\u9a57\u8b49\u5b89\u88dd\u53ca\u8a2d\u5099<\/strong><\/p>\n\n\n\n\u6253\u958b\u4f60\u7684 Python \u74b0\u5883\uff08Jupyter Notebook, VS Code, PyCharm \u7b49\uff09\uff0c\u57f7\u884c\u4ee5\u4e0b\u4ee3\u78bc\u6aa2\u67e5\u8a2d\u5099\uff1a<\/p>\n\n\n\n
import torch\n\nprint(f\"PyTorch version: {torch.__version__}\")\nprint(f\"CUDA (NVIDIA GPU) available: {torch.cuda.is_available()}\")\nif torch.cuda.is_available():\n print(f\"CUDA device name: {torch.cuda.get_device_name(0)}\")\n\nprint(f\"MPS (Apple Silicon) available: {torch.backends.mps.is_available()}\")\n\n# Select which device to use\ndevice = \"cuda\" if torch.cuda.is_available() else \"mps\" if torch.backends.mps.is_available() else \"cpu\"\nprint(f\"Using device: {device}\")<\/code><\/pre>\n\n\n\n\u5c08\u6848\u7d50\u69cb<\/strong><\/h2>\n\n\n\nmy_python_project\/ # \u5c08\u6848\u6839\u76ee\u9304\n\u251c\u2500\u2500 my_project_env\/ # \u865b\u64ec\u74b0\u5883\u8cc7\u6599\u593e\uff08\u4e0d\u6703\u4e0a\u50b3\u5230 Git\uff09\n\u251c\u2500\u2500 app.py # \u4e3b\u7a0b\u5f0f\n\u251c\u2500\u2500 requirements.txt # \u5957\u4ef6\u4f9d\u8cf4\u6e05\u55ae\n\u2514\u2500\u2500 .gitignore # Git \u5ffd\u7565\u898f\u5247<\/code><\/pre>\n\n\n\nCode<\/h2>\n\n\n\n
\u8b93\u6211\u5011\u5be6\u4f5c\u4e00\u500b\u6700\u7d93\u5178\u7684\u7bc4\u4f8b\uff1a\u5728 Fashion-MNIST<\/strong> \u8cc7\u6599\u96c6\u4e0a\u8a13\u7df4\u4e00\u500b\u7c21\u55ae\u7684\u795e\u7d93\u7db2\u8def\u4f86\u5c0d\u8863\u7269\u5716\u7247\u9032\u884c\u5206\u985e\u3002\u6b64\u4ee3\u78bc\u5c55\u793a\u4e86\u5b8c\u6574\u7684\u6d41\u7a0b\uff0c\u4e26\u81ea\u52d5\u5229\u7528\u53ef\u7528\u7684\u8a2d\u5099\u9032\u884c\u52a0\u901f\u3002<\/p>\n\n\n\n# import_libraries.py\nimport torch\nfrom torch import nn, optim\nfrom torch.utils.data import DataLoader\nfrom torchvision import datasets, transforms\nimport matplotlib.pyplot as plt\n\n# 1. Set device\ndevice = \"cuda\" if torch.cuda.is_available() else \"mps\" if torch.backends.mps.is_available() else \"cpu\"\nprint(f\"Using {device} device\")\n\n# 2. Load and prepare dataset\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\ntrain_dataset = datasets.FashionMNIST(root='.\/data', train=True, download=True, transform=transform)\ntest_dataset = datasets.FashionMNIST(root='.\/data', train=False, download=True, transform=transform)\n\ntrain_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)\ntest_loader = DataLoader(test_dataset, batch_size=64, shuffle=False)\n\n# Check the shape of one batch\nfor X, y in train_loader:\n print(f\"Shape of X [Batch, Channel, Height, Width]: {X.shape}\")\n print(f\"Shape of y: {y.shape} {y.dtype}\")\n break\n\n# 3. Build neural network model\nclass NeuralNetwork(nn.Module):\n def __init__(self):\n super().