File size: 35,275 Bytes
d8dd320 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 | <!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Transformer架构交互式教学</title>
<script src="https://cdn.tailwindcss.com"></script>
<script src="https://d3js.org/d3.v7.min.js"></script>
<script src="https://cdn.plot.ly/plotly-2.24.1.min.js"></script>
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.4.0/css/all.min.css">
<script>
tailwind.config = {
theme: {
extend: {
colors: {
primary: '#4F46E5',
secondary: '#818CF8',
accent: '#F472B6',
dark: '#1E293B',
light: '#F1F5F9'
}
}
}
}
</script>
<style>
.attention-node {
transition: all 0.3s ease;
cursor: pointer;
}
.attention-node:hover {
stroke-width: 3px;
stroke: #F472B6;
}
.attention-line {
stroke-opacity: 0.3;
transition: stroke-opacity 0.3s ease;
}
.attention-line:hover {
stroke-opacity: 0.8;
}
.token-highlight {
background-color: rgba(79, 70, 229, 0.2);
transition: all 0.3s ease;
}
.fade-in {
animation: fadeIn 0.5s ease-in;
}
@keyframes fadeIn {
from { opacity: 0; transform: translateY(10px); }
to { opacity: 1; transform: translateY(0); }
}
.architecture-diagram {
background-image: linear-gradient(to right, #F1F5F9 1px, transparent 1px),
linear-gradient(to bottom, #F1F5F9 1px, transparent 1px);
background-size: 30px 30px;
}
</style>
</head>
<body class="bg-light text-dark font-sans">
<!-- 导航栏 -->
<nav class="bg-primary text-white shadow-lg">
<div class="container mx-auto px-4 py-3 flex justify-between items-center">
<div class="flex items-center space-x-2">
<i class="fas fa-brain text-2xl"></i>
<h1 class="text-xl font-bold">Transformer架构交互式教学</h1>
</div>
<div class="flex space-x-4">
<a href="#architecture" class="hover:text-secondary transition">架构</a>
<a href="#attention" class="hover:text-secondary transition">注意力机制</a>
<a href="#generation" class="hover:text-secondary transition">文本生成</a>
<a href="#deployment" class="hover:text-secondary transition">部署指南</a>
</div>
</div>
</nav>
<!-- 介绍部分 -->
<section class="py-16 bg-gradient-to-br from-primary to-secondary text-white">
<div class="container mx-auto px-4 text-center">
<h2 class="text-4xl font-bold mb-6">探索Transformer架构的奥秘</h2>
<p class="text-xl max-w-3xl mx-auto mb-8">
通过交互式可视化,直观理解大语言模型的核心原理 - Transformer架构和注意力机制
</p>
<div class="flex justify-center space-x-4">
<a href="#architecture" class="bg-white text-primary px-6 py-3 rounded-lg font-semibold hover:bg-opacity-90 transition">开始探索</a>
<a href="#generation" class="bg-accent text-white px-6 py-3 rounded-lg font-semibold hover:bg-opacity-90 transition">体验文本生成</a>
</div>
</div>
</section>
<!-- Transformer架构可视化 -->
<section id="architecture" class="py-16 bg-white">
<div class="container mx-auto px-4">
<h2 class="text-3xl font-bold text-center mb-4">Transformer架构解析</h2>
<p class="text-center text-gray-600 max-w-3xl mx-auto mb-12">
Transformer是当今大语言模型的核心架构,由编码器和解码器组成,通过自注意力机制处理序列数据
</p>
<div class="grid grid-cols-1 lg:grid-cols-2 gap-8 items-center">
<div>
<div class="architecture-diagram bg-white rounded-xl shadow-lg p-4 h-96" id="transformer-viz"></div>
</div>
<div>
<div class="bg-gray-50 rounded-xl p-6 shadow">
<h3 class="text-xl font-semibold mb-4">核心组件</h3>
<ul class="space-y-3">
<li class="flex items-start">
<i class="fas fa-cube text-secondary mt-1 mr-3"></i>
<div>
<strong>输入嵌入层</strong> - 将单词转换为向量表示
</div>
</li>
<li class="flex items-start">
<i class="fas fa-project-diagram text-secondary mt-1 mr-3"></i>
