\uc11c \uc218\ubc31 \ubc30\uae4c\uc9c0 \ube60\ub97c \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\ud30c\uc774\ud1a0\uce58\uc5d0\uc11c\ub294 CPU\uc640 GPU\ub97c \uac04\ud3b8\ud558\uac8c \uc120\ud0dd\ud558\uc5ec \uc0ac\uc6a9 \uac00\ub2a5\ud569\ub2c8\ub2e4. \uae30\ubcf8\uc801\uc73c\ub85c\ub294 CPU\ub97c \uc0ac\uc6a9\ud558\uc9c0\ub9cc, GPU\ub97c \uc0ac\uc6a9\ud560 \uc218 \uc788\ub294 \uacbd\uc6b0\uc5d0\ub294 \ucf54\ub4dc\uc5d0\uc11c \uba87 \uac00\uc9c0 \uac04\ub2e8\ud55c \ubcc0\uacbd\ub9cc\uc73c\ub85c GPU\ub97c \ud65c\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc544\ub798 \uc608\uc81c \ucf54\ub4dc\ub97c \ud1b5\ud574 \uc774\ub97c \uc0b4\ud3b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torchvision import datasets, transforms\n\n# \ub370\uc774\ud130\uc14b \uc900\ube44\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\ntrain_dataset = datasets.MNIST(root='.\/data', train=True, transform=transform, download=True)\ntrain_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)\n\n# \uc2e0\uacbd\ub9dd \ubaa8\ub378 \uc815\uc758\nclass SimpleNN(nn.Module):\n def __init__(self):\n super(SimpleNN, self).__init__()\n self.fc1 = nn.Linear(28 * 28, 128)\n self.fc2 = nn.Linear(128, 10)\n \n def forward(self, x):\n x = x.view(-1, 28 * 28) # flatten\n x = torch.relu(self.fc1(x))\n x = self.fc2(x)\n return x\n\n# \ub514\ubc14\uc774\uc2a4 \uc124\uc815\ndevice = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\nmodel = SimpleNN().to(device)\n\n# \uc190\uc2e4 \ud568\uc218\uc640 \ucd5c\uc801\ud654\uae30 \uc124\uc815\ncriterion = nn.CrossEntropyLoss()\noptimizer = optim.SGD(model.parameters(), lr=0.01)\n\n# \ubaa8\ub378 \ud6c8\ub828\nfor epoch in range(5): # \ud6c8\ub828 \uc5d0\ud3ed \uc218\n for images, labels in train_loader:\n images, labels = images.to(device), labels.to(device) # GPU\ub85c \ub370\uc774\ud130 \uc774\ub3d9\n optimizer.zero_grad() # \uadf8\ub798\ub514\uc5b8\ud2b8 \ucd08\uae30\ud654\n outputs = model(images) # \uc608\uce21\n loss = criterion(outputs, labels) # \uc190\uc2e4 \uacc4\uc0b0\n loss.backward() # \uc5ed\uc804\ud30c\n optimizer.step() # \uac00\uc911\uce58 \uc5c5\ub370\uc774\ud2b8\n \n print(f'Epoch [{epoch + 1}\/5], Loss: {loss.item():.4f}')\n<\/code><\/pre>\n\ucf54\ub4dc \uc124\uba85<\/h3>\n\n- \ub370\uc774\ud130\uc14b \uc900\ube44<\/strong>: MNIST \ub370\uc774\ud130\uc14b\uc744 \ub85c\ub4dc\ud558\uace0 \uc804\ucc98\ub9ac\ud558\uc5ec DataLoader\ub85c \uc900\ube44\ud569\ub2c8\ub2e4.<\/li>\n
- \uc2e0\uacbd\ub9dd \ubaa8\ub378 \uc815\uc758<\/strong>: \uac04\ub2e8\ud55c \ub450 \uce35 \uad6c\uc870\uc758 \uc2e0\uacbd\ub9dd\uc744 \uc815\uc758\ud569\ub2c8\ub2e4.<\/li>\n
- \ub514\ubc14\uc774\uc2a4 \uc124\uc815<\/strong>: GPU\uac00 \uc0ac\uc6a9 \uac00\ub2a5\ud55c \uacbd\uc6b0 GPU\ub97c \uc0ac\uc6a9\ud558\uace0, \uadf8\ub807\uc9c0 \uc54a\uc73c\uba74 CPU\ub97c \uc0ac\uc6a9\ud569\ub2c8\ub2e4.<\/li>\n
- \ubaa8\ub378 \ud6c8\ub828<\/strong>: \uc815\uc758\ud55c \ub370\uc774\ud130\uc640 \ubaa8\ub378\uc744 \uc0ac\uc6a9\ud558\uc5ec \ud6c8\ub828\uc744 \uc9c4\ud589\ud569\ub2c8\ub2e4. \uc774\ub54c \ub370\uc774\ud130\ub97c GPU\ub85c \uc774\ub3d9\uc2dc\ud0a4\ub294 \uac83\uc744 \uc78a\uc9c0 \uc54a\uc544\uc57c \ud569\ub2c8\ub2e4.<\/li>\n<\/ul>\n
CPU\uc640 GPU \uc131\ub2a5 \ube44\uad50<\/h2>\n