__init__()\n self.flatten = nn.Flatten()\n self.linear_relu_stack = nn.Sequential(\n nn.Linear(28*28, 512),\n nn.ReLU(),\n nn.Linear(512, 512),\n nn.ReLU(),\n nn.Linear(512, 10)\n )\n\n def forward(self, x):\n x = self.flatten(x)\n logits = self.linear_relu_stack(x)\n return logits\n\nmodel = NeuralNetwork().to(device) # Move model to device (GPU\/MPS\/CPU)\nprint(model)\n\n# 4. Define loss function and optimizer\nloss_fn = nn.CrossEntropyLoss()\noptimizer = optim.Adam(model.parameters(), lr=1e-3)\n\n# 5. Training loop\ndef train(dataloader, model, loss_fn, optimizer):\n size = len(dataloader.dataset)\n model.train()\n for batch, (X, y) in enumerate(dataloader):\n X, y = X.to(device), y.to(device) # Move data to the same device\n\n # Compute prediction error\n pred = model(X)\n loss = loss_fn(pred, y)\n\n # Backpropagation\n optimizer.zero_grad()\n loss.backward()\n optimizer.step()\n\n if batch % 100 == 0:\n loss, current = loss.item(), batch * len(X)\n print(f\"loss: {loss:>7f} [{current:>5d}\/{size:>5d}]\")\n\n# 6. Test loop\ndef test(dataloader, model, loss_fn):\n size = len(dataloader.dataset)\n num_batches = len(dataloader)\n model.eval()\n test_loss, correct = 0, 0\n with torch.no_grad():\n for X, y in dataloader:\n X, y = X.to(device), y.to(device)\n pred = model(X)\n test_loss += loss_fn(pred, y).item()\n correct += (pred.argmax(1) == y).type(torch.float).sum().item()\n test_loss \/= num_batches\n correct \/= size\n print(f\"Test Results: \\n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f} \\n\")\n return correct\n\n# 7. Run training\nepochs = 5\naccuracy_history = []\nfor t in range(epochs):\n print(f\"Epoch {t+1}\\n-------------------------------\")\n train(train_loader, model, loss_fn, optimizer)\n acc = test(test_loader, model, loss_fn)\n accuracy_history.append(acc)\nprint(\"Training Done!\")\n\n# 8. Save trained model\ntorch.save(model.state_dict(), \"outputs\/model.pth\")\nprint(\"Saved model state to model.pth\")<\/code><\/pre>\n\n\n\n\u8f38\u51fa<\/h2>\n\n\n\n
\u7576\u4f60\u904b\u884c\u4e0a\u8ff0\u4ee3\u78bc\u5f8c\uff0c\u4f60\u5c07\u6703\u5728\u63a7\u5236\u53f0\u770b\u5230\u985e\u4f3c\u4ee5\u4e0b\u7684\u8f38\u51fa\uff0c\u4e26\u5728 outputs<\/code> \u8cc7\u6599\u593e\u4e2d\u751f\u6210\u6a21\u578b\u6587\u4ef6\u3002<\/p>\n\n\n\n\u63a7\u5236\u53f0\u8f38\u51fa (Console Output):<\/strong><\/p>\n\n\n\nUsing cuda device # \u6216 Using mps device \/ Using cpu device\nShape of X [Batch, Channel, Height, Width]: torch.Size([64, 1, 28, 28])\nShape of y: torch.Size([64]) torch.int64\nNeuralNetwork(\n (flatten): Flatten(start_dim=1, end_dim=-1)\n (linear_relu_stack): Sequential(\n (0): Linear(in_features=784, out_features=512, bias=True)\n (1): ReLU()\n (2): Linear(in_features=512, out_features=512, bias=True)\n (3): ReLU()\n (4): Linear(in_features=512, out_features=10, bias=True)\n )\n)\nEpoch 1\n-------------------------------\nloss: 2.301106 [ 0\/60000]\nloss: 0.558233 [ 6400\/60000]\n...\nTest Error:\n Accuracy: 84.0%, Avg loss: 0.412345\n\nEpoch 5\n-------------------------------\n...\nTest Error:\n Accuracy: 87.2%, Avg loss: 0.352123\n\nTraining Done!\nSaved PyTorch Model State to model.pth<\/code><\/pre>\n\n\n\n\u7d50\u8ad6<\/h2>\n\n\n\n
\u901a\u904e\u9019\u7bc7\u300c\u5f9e\u96f6\u958b\u59cb\u300d\u7684 PyTorch \u6307\u5357\uff0c\u6211\u5011\u89e3\u5bc6\u4e86 2025 \u5e74\u4eba\u5de5\u667a\u6167\u958b\u767c\u7684\u6838\u5fc3\u5de5\u5177\u8207\u5be6\u6230\u61c9\u7528\u3002\u4f60\u5df2\u7d93\u5b78\u6703\u4e86\uff1a<\/p>\n\n\n\n