<div>
<strong>位置编码</strong> - 为输入添加位置信息
</div>
</li>
<li class="flex items-start">
<i class="fas fa-bullseye text-accent mt-1 mr-3"></i>
<div>
<strong>多头注意力</strong> - 模型的核心机制,计算单词间相关性
</div>
</li>
<li class="flex items-start">
<i class="fas fa-layer-group text-secondary mt-1 mr-3"></i>
<div>
<strong>前馈神经网络</strong> - 对注意力结果进行非线性变换
</div>
</li>
<li class="flex items-start">
<i class="fas fa-code-branch text-secondary mt-1 mr-3"></i>
<div>
<strong>残差连接和层归一化</strong> - 稳定训练过程
</div>
</li>
</ul>
<div class="mt-8">
<h4 class="font-semibold mb-2">探索架构</h4>
<div class="flex flex-wrap gap-2">
<button onclick="highlightComponent('encoder')" class="bg-secondary text-white px-3 py-1 rounded hover:bg-opacity-90 transition">编码器</button>
<button onclick="highlightComponent('decoder')" class="bg-accent text-white px-3 py-1 rounded hover:bg-opacity-90 transition">解码器</button>
<button onclick="highlightComponent('attention')" class="bg-primary text-white px-3 py-1 rounded hover:bg-opacity-90 transition">注意力层</button>
<button onclick="resetHighlight()" class="bg-gray-300 text-gray-700 px-3 py-1 rounded hover:bg-gray-400 transition">重置</button>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- 注意力机制可视化 -->
<section id="attention" class="py-16 bg-gray-50">
<div class="container mx-auto px-4">
<h2 class="text-3xl font-bold text-center mb-4">注意力机制详解</h2>
<p class="text-center text-gray-600 max-w-3xl mx-auto mb-12">
注意力机制让模型能够"关注"输入序列中不同部分之间的关系,是Transformer的核心创新
</p>
<div class="grid grid-cols-1 lg:grid-cols-2 gap-8">
<div class="bg-white rounded-xl shadow-lg p-4">
<h3 class="text-xl font-semibold mb-4">注意力权重可视化</h3>
<div id="attention-heatmap" class="h-80"></div>
<div class="mt-4">
<label class="block text-sm font-medium mb-2">输入文本</label>
<input type="text" id="attention-text" value="人工智能正在改变世界" class="w-full p-2 border rounded">
<button onclick="updateAttentionViz()" class="mt-3 bg-primary text-white px-4 py-2 rounded hover:bg-opacity-90 transition">计算注意力</button>
</div>
</div>
<div>
<div class="bg-white rounded-xl shadow-lg p-6 mb-6">
<h3 class="text-xl font-semibold mb-4">注意力机制如何工作</h3>
<p class="mb-4">注意力机制通过三个步骤计算单词之间的相关性:</p>
<ol class="list-decimal pl-6 space-y-2">
<li><strong>查询(Query)</strong> - 当前处理的单词</li>
<li><strong>键(Key)</strong> - 序列中的所有单词</li>
<li><strong>值(Value)</strong> - 每个键对应的内容</li>
</ol>
<div class="mt-4 p-4 bg-blue-50 rounded-lg">
<p class="font-mono text-sm">注意力分数 = Softmax(Query • Key<sup>T</sup> / √d<sub>k</sub>) • Value</p>
</div>
</div>
<div class="bg-white rounded-xl shadow-lg p-6">
<h3 class="text-xl font-semibold mb-4">多头注意力</h3>
<p>Transformer使用多个"注意力头",每个头学习不同的表示方式,使模型能够捕捉不同方面的关系。</p>
<div class="mt-4 flex justify-center">
<div class="relative">
<div class="w-12 h-12 bg-secondary rounded-full flex items-center justify-center text-white">头1</div>
<div class="w-12 h-12 bg-primary rounded-full flex items-center justify-center text-white absolute top-0 left-14">头2</div>
<div class="w-12 h-12 bg-accent rounded-full flex items-center justify-center text-white absolute top-0 left-28">头3</div>
<div class="w-12 h-12 bg-purple-500 rounded-full flex items-center justify-center text-white absolute top-0 left-42">头4</div>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- 文本生成体验 -->
<section id="generation" class="py-16 bg-white">
<div class="container mx-auto px-4">
<h2 class="text-3xl font-bold text-center mb-4">文本生成体验</h2>
