flykhan
2536c937e3
翻译自英文原版 maths-cs-ai-compendium，共 20 章全部完成。 第01章 向量 | 第02章 矩阵 | 第03章 微积分 第04章 统计学 | 第05章 概率论 | 第06章 机器学习 第07章 计算语言学 | 第08章 计算机视觉 | 第09章 音频与语音 第10章 多模态学习 | 第11章 自主系统 | 第12章 图神经网络 第13章 计算与操作系统 | 第14章 数据结构与算法 第15章 生产级软件工程 | 第16章 SIMD与GPU编程 第17章 AI推理 | 第18章 ML系统设计 第19章 应用人工智能 | 第20章 前沿人工智能 翻译说明： - 所有数学公式 $...$ / $$...$$、代码块、图片引用完整保留 - mkdocs.yml 配置中文导航 + language: zh - README.md 已翻译为中文（兼 docs/index.md） - docs/ 目录包含指向各章文件的 symlink - 约 29,000 行中文内容，排除 .cache/ 构建缓存
115 lines
7.2 KiB
XML
115 lines
7.2 KiB
XML
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 800 280" width="800" height="280" font-family="Arial, sans-serif">
|
|
<!-- Title -->
|
|
<text x="400" y="22" text-anchor="middle" font-size="14" font-weight="bold" fill="#333">Image Tokeniser Architectures</text>
|
|
|
|
<defs>
|
|
<marker id="cArr" markerWidth="7" markerHeight="5" refX="7" refY="2.5" orient="auto">
|
|
<path d="M0,0 L7,2.5 L0,5 Z" fill="#666"/>
|
|
</marker>
|
|
</defs>
|
|
|
|
<!-- ========= Column 1: dVAE (DALL-E) ========= -->
|
|
<rect x="15" y="36" width="240" height="236" rx="8" fill="#e74c3c" fill-opacity="0.04" stroke="#e74c3c" stroke-width="1" stroke-dasharray="4,2"/>
|
|
<text x="135" y="54" text-anchor="middle" font-size="12" font-weight="bold" fill="#e74c3c">dVAE (DALL-E)</text>
|
|
|
|
<!-- Input -->
|
|
<rect x="50" y="66" width="44" height="32" rx="4" fill="#3498db" opacity="0.4" stroke="#3498db" stroke-width="1"/>
|
|
<text x="72" y="86" text-anchor="middle" font-size="8" fill="#333">Input</text>
|
|
<!-- Arrow -->
|
|
<line x1="98" y1="82" x2="118" y2="82" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
<!-- Encoder -->
|
|
<rect x="124" y="66" width="56" height="32" rx="6" fill="#3498db" fill-opacity="0.12" stroke="#3498db" stroke-width="1.2"/>
|
|
<text x="152" y="86" text-anchor="middle" font-size="9" font-weight="bold" fill="#3498db">Encoder</text>
|
|
<!-- Arrow down -->
|
|
<line x1="152" y1="102" x2="152" y2="118" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- Gumbel-Softmax -->
|
|
<rect x="88" y="124" width="128" height="36" rx="6" fill="#f39c12" fill-opacity="0.12" stroke="#f39c12" stroke-width="1.5"/>
|
|
<text x="152" y="140" text-anchor="middle" font-size="9" font-weight="bold" fill="#f39c12">Gumbel-Softmax</text>
|
|
<text x="152" y="152" text-anchor="middle" font-size="7" fill="#f39c12">(differentiable sampling)</text>
|
|
<!-- Arrow down -->
|
|
<line x1="152" y1="164" x2="152" y2="178" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- Soft tokens -->
|
|
<rect x="104" y="184" width="96" height="26" rx="5" fill="#f39c12" fill-opacity="0.1" stroke="#f39c12" stroke-width="1"/>
|
|
<text x="152" y="200" text-anchor="middle" font-size="9" fill="#f39c12">Soft token probs</text>
|
|
<!-- Arrow down -->
|
|
<line x1="152" y1="214" x2="152" y2="228" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- Decoder -->
|
|
<rect x="124" y="234" width="56" height="26" rx="6" fill="#e74c3c" fill-opacity="0.12" stroke="#e74c3c" stroke-width="1.2"/>
|
|
<text x="152" y="251" text-anchor="middle" font-size="9" font-weight="bold" fill="#e74c3c">Decoder</text>
|
|
|
|
<!-- Highlight -->
|
|
<text x="135" y="274" text-anchor="middle" font-size="8" fill="#f39c12" font-style="italic">soft / differentiable</text>
|
|
|
|
<!-- ========= Column 2: VQ-GAN ========= -->
|
|
<rect x="280" y="36" width="240" height="236" rx="8" fill="#3498db" fill-opacity="0.04" stroke="#3498db" stroke-width="1" stroke-dasharray="4,2"/>
|
|
<text x="400" y="54" text-anchor="middle" font-size="12" font-weight="bold" fill="#3498db">VQ-GAN</text>
|
|
|
|
<!-- Input -->
|
|
<rect x="316" y="66" width="44" height="32" rx="4" fill="#3498db" opacity="0.4" stroke="#3498db" stroke-width="1"/>
|
|
<text x="338" y="86" text-anchor="middle" font-size="8" fill="#333">Input</text>
|
|
<!-- Arrow -->
|
|
<line x1="364" y1="82" x2="384" y2="82" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
<!-- Encoder -->
|
|
