D2l Zh
《动手学深度学习》：每个小节都是可运行的 Jupyter 记事本，含 PyTorch、TensorFlow、PaddlePaddle、MXNet 四种框架实现的深度学习教科书。
想扎实打牢深度学习基础的中文读者，免费方案里很难找到比它更划算的：概念、公式、图示、可跑代码四位一体，ResNet 到 Transformer 都有从零实现，第二版还补齐了 PaddlePaddle 四框架对照。代价是内容止步于 BERT 与 Transformer，扩散模型、LLM 微调这类后续方向没覆盖，代码基于较早框架版本，跑起来可能要自行适配。当基础教材用，别指望它带你追前沿。
Verified 12h ago · liveness 76/100 · cite: rightaichoice.com/tools/d2l-zh
- 希望从零动手实践、边跑边学的中文深度学习初学者
- 需要用可运行代码补齐实现能力的机器学习工程师
- 把本书用作课程教材或参考书的高校教师
- 使用 PyTorch、TensorFlow 或 PaddlePaddle 技术栈的中文开发者
- 追踪扩散模型、LLM 微调、RLHF 等较新方向的研究者
- 偏好纯理论推导、不想跑代码的读者
- 不使用 Python 或主流深度学习框架的开发者
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip D2L if you need material that covers LLM fine-tuning, diffusion models or RLHF, since the book's content ends at BERT and Transformer.
在线版免费，16 章内容、四框架实现、在线文档与 PDF 下载都不收费；只有想拿纸质教材或用作课程课本时才需要花钱——第二版《动手学深度学习（PyTorch版）》黑白平装版约 ¥100，第一版全彩精装约 ¥100-200。相比动辄数百美元的付费课程与认证，它的学习成本几乎压到了最低。
In short
D2l Zh — 《动手学深度学习》：每个小节都是可运行的 Jupyter 记事本，含 PyTorch、TensorFlow、PaddlePaddle、MXNet 四种框架实现的深度学习教科书。. Best for 希望从零动手实践、边跑边学的中文深度学习初学者, 需要用可运行代码补齐实现能力的机器学习工程师, 把本书用作课程教材或参考书的高校教师. Free to start; it also has paid plans, priced in a currency we have not confirmed — see the pricing table for the vendor’s own figures.
What people actually say about D2l Zh — is it worth it?
We scanned public community sources for D2l Zh on Jul 3, 2026 and could not establish that the discussion we found is about this tool rather than something else sharing its name. Our own analysis of that scan says the posts were off-subject. Rather than publish a sentiment score built on the wrong subject, we publish nothing here and re-run the scan.
Viability Score
How well maintained and how widely used is D2l Zh? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- 每小节一个可运行的 Jupyter 记事本，代码与超参数可自由修改
- 支持 PyTorch、NumPy/MXNet、TensorFlow、PaddlePaddle 四种框架实现
- 同一算法提供“从零实现”与“简洁实现”双轨代码对照
- Kaggle 实战：预测房价、CIFAR-10 图像分类、狗的品种识别
- 现代卷积网络章节：AlexNet、VGG、NiN、GoogLeNet、ResNet、DenseNet
- 现代循环网络章节：GRU、LSTM、深度双向 RNN、seq2seq、束搜索
- 注意力机制章节：Bahdanau 注意力、多头注意力、自注意力、Transformer
- NLP 章节：word2vec、GloVe、BERT 预训练与微调、情感分析、自然语言推断
- 优化算法章节：SGD、动量法、AdaGrad、RMSProp、Adadelta、Adam、学习率调度器
- 计算性能章节：异步计算、自动并行、多 GPU 训练、参数服务器
- 计算机视觉章节：图像增广、微调、锚框、SSD、R-CNN 系列、语义分割、风格迁移
- 免费在线文档与 PDF 下载，配套课件、作业与教学视频
- 可在本地、Amazon SageMaker Studio Lab、Amazon SageMaker、Google Colab 运行
- 每章末尾附社区讨论链接，200 多位贡献者持续维护
About D2l Zh