<p class="text-center text-gray-600 max-w-3xl mx-auto mb-12">
体验Transformer模型如何生成文本,并可视化生成过程中的注意力机制
</p>
<div class="max-w-4xl mx-auto bg-gray-50 rounded-xl shadow-lg p-6">
<div class="grid grid-cols-1 md:grid-cols-3 gap-6 mb-6">
<div>
<label class="block text-sm font-medium mb-2">选择模型</label>
<select id="model-select" class="w-full p-2 border rounded">
<option value="gpt-mini">硅基流动-GPT-Mini</option>
<option value="gpt-medium">硅基流动-GPT-Medium</option>
</select>
</div>
<div>
<label class="block text-sm font-medium mb-2">生成长度</label>
<input type="range" id="length-slider" min="20" max="100" value="50" class="w-full">
<div class="text-center text-sm"><span id="length-value">50</span> tokens</div>
</div>
<div>
<label class="block text-sm font-medium mb-2">温度</label>
<input type="range" id="temp-slider" min="0.1" max="1.0" step="0.1" value="0.7" class="w-full">
<div class="text-center text-sm"><span id="temp-value">0.7</span></div>
</div>
</div>
<div class="mb-6">
<label class="block text-sm font-medium mb-2">输入提示</label>
<textarea id="prompt-input" class="w-full p-3 border rounded h-32">人工智能是</textarea>
</div>
<div class="flex justify-center mb-6">
<button id="generate-btn" class="bg-primary text-white px-6 py-3 rounded-lg font-semibold hover:bg-opacity-90 transition flex items-center">
<i class="fas fa-bolt mr-2"></i>生成文本
</button>
</div>
<div class="bg-white rounded-lg shadow p-4 mb-6">
<h3 class="font-semibold mb-2">生成结果</h3>
<div id="generated-text" class="min-h-32 p-3 bg-gray-100 rounded">
<p class="text-gray-500">点击"生成文本"按钮查看结果...</p>
</div>
</div>
<div class="bg-white rounded-lg shadow p-4">
<h3 class="font-semibold mb-2">生成过程可视化</h3>
<div id="generation-viz" class="h-64"></div>
</div>
</div>
</div>
</section>
<!-- 部署指南 -->
<section id="deployment" class="py-16 bg-gray-50">
<div class="container mx-auto px-4">
<h2 class="text-3xl font-bold text-center mb-4">Windows 11部署指南</h2>
<p class="text-center text-gray-600 max-w-3xl mx-auto mb-12">
面向小白的详细部署教程,轻松在Windows 11上运行此交互式教学网站
</p>
<div class="max-w-4xl mx-auto bg-white rounded-xl shadow-lg overflow-hidden">
<div class="grid grid-cols-1 md:grid-cols-2">
<div class="p-6">
<h3 class="text-xl font-semibold mb-4">部署步骤</h3>
<ol class="list-decimal pl-6 space-y-4">
<li>
<strong>下载网站文件</strong>
<p class="text-sm text-gray-600 mt-1">将本网页保存为HTML文件(右键点击 → 另存为)</p>
</li>
<li>
<strong>安装本地服务器(可选)</strong>
<p class="text-sm text-gray-600 mt-1">安装Node.js和http-server模块:<code class="bg-gray-100 p-1 rounded">npm install -g http-server</code></p>
</li>
<li>
<strong>运行网站</strong>
<p class="text-sm text-gray-600 mt-1">双击HTML文件直接打开,或在命令行中运行:<code class="bg-gray-100 p-1 rounded">http-server</code></p>
</li>
<li>
<strong>访问网站</strong>
<p class="text-sm text-gray-600 mt-1">在浏览器中访问 <code class="bg-gray-100 p-1 rounded">http://localhost:8080</code></p>
</li>
</ol>
</div>
<div class="bg-gray-800 text-white p-6">
<h3 class="text-xl font-semibold mb-4">常见问题</h3>
<div class="space-y-4">
<div>
<h4 class="font-medium">Q: 为什么需要本地服务器?</h4>
<p class="text-gray-300 text-sm mt-1">A: 直接打开HTML文件时,某些API调用可能被浏览器安全策略阻止。使用本地服务器可以避免此问题。</p>
</div>
<div>
<h4 class="font-medium">Q: 如何安装Node.js?</h4>
<p class="text-gray-300 text-sm mt-1">A: 访问 <a href="https://nodejs.org" class="text-secondary underline">nodejs.org</a> 下载安装包,按提示安装即可。</p>
</div>
<div>
<h4 class="font-medium">Q: 网站无法加载可视化?</h4>
<p class="text-gray-300 text-sm mt-1">A: 确保浏览器支持JavaScript,并检查网络连接是否正常。</p>