GPU\ub97c \uc0ac\uc6a9\ud588\uc744 \ub54c\uc758 \uc131\ub2a5 \uc774\uc810\uc740 \ub2e4\uc74c\uacfc \uac19\uc740 \uce21\uc815\uc744 \ud1b5\ud574 \ud655\uc778\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ubcf4\ud1b5 CPU\uc640 GPU \ubaa8\ub450 \ud6c8\ub828 \uc2dc\uac04\uacfc \uc815\ud655\ub3c4 \uce21\uba74\uc5d0\uc11c \ucc28\uc774\ub97c \ubcf4\uc785\ub2c8\ub2e4. \ub2e4\uc74c\uc740 CPU\uc640 GPU\ub97c \uc0ac\uc6a9\ud560 \ub54c\uc758 \ud6c8\ub828 \uc2dc\uac04 \uc608\uc2dc\uc785\ub2c8\ub2e4:<\/p>\n
\nimport time\n\n# CPU \uc131\ub2a5 \ud14c\uc2a4\ud2b8\ndevice_cpu = torch.device('cpu')\nmodel_cpu = SimpleNN().to(device_cpu)\n\nstart_time = time.time()\nfor epoch in range(5):\n for images, labels in train_loader:\n images, labels = images.to(device_cpu), labels.to(device_cpu)\n optimizer.zero_grad()\n outputs = model_cpu(images)\n loss = criterion(outputs, labels)\n loss.backward()\n optimizer.step()\nend_time = time.time()\nprint(f'CPU Training Time: {end_time - start_time:.2f} seconds')\n\n# GPU \uc131\ub2a5 \ud14c\uc2a4\ud2b8\ndevice_gpu = torch.device('cuda')\nmodel_gpu = SimpleNN().to(device_gpu)\n\nstart_time = time.time()\nfor epoch in range(5):\n for images, labels in train_loader:\n images, labels = images.to(device_gpu), labels.to(device_gpu)\n optimizer.zero_grad()\n outputs = model_gpu(images)\n loss = criterion(outputs, labels)\n loss.backward()\n optimizer.step()\nend_time = time.time()\nprint(f'GPU Training Time: {end_time - start_time:.2f} seconds')\n<\/code><\/pre>\n\uc704 \ucf54\ub4dc\ub97c \uc2e4\ud589\ud558\uba74 CPU\uc640 GPU\uc758 \ud6c8\ub828 \uc18c\uc694 \uc2dc\uac04\uc744 \ube44\uad50\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc77c\ubc18\uc801\uc73c\ub85c\ub294 GPU\uac00 \ub354 \ube60\ub978 \ud6c8\ub828 \uc131\ub2a5\uc744 \ubcf4\uc5ec\uc8fc\uc9c0\ub9cc, \ubaa8\ub378\uc758 \ubcf5\uc7a1\ub3c4\uc640 \ub370\uc774\ud130 \ud06c\uae30, \ud558\ub4dc\uc6e8\uc5b4 \uc131\ub2a5\uc5d0 \ub530\ub77c \ucc28\uc774\uac00 \uc788\uc744 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\uacb0\ub860<\/h2>\n
\ub525\ub7ec\ub2dd \ubaa8\ub378\uc744 \ud6a8\uc728\uc801\uc73c\ub85c \ud6c8\ub828\uc2dc\ud0a4\uae30 \uc704\ud574\uc11c\ub294 CPU\uc640 GPU\uc758 \ud2b9\uc9d5\uacfc \uc7a5\ub2e8\uc810\uc744 \uc798 \uc774\ud574\ud574\uc57c \ud569\ub2c8\ub2e4. CPU\ub294 \ubc94\uc6a9\uc131\uc744 \uac16\ucd98 \ubc18\uba74, GPU\ub294 \ub300\ub7c9\uc758 \ub370\uc774\ud130 \ucc98\ub9ac\ub97c \ud6a8\uacfc\uc801\uc73c\ub85c \uc218\ud589\ud558\ub294 \ub370 \ucd5c\uc801\ud654\ub418\uc5b4 \uc788\uc2b5\ub2c8\ub2e4. \ub530\ub77c\uc11c \ub2f9\uc2e0\uc758 \ud504\ub85c\uc81d\ud2b8\uc5d0 \ub9de\ub294 \ud558\ub4dc\uc6e8\uc5b4\ub97c \uc120\ud0dd\ud558\uace0, \ud30c\uc774\ud1a0\uce58\ub97c \ud1b5\ud574 \uc774\uc5d0 \ub9de\uac8c \ucf54\ub4dc\ub97c \uc791\uc131\ud55c\ub2e4\uba74, \ub354\uc6b1 \ud6a8\uc728\uc801\uc73c\ub85c \ub525\ub7ec\ub2dd \ubaa8\ub378\uc744 \uad6c\ucd95\ud560 \uc218 \uc788\uc744 \uac83\uc785\ub2c8\ub2e4.<\/p>\n
\ucd94\uac00\uc801\uc73c\ub85c, GPU\ub97c \ud65c\uc6a9\ud560 \ub54c \uc8fc\uc758\ud574\uc57c \ud560 \uc810\uc740 GPU \uba54\ubaa8\ub9ac\uc758 \ud55c\uacc4\ub97c \uc778\uc2dd\ud558\uace0 \ud544\uc694\ud55c \uacbd\uc6b0 \ubbf8\ub2c8\ubc30\uce58\ub97c \uc870\uc815\ud558\uc5ec \uc0ac\uc6a9\ud574\uc57c \ud55c\ub2e4\ub294 \uc810\uc785\ub2c8\ub2e4. \uc774\ub7ec\ud55c \uc810\ub4e4\uc774 \ud30c\uc774\ud1a0\uce58\uc640 \ub525\ub7ec\ub2dd\uc758 \ud65c\uc6a9\ub3c4\ub97c \ub192\uc5ec\uc904 \uac83\uc785\ub2c8\ub2e4.<\/p>\n
<\/body><\/p>\n","protected":false},"excerpt":{"rendered":"