《动手学深度学习》（D2L）是一本面向中文读者、能运行可讨论的深度学习教科书，最鲜明的设计是全书 16 章的每一小节都是一个可运行的 Jupyter 记事本——代码可改、超参数可调、结果立刻可见，而不是只推导公式。当前提供 PyTorch、NumPy/MXNet、TensorFlow 和 PaddlePaddle 四种实现，第二版在线内容新增 PaddlePaddle；多数章节同时给出“从零实现”与“简洁实现”两套写法，便于对照。 内容从数据操作、线性回归、softmax 回归、多层感知机起步，经卷积神经网络（LeNet、AlexNet、VGG、NiN、GoogLeNet、ResNet、DenseNet）与循环神经网络（GRU、LSTM、深度双向 RNN、seq2seq、束搜索），讲到注意力机制与 Transformer，另有优化算法和计算性能两章，并用预测房价、CIFAR-10、狗的品种识别三个 Kaggle 实战收束。 运行环境无需自配：本地、Amazon SageMaker Studio Lab、Amazon SageMaker、Google Colab 都能跑。配套免费资源包括课件、作业与 B 站教学视频；官方称本书中英文版被全球 70 多个国家 500 多所大学用作教材或参考书，社区有 200 多位贡献者。 定位上，它比吴恩达《深度学习专项课程》更偏动手编码，比《深度学习》（花书）上手更平缓；与 PyTorch 官方教程相比又成体系得多。想要系统、可执行、中文讲解的基础教材，它是最省心的起点之一。
Behind the Verdict
我们会在两种情况下直接推荐它：一是零基础想真正跑通第一个神经网络，二是工程师想补齐“知道原理但没写过”的短板。全书没有把公式和代码割裂，反而强制你打开记事本改一行超参数再看结果，这种反馈回路是大多数视频课给不了的。 什么时候该绕过它？如果你的目标是微调大模型、做 RLHF 或研究扩散模型，这本书帮不上忙——章节结构停在 BERT 与 Transformer 的预训练/微调，之后的方向需要另找资料。同样，想要按 API 索引快速查某个函数用法的人，会嫌它按教学顺序而非字典顺序组织。 和吴恩达的《深度学习专项课程》比，D2L 更重手写实现，适合已经会一点 Python 的人；和花书比，它把数学门槛压低了不少，代价是理论深度没那么极致。真正想两样都要的读者，通常是先过 D2L 再回头啃花书。 实际使用时有几个坑要提前知道：代码基于较早的框架版本，直接照抄最新 PyTorch/TensorFlow 可能报错，得自己适配；GPU 章节和多 GPU 训练部分对没机器的人只能读不能练，用 Colab 或 SageMaker Studio Lab 免费额度能缓解。 如果只把它当电子书翻，收益会大打折扣。它真正的价值在于“每一小节都能跑”这条设计约束——你改了代码、看到损失曲线动了，才算学进去。对高校教师来说，配套课件和作业省下的备课时间也值得一提。
Researching D2l Zh? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas D2l Zh actually fits — and what changes day-one when you adopt it.
打开在线版第 3 章，在 Google Colab 里运行线性回归的从零实现，改学习率观察损失曲线变化
Outcome: 在不装环境的前提下跑通第一个模型，理解训练循环与梯度下降的实际行为
对照第 10 章的从零实现与简洁实现，拆解多头注意力与自注意力的矩阵维度，再迁移到自己的代码里
Outcome: 补上此前只会调 API、不清楚内部实现的知识缺口
取用中文版课件与作业，按第 6-9 章组织一学期课程，让学生在 SageMaker Studio Lab 上做实验
Outcome: 省去从零编写教材与实验环境的时间，课程直接可用
Use Cases
- 从零学习深度学习基础概念，并在 Jupyter 记事本里动手验证
- 用 PyTorch 或 PaddlePaddle 实现卷积神经网络做图像分类
- 跟随第 10 章拆解 Transformer 与自注意力机制
- 用第 13 章的 Kaggle 项目（CIFAR-10、狗的品种识别）练实战
- 高校教师取用课件与作业设计深度学习课程
- 在 Colab 或 SageMaker Studio Lab 上不加配置直接跑通全书代码
Limitations
内容覆盖基础到中阶，止于 Transformer 与 BERT，对 2023 年后的 LLM 微调、扩散模型等方向几乎没有涉及。代码基于较早的框架版本，与最新 API 可能不完全兼容，需要自行适配。中文版与英文版并行但中文更新优先。作为教材，它缺少实时反映最新研究进展的机制——你拿到的是经过整理的教学体系，而不是前沿追踪。另外，书中的实现是教学版本，直接照搬到生产环境需要自行重写数据管道与分布式部分。
as of 2026-10-11
Verification history
We have re-verified D2l Zh 10 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 10 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published D2l Zh tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
在线版（第二版）
$0/mo
第二版纸质书《动手学深度学习（PyTorch版）》（黑白平装）
约 ¥100
第一版纸质书（全彩精装 / 黑白平装）
约 ¥100-200
Where the pricing makes sense
The company stage and team size where D2l Zh's pricing actually pencils out — and where peers do it cheaper.