</div>
</div>
</div>
</div>
</div>
<div class="mt-12 max-w-4xl mx-auto">
<h3 class="text-xl font-semibold mb-4">视频教程</h3>
<div class="bg-gray-200 border-2 border-dashed rounded-xl w-full h-64 flex items-center justify-center">
<div class="text-center">
<i class="fas fa-video text-4xl text-gray-400 mb-3"></i>
<p class="text-gray-500">Windows 11部署视频教程</p>
</div>
</div>
</div>
</div>
</section>
<!-- 页脚 -->
<footer class="bg-dark text-white py-8">
<div class="container mx-auto px-4 text-center">
<p class="mb-4">Transformer架构交互式教学网站 | 帮助初学者理解大语言模型原理</p>
<div class="flex justify-center space-x-4">
<a href="#" class="hover:text-secondary transition"><i class="fab fa-github"></i></a>
<a href="#" class="hover:text-secondary transition"><i class="fab fa-twitter"></i></a>
<a href="#" class="hover:text-secondary transition"><i class="fab fa-linkedin"></i></a>
</div>
<p class="mt-4 text-gray-400 text-sm">© 2023 AI Education Project. 本网站仅用于教学目的。</p>
</div>
</footer>
<script>
// 初始化Transformer架构可视化
function initTransformerViz() {
const svg = d3.select("#transformer-viz")
.append("svg")
.attr("width", "100%")
.attr("height", "100%")
.attr("viewBox", "0 0 800 400");
// 绘制编码器
const encoderGroup = svg.append("g").attr("transform", "translate(100, 50)");
// 编码器标签
encoderGroup.append("text")
.text("编码器")
.attr("x", 0)
.attr("y", -10)
.attr("font-weight", "bold")
.attr("fill", "#4F46E5");
// 绘制编码器层
for (let i = 0; i < 3; i++) {
const layerGroup = encoderGroup.append("g").attr("transform", `translate(0, ${i * 100})`);
// 层矩形
layerGroup.append("rect")
.attr("id", `encoder-layer-${i}`)
.attr("x", 0)
.attr("y", 0)
.attr("width", 200)
.attr("height", 80)
.attr("rx", 10)
.attr("fill", i === 1 ? "#E0E7FF" : "#F1F5F9")
.attr("stroke", "#4F46E5")
.attr("stroke-width", 2);
// 层标签
layerGroup.append("text")
.text(`编码器层 ${i+1}`)
.attr("x", 100)
.attr("y", 40)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#4F46E5");
// 内部组件
layerGroup.append("rect")
.attr("x", 20)
.attr("y", 20)
.attr("width", 60)
.attr("height", 40)
.attr("rx", 5)
.attr("fill", i === 1 ? "#C7D2FE" : "#DBEAFE")
.attr("stroke", "#818CF8")
.attr("id", `attention-${i}`);
layerGroup.append("text")
.text("注意力")
.attr("x", 50)
.attr("y", 40)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#4F46E5")
.attr("font-size", "12px");
layerGroup.append("rect")
.attr("x", 120)
.attr("y", 20)
.attr("width", 60)
.attr("height", 40)
.attr("rx", 5)
.attr("fill", i === 1 ? "#C7D2FE" : "#DBEAFE")
.attr("stroke", "#818CF8");
layerGroup.append("text")
.text("前馈")
.attr("x", 150)
.attr("y", 40)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#4F46E5")
.attr("font-size", "12px");
}
// 绘制解码器
const decoderGroup = svg.append("g").attr("transform", "translate(500, 50)");
// 解码器标签
decoderGroup.append("text")
.text("解码器")
.attr("x", 0)
.attr("y", -10)
.attr("font-weight", "bold")
.attr("fill", "#F472B6");
// 绘制解码器层
for (let i = 0; i < 3; i++) {
const layerGroup = decoderGroup.append("g").attr("transform", `translate(0, ${i * 100})`);
// 层矩形
layerGroup.append("rect")
.attr("id", `decoder-layer-${i}`)
.attr("x", 0)
.attr("y", 0)
.attr("width", 200)
.attr("height", 80)
.attr("rx", 10)
.attr("fill", i === 1 ? "#FCE7F3" : "#FDF2F8")
.attr("stroke", "#F472B6")
.attr("stroke-width", 2);
// 层标签
layerGroup.append("text")
.text(`解码器层 ${i+1}`)
.attr("x", 100)
.attr("y", 40)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#F472B6");