\n- \u70ba\u4ec0\u9ebc\u5b83\u662f\u73fe\u4ee3 AI \u958b\u767c\u7684\u5229\u5668\uff1a\u76f4\u89ba\u3001\u9748\u6d3b\u4e14\u751f\u614b\u5f37\u5927\u3002<\/li>\n\n\n\n
- \u5982\u4f55\u8a2d\u7f6e\u74b0\u5883\u4e26\u5229\u7528\u786c\u9ad4\u52a0\u901f\uff1a\u900f\u904e
.to(device)<\/code> \u8f15\u9b06\u5728 CPU\u3001GPU (CUDA) \u548c MPS \u4e4b\u9593\u5207\u63db\u3002<\/li>\n\n\n\n- \u4e00\u500b\u6a19\u6e96\u7684\u5c08\u6848\u6d41\u7a0b\uff1a\u5f9e\u8cc7\u6599\u8f09\u5165\u3001\u6a21\u578b\u5b9a\u7fa9\u3001\u8a13\u7df4\u5faa\u74b0\u5230\u6a21\u578b\u4fdd\u5b58\u3002<\/li>\n<\/ul>\n\n\n\n
\u9019\u50c5\u50c5\u662f\u958b\u59cb\u3002AI \u7684\u4e16\u754c\u5ee3\u95ca\u800c\u6df1\u9083\uff0c\u7b49\u5f85\u4f60\u63a2\u7d22\u66f4\u8907\u96dc\u7684\u6a21\u578b\uff08CNN, RNN, Transformer\uff09\u3001\u66f4\u9ad8\u7d1a\u7684\u6280\u8853\uff08\u9077\u79fb\u5b78\u7fd2\u3001\u751f\u6210\u5c0d\u6297\u7db2\u8def\uff09\u3001\u4ee5\u53ca\u5982\u4f55\u5c07\u6a21\u578b\u90e8\u7f72\u5230\u771f\u5be6\u4e16\u754c\u3002<\/p>\n\n\n\n
\u73fe\u5728\uff0c\u4f60\u5df2\u7d93\u64c1\u6709\u4e86\u52a0\u901f\u958b\u767c\u7684\u9470\u5319\u3002\u6253\u958b\u4f60\u7684\u7de8\u8f2f\u5668\uff0c\u958b\u59cb\u69cb\u5efa\u5c6c\u65bc\u4f60\u7684\u667a\u80fd\u672a\u4f86\u5427\uff01\uff01<\/p>\n\n\n\n
\n\n\n\n<\/p>\n","protected":false},"excerpt":{"rendered":"
\u4eba\u5de5\u667a\u6167\u7684\u767c\u5c55\u5728\u8fd1\u5e74\u4f86\u7a81\u98db\u731b\u9032\uff0c\u5f9e\u81ea\u7136\u8a9e\u8a00\u8655\u7406\u5230\u96fb\u8166\u8996\u89ba\uff0c\u5404\u500b\u9818\u57df\u90fd\u53d6\u5f97\u4e86\u9a5a\u4eba\u7684\u6210\u5c31\u3002\u800c\u652f\u6490\u9019\u5834\u9769\u547d\u7684\uff0c\u9664\u4e86\u5f37 […]<\/p>\n","protected":false},"author":1,"featured_media":3955,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3,87],"tags":[27,21,44,9,86,11],"class_list":["post-3948","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-python","category-python-ai","tag-c","tag-ide","tag-programming-language","tag-python","tag-rtos","tag-tutorial"],"blocksy_meta":[],"_links":{"self":[{"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/posts\/3948","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/comments?post=3948"}],"version-history":[{"count":16,"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/posts\/3948\/revisions"}],"predecessor-version":[{"id":3980,"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/posts\/3948\/revisions\/3980"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/media\/3955"}],"wp:attachment":[{"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/media?parent=3948"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/categories?post=3948"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/saludpcb.com\/zh\/wp-json\/wp\/v2\/tags?post=3948"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"PyTorch\"](\"https:\/\/saludpcb.com\/zh\/wp-content\/uploads\/2025\/09\/Unveiling-Python-and-PyTorch-2025-Supercharge-Your-AI-Journey-from-Scratch.jpg\" "\\"\\"")
<\/figure>\n\n\n\n