<rect x="390" y="66" width="56" height="32" rx="6" fill="#3498db" fill-opacity="0.12" stroke="#3498db" stroke-width="1.2"/>
|
|
<text x="418" y="86" text-anchor="middle" font-size="9" font-weight="bold" fill="#3498db">Encoder</text>
|
|
<!-- Arrow down -->
|
|
<line x1="418" y1="102" x2="418" y2="118" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- NN Codebook Lookup -->
|
|
<rect x="340" y="124" width="156" height="36" rx="6" fill="#f39c12" fill-opacity="0.12" stroke="#f39c12" stroke-width="1.5"/>
|
|
<text x="418" y="140" text-anchor="middle" font-size="9" font-weight="bold" fill="#f39c12">NN Codebook Lookup</text>
|
|
<text x="418" y="152" text-anchor="middle" font-size="7" fill="#f39c12">(nearest-neighbour, hard)</text>
|
|
<!-- Arrow down -->
|
|
<line x1="418" y1="164" x2="418" y2="178" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- Discrete tokens -->
|
|
<rect x="362" y="184" width="112" height="26" rx="5" fill="#27ae60" fill-opacity="0.12" stroke="#27ae60" stroke-width="1"/>
|
|
<text x="418" y="200" text-anchor="middle" font-size="9" fill="#27ae60" font-weight="bold">Discrete tokens</text>
|
|
<!-- Arrow down -->
|
|
<line x1="418" y1="214" x2="418" y2="228" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- Decoder + Discriminator -->
|
|
<rect x="370" y="234" width="96" height="26" rx="6" fill="#e74c3c" fill-opacity="0.12" stroke="#e74c3c" stroke-width="1.2"/>
|
|
<text x="418" y="248" text-anchor="middle" font-size="8" font-weight="bold" fill="#e74c3c">Decoder + Discrim.</text>
|
|
|
|
<!-- Highlight -->
|
|
<text x="400" y="274" text-anchor="middle" font-size="8" fill="#3498db" font-style="italic">hard quantisation + adversarial</text>
|
|
|
|
<!-- ========= Column 3: FSQ ========= -->
|
|
<rect x="545" y="36" width="240" height="236" rx="8" fill="#27ae60" fill-opacity="0.04" stroke="#27ae60" stroke-width="1" stroke-dasharray="4,2"/>
|
|
<text x="665" y="54" text-anchor="middle" font-size="12" font-weight="bold" fill="#27ae60">FSQ</text>
|
|
|
|
<!-- Input -->
|
|
<rect x="580" y="66" width="44" height="32" rx="4" fill="#3498db" opacity="0.4" stroke="#3498db" stroke-width="1"/>
|
|
<text x="602" y="86" text-anchor="middle" font-size="8" fill="#333">Input</text>
|
|
<!-- Arrow -->
|
|
<line x1="628" y1="82" x2="648" y2="82" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
<!-- Encoder -->
|
|
<rect x="654" y="66" width="56" height="32" rx="6" fill="#3498db" fill-opacity="0.12" stroke="#3498db" stroke-width="1.2"/>
|
|
<text x="682" y="86" text-anchor="middle" font-size="9" font-weight="bold" fill="#3498db">Encoder</text>
|
|
<!-- Arrow down -->
|
|
<line x1="682" y1="102" x2="682" y2="118" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- Round to fixed levels -->
|
|
<rect x="610" y="124" width="144" height="36" rx="6" fill="#27ae60" fill-opacity="0.12" stroke="#27ae60" stroke-width="1.5"/>
|
|
<text x="682" y="140" text-anchor="middle" font-size="9" font-weight="bold" fill="#27ae60">Round to Fixed Levels</text>
|
|
<text x="682" y="152" text-anchor="middle" font-size="7" fill="#27ae60">(e.g. [-2, -1, 0, 1, 2] per dim)</text>
|
|
<!-- Arrow down -->
|
|
<line x1="682" y1="164" x2="682" y2="178" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- Discrete tokens -->
|
|
<rect x="626" y="184" width="112" height="26" rx="5" fill="#27ae60" fill-opacity="0.12" stroke="#27ae60" stroke-width="1"/>
|
|
<text x="682" y="200" text-anchor="middle" font-size="9" fill="#27ae60" font-weight="bold">Discrete tokens</text>
|
|
<!-- Arrow down -->
|
|
<line x1="682" y1="214" x2="682" y2="228" stroke="#666" stroke-width="1" marker-end="url(#cArr)"/>
|
|
|
|
<!-- Decoder -->
|
|
<rect x="654" y="234" width="56" height="26" rx="6" fill="#e74c3c" fill-opacity="0.12" stroke="#e74c3c" stroke-width="1.2"/>
|
|
<text x="682" y="251" text-anchor="middle" font-size="9" font-weight="bold" fill="#e74c3c">Decoder</text>
|
|
|
|
<!-- Highlight -->
|
|
<text x="665" y="274" text-anchor="middle" font-size="8" fill="#27ae60" font-style="italic">no codebook needed</text>
|
|
</svg> |
