{"id":25581,"date":"2024-10-13T07:49:19","date_gmt":"2024-10-12T22:49:19","guid":{"rendered":"http:\/\/www.tyosuke20xx.com\/blog\/?p=25581"},"modified":"2025-01-06T17:56:44","modified_gmt":"2025-01-06T08:56:44","slug":"%e7%b0%a1%e5%8d%98%e3%81%aa%e7%94%bb%e5%83%8f%e7%94%9f%e6%88%90ai-python","status":"publish","type":"post","link":"http:\/\/www.tyosuke20xx.com\/blog\/?p=25581","title":{"rendered":"\u7c21\u5358\u306a\u753b\u50cf\u751f\u6210AI python"},"content":{"rendered":"\n<pre class=\"wp-block-code\"><code>import torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torchvision import datasets, transforms\nfrom torch.utils.data import DataLoader\nfrom torchvision.utils import save_image  # \u753b\u50cf\u4fdd\u5b58\u306e\u305f\u3081\u306e\u95a2\u6570\nimport os  # \u30d5\u30a1\u30a4\u30eb\u306e\u4fdd\u5b58\u5148\u3092\u6307\u5b9a\u3059\u308b\u305f\u3081\u306e\u30e9\u30a4\u30d6\u30e9\u30ea\n\n# VAE\u306e\u30a8\u30f3\u30b3\u30fc\u30c0\u30fc\u3068\u30c7\u30b3\u30fc\u30c0\u30fc\nclass VAE(nn.Module):\n    def __init__(self, input_dim=784, hidden_dim=400, latent_dim=20):\n        super(VAE, self).__init__()\n        self.fc1 = nn.Linear(input_dim, hidden_dim)\n        self.fc21 = nn.Linear(hidden_dim, latent_dim)\n        self.fc22 = nn.Linear(hidden_dim, latent_dim)\n        self.fc3 = nn.Linear(latent_dim, hidden_dim)\n        self.fc4 = nn.Linear(hidden_dim, input_dim)\n\n    def encode(self, x):\n        h1 = torch.relu(self.fc1(x))\n        return self.fc21(h1), self.fc22(h1)\n\n    def reparameterize(self, mu, logvar):\n        std = torch.exp(0.5 * logvar)\n        eps = torch.randn_like(std)\n        return mu + eps * std\n\n    def decode(self, z):\n        h3 = torch.relu(self.fc3(z))\n        return torch.sigmoid(self.fc4(h3))\n\n    def forward(self, x):\n        mu, logvar = self.encode(x.view(-1, 784))\n        z = self.reparameterize(mu, logvar)\n        return self.decode(z), mu, logvar\n\n# VAE\u306e\u640d\u5931\u95a2\u6570\ndef loss_function(recon_x, x, mu, logvar):\n    BCE = nn.functional.binary_cross_entropy(recon_x, x.view(-1, 784), reduction='sum')\n    KLD = -0.5 * torch.sum(1 + logvar - mu.pow(2) - logvar.exp())\n    return BCE + KLD\n\n# \u30e2\u30c7\u30eb\u306e\u5b9a\u7fa9\u3068\u30c7\u30fc\u30bf\u30ed\u30fc\u30c0\u30fc\u306e\u6e96\u5099\ndevice = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\nvae = VAE().to(device)\noptimizer = optim.Adam(vae.parameters(), lr=1e-3)\n\ntransform = transforms.ToTensor()\ntrain_dataset = datasets.MNIST('.', train=True, download=True, transform=transform)\ntrain_loader = DataLoader(train_dataset, batch_size=128, shuffle=True)\n\n# \u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30eb\u30fc\u30d7\nvae.train()\nfor epoch in range(10):  # 10\u30a8\u30dd\u30c3\u30af\u3067\u5b66\u7fd2\n    train_loss = 0\n    for batch_idx, (data, _) in enumerate(train_loader):\n        data = data.to(device)\n        optimizer.zero_grad()\n        recon_batch, mu, logvar = vae(data)\n        loss = loss_function(recon_batch, data, mu, logvar)\n        loss.backward()\n        train_loss += loss.item()\n        optimizer.step()\n    \n    print(f'Epoch {epoch + 1}, Loss: {train_loss \/ len(train_loader.dataset)}')\n\n# \u5b66\u7fd2\u3057\u305f\u30e2\u30c7\u30eb\u3067\u753b\u50cf\u751f\u6210\nvae.eval()\nwith torch.no_grad():\n    z = torch.randn(64, 20).to(device)\n    sample = vae.decode(z).cpu()\n    sample = sample.view(64, 1, 28, 28)  # 28x28\u306e\u753b\u50cf\u30b5\u30a4\u30ba\u306b\u5909\u63db (MNIST\u30c7\u30fc\u30bf\u306e\u30d5\u30a9\u30fc\u30de\u30c3\u30c8)\n    \n    # \u4fdd\u5b58\u5148\u30c7\u30a3\u30ec\u30af\u30c8\u30ea\u3092\u6307\u5b9a\n    os.makedirs('generated_images', exist_ok=True)\n    save_image(sample, 'generated_images\/sample.png')\n\n    print(\"\u753b\u50cf\u751f\u6210\u5b8c\u4e86: 'generated_images\/sample.png' \u306b\u4fdd\u5b58\u3055\u308c\u307e\u3057\u305f\")\n<\/code><\/pre>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"_uf_show_specific_survey":0,"_uf_disable_surveys":false,"footnotes":""},"categories":[4,65],"tags":[3,39],"class_list":["post-25581","post","type-post","status-publish","format-standard","hentry","category-programming","category-python","tag-programming","tag-python"],"aioseo_notices":[],"jetpack_featured_media_url":"","_links":{"self":[{"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/25581","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=25581"}],"version-history":[{"count":2,"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/25581\/revisions"}],"predecessor-version":[{"id":25774,"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=\/wp\/v2\/posts\/25581\/revisions\/25774"}],"wp:attachment":[{"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=25581"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=25581"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.tyosuke20xx.com\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=25581"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}