// 内部组件
layerGroup.append("rect")
.attr("x", 20)
.attr("y", 20)
.attr("width", 50)
.attr("height", 40)
.attr("rx", 5)
.attr("fill", i === 1 ? "#FBCFE8" : "#FCE7F3")
.attr("stroke", "#F472B6");
layerGroup.append("text")
.text("自注意")
.attr("x", 45)
.attr("y", 40)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#F472B6")
.attr("font-size", "12px");
layerGroup.append("rect")
.attr("x", 85)
.attr("y", 20)
.attr("width", 50)
.attr("height", 40)
.attr("rx", 5)
.attr("fill", i === 1 ? "#FBCFE8" : "#FCE7F3")
.attr("stroke", "#F472B6");
layerGroup.append("text")
.text("编码注意")
.attr("x", 110)
.attr("y", 40)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#F472B6")
.attr("font-size", "10px");
layerGroup.append("rect")
.attr("x", 150)
.attr("y", 20)
.attr("width", 40)
.attr("height", 40)
.attr("rx", 5)
.attr("fill", i === 1 ? "#FBCFE8" : "#FCE7F3")
.attr("stroke", "#F472B6");
layerGroup.append("text")
.text("前馈")
.attr("x", 170)
.attr("y", 40)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#F472B6")
.attr("font-size", "12px");
}
// 绘制连接线
svg.append("line")
.attr("x1", 300)
.attr("y1", 200)
.attr("x2", 500)
.attr("y2", 200)
.attr("stroke", "#94A3B8")
.attr("stroke-width", 2)
.attr("marker-end", "url(#arrow)");
// 绘制输入输出
svg.append("rect")
.attr("x", 20)
.attr("y", 180)
.attr("width", 60)
.attr("height", 40)
.attr("rx", 5)
.attr("fill", "#D1FAE5")
.attr("stroke", "#10B981");
svg.append("text")
.text("输入")
.attr("x", 50)
.attr("y", 200)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#10B981");
svg.append("rect")
.attr("x", 720)
.attr("y", 180)
.attr("width", 60)
.attr("height", 40)
.attr("rx", 5)
.attr("fill", "#D1FAE5")
.attr("stroke", "#10B981");
svg.append("text")
.text("输出")
.attr("x", 750)
.attr("y", 200)
.attr("text-anchor", "middle")
.attr("dominant-baseline", "middle")
.attr("fill", "#10B981");
// 添加箭头标记
svg.append("defs").append("marker")
.attr("id", "arrow")
.attr("viewBox", "0 -5 10 10")
.attr("refX", 5)
.attr("refY", 0)
.attr("markerWidth", 6)
.attr("markerHeight", 6)
.attr("orient", "auto")
.append("path")
.attr("d", "M0,-5L10,0L0,5")
.attr("fill", "#94A3B8");
}
// 初始化注意力热力图
function initAttentionViz() {
const text = "人工智能正在改变世界";
const tokens = text.split('');
const data = [{
z: generateAttentionMatrix(tokens.length),
x: tokens,
y: tokens,
type: 'heatmap',
colorscale: 'Blues',
hoverongaps: false
}];
const layout = {
title: '注意力权重分布',
xaxis: { title: '键 (Key)' },
yaxis: { title: '查询 (Query)', autorange: 'reversed' },
height: 300,
margin: { l: 60, r: 20, t: 40, b: 60 }
};
Plotly.newPlot('attention-heatmap', data, layout);
}
// 生成模拟注意力矩阵
function generateAttentionMatrix(size) {
const matrix = [];
for (let i = 0; i < size; i++) {
const row = [];
for (let j = 0; j < size; j++) {
// 模拟对角线附近有较高权重
const dist = Math.abs(i - j);
let value = Math.exp(-dist * 0.5);
// 添加一些随机性
value += (Math.random() - 0.5) * 0.2;
value = Math.max(0, Math.min(1, value));
row.push(value);
}
matrix.push(row);
}
return matrix;
}
// 更新注意力可视化
function updateAttentionViz() {
const text = document.getElementById('attention-text').value || "人工智能正在改变世界";
const tokens = text.split('');
Plotly.update('attention-heatmap', {
z: [generateAttentionMatrix(tokens.length)],
x: [tokens],
y: [tokens]
});
}
// 高亮架构组件
function highlightComponent(component) {
d3.selectAll("rect").attr("stroke-width", 1);
if (component === 'encoder') {
d3.selectAll("#encoder-layer-1")
.attr("stroke-width", 3)