在线版免费，16 章内容、四框架实现、在线文档与 PDF 下载都不收费；只有想拿纸质教材或用作课程课本时才需要花钱——第二版《动手学深度学习（PyTorch版）》黑白平装版约 ¥100，第一版全彩精装约 ¥100-200。相比动辄数百美元的付费课程与认证，它的学习成本几乎压到了最低。
Setup time & first value
How long it actually takes to get something useful out of D2l Zh — broken out by persona, not the marketing-page minute.
在线版零安装：打开 Colab 或 SageMaker Studio Lab 直接运行任意小节，几分钟内可看到第一个训练结果。本地运行需自行装 Python 与框架（PyTorch/MXNet/TensorFlow/PaddlePaddle 之一）及 d2l 包，通常半小时到一小时。作为课程教材，教师按章节组织一学期内容，取用现成课件与作业可省去大部分备课时间。
Switching to or from D2l Zh
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →从吴恩达《深度学习专项课程》：偏理论的课程看完后转到 D2L，用可运行代码把同一批概念落地
- →从《深度学习》（花书）：理论推导读完转来练实现，第 4 章多层感知机与第 6 章 CNN 起点一致
- ↗到前沿方向：学完第 14-15 章 BERT 后，转向 LLM 微调、扩散模型等 2023 年后的专门资料
- ↗到框架官方文档：需要按 API 索引查函数用法时，改用 PyTorch/TensorFlow/PaddlePaddle 官方文档
Integrations
Resources & Guides
- Resourcezh.d2l.ai
Home · D2l Zh
Helpful link from zh.d2l.ai
- Resourcezh.d2l.ai
Index · D2l Zh
Helpful link from zh.d2l.ai
- Resourcezh.d2l.ai
Index · D2l Zh
Helpful link from zh.d2l.ai
- Resourcezh.d2l.ai
Index · D2l Zh
Helpful link from zh.d2l.ai
- Resourcezh.d2l.ai
Jupyter · D2l Zh
Helpful link from zh.d2l.ai
- Resourcezh.d2l.ai
Sagemaker · D2l Zh
Helpful link from zh.d2l.ai
- Resourcezh.d2l.ai
D2l Api · D2l Zh
Helpful link from zh.d2l.ai
- Resourcezh.d2l.ai
Contributing · D2l Zh
Helpful link from zh.d2l.ai
Tutorials & Learning
YouTube returned 6 videos for “D2l Zh”, and we withheld 6: 6 could not be judged, because “D2l Zh” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about D2l Zh.
Official links
Tools that pair well with D2l Zh
Common stack mates teams adopt alongside D2l Zh, with the specific reason each pairing earns its keep.
LLMs From Scratch
Build a large language model from scratch in PyTorch — a code-first book that walks you through coding a GPT-2 class LLM line by line
Ailearning
ApacheCN 维护的免费开源中文 AI 学习资源库，覆盖 Python、机器学习、深度学习与 NLP 全栈教程。
Kimi Chat
Kimi Chat is Moonshot AI's agentic workspace where deep research, slides, sheets and sites come back as finished deliverables.
Featured Head-to-Head Comparisons
D2l Zh vs Surge Ai
D2l Zh and Surge AI serve diametrically opposite needs. D2l Zh is a free, comprehensive textbook for learning deep learning hands-on, ideal for students and self-learners. Surge AI is a paid enterprise platform that provides expert human feedback and advanced benchmarks for frontier AI alignment. Choose D2l if you want to learn; choose Surge if you need to rigorously evaluate or align cutting-edge models.
D2l Zh vs Praktika
Praktika and D2L Zh serve completely different needs: one is a mobile AI-tutor app for spoken language practice, the other is a free interactive textbook for deep learning. Choose Praktika if you want to improve conversational fluency in a new language via AI tutors; choose D2L Zh if you're a beginner or intermediate learner looking to understand and implement deep learning models with hands-on code. There is no direct competition.
Alternatives to D2l Zh
View all
LLMs From Scratch
Build a large language model from scratch in PyTorch — a code-first book that walks you through coding a GPT-2 class LLM line by line
Ailearning
ApacheCN 维护的免费开源中文 AI 学习资源库，覆盖 Python、机器学习、深度学习与 NLP 全栈教程。
Frequently Asked Questions
Categories
Best-of guides
Topics
Used D2l Zh? Help shape our editorial sentiment research.