.attr("stroke", "#4F46E5");
} else if (component === 'decoder') {
d3.selectAll("#decoder-layer-1")
.attr("stroke-width", 3)
.attr("stroke", "#F472B6");
} else if (component === 'attention') {
d3.selectAll("[id^='attention-']")
.attr("stroke-width", 3)
.attr("stroke", "#818CF8");
}
}
// 重置高亮
function resetHighlight() {
d3.selectAll("rect").attr("stroke-width", 1);
}
// 文本生成函数
async function generateText() {
const prompt = document.getElementById('prompt-input').value;
const model = document.getElementById('model-select').value;
const length = document.getElementById('length-slider').value;
const temperature = document.getElementById('temp-slider').value;
if (!prompt.trim()) {
alert("请输入提示文本!");
return;
}
document.getElementById('generate-btn').innerHTML = '<i class="fas fa-spinner fa-spin mr-2"></i>生成中...';
try {
// 模拟API调用延迟
await new Promise(resolve => setTimeout(resolve, 1500));
// 模拟生成结果
const generatedText = prompt + "一项革命性技术,它通过模拟人类智能过程,使计算机能够执行复杂的任务。从语音助手到自动驾驶,AI正在渗透到我们生活的方方面面。";
// 显示生成结果
document.getElementById('generated-text').innerHTML =
`<p class="fade-in">${generatedText}</p>`;
// 更新生成过程可视化
updateGenerationViz(prompt, generatedText);
} catch (error) {
console.error("生成错误:", error);
document.getElementById('generated-text').innerHTML =
`<p class="text-red-500">生成失败: ${error.message}</p>`;
} finally {
document.getElementById('generate-btn').innerHTML = '<i class="fas fa-bolt mr-2"></i>生成文本';
}
}
// 更新生成过程可视化
function updateGenerationViz(prompt, fullText) {
const promptTokens = prompt.split('');
const allTokens = fullText.split('');
const data = [{
z: generateAttentionMatrix(allTokens.length),
x: allTokens,
y: allTokens,
type: 'heatmap',
colorscale: 'Blues',
hoverongaps: false
}];
const layout = {
title: '生成过程注意力权重',
xaxis: { title: '位置' },
yaxis: { title: '生成步骤', autorange: 'reversed' },
height: 250,
margin: { l: 50, r: 20, t: 40, b: 50 }
};
Plotly.newPlot('generation-viz', data, layout);
}
// 初始化页面
document.addEventListener('DOMContentLoaded', function() {
// 初始化可视化
initTransformerViz();
initAttentionViz();
// 设置滑块事件
document.getElementById('length-slider').addEventListener('input', function() {
document.getElementById('length-value').textContent = this.value;
});
document.getElementById('temp-slider').addEventListener('input', function() {
document.getElementById('temp-value').textContent = this.value;
});
// 设置生成按钮事件
document.getElementById('generate-btn').addEventListener('click', generateText);
// 平滑滚动
document.querySelectorAll('a[href^="#"]').forEach(anchor => {
anchor.addEventListener('click', function(e) {
e.preventDefault();
document.querySelector(this.getAttribute('href')).scrollIntoView({
behavior: 'smooth'
});
});
});
});
</script>
<p style="border-radius: 8px; text-align: center; font-size: 12px; color: #fff; margin-top: 16px;position: fixed; left: 8px; bottom: 8px; z-index: 10; background: rgba(0, 0, 0, 0.8); padding: 4px 8px;">Made with <img src="https://enzostvs-deepsite.hf.space/logo.svg" alt="DeepSite Logo" style="width: 16px; height: 16px; vertical-align: middle;display:inline-block;margin-right:3px;filter:brightness(0) invert(1);"><a href="https://enzostvs-deepsite.hf.space" style="color: #fff;text-decoration: underline;" target="_blank" >DeepSite</a> - 🧬 <a href="https://enzostvs-deepsite.hf.space?remix=ahepc/ai" style="color: #fff;text-decoration: underline;" target="_blank" >Remix</a></p></body>
</html> |