No. 144 of 500 No. 12 in Deep Learning

Deep Learning for Coders with fastai and PyTorch

AI Applications Without a PhD

By Jeremy Howard and Sylvain Gugger

Print length
621 pages
Publisher
O'Reilly Media
Reading level
Intermediate
ISBN-13
978-1492045526
Published
August 25, 2020
Written for
Python programmers who want to train and deploy deep learning models without a math-heavy background.

Synopsis

Jeremy Howard and Sylvain Gugger based this book on fast.ai's free online deep learning course, and it follows the same top-down method: readers train a working image classifier in the first chapter and only later open the hood. Early chapters cover computer vision, deploying a model as a web application, data ethics and the mechanics of the training loop. Middle chapters extend the approach to tabular data, collaborative filtering and natural language processing, including the ULMFiT transfer-learning method Howard co-created. The final chapters rebuild key pieces from scratch, from convolutions and optimizers to a complete neural network assembled from basic tensor operations.

The prose assumes Python fluency but little mathematics, and every chapter is available as a Jupyter notebook in the public fastbook repository, so readers can run and modify the code as they go. The book's lasting value lies in practical judgment: how to read a learning curve, pick a learning rate, spot overfitting and validate a model properly. Its main limitation is age. The language chapters center on recurrent networks rather than transformers, and the fastai library has continued to change, so some code needs adjusting. Readers gain intuition and experimental habits that transfer to any framework.

Scorecard

Editorial score

How it scored

Editorial score: 79.3 / 1005 weighted criteria

  • Influence30%

    24.0/30
  • Authority25%

    21.3/25
  • Practical value20%

    17.0/20
  • Currency15%

    9.0/15
  • Readability10%

    8.0/10
  • Editorial score

    79.3/100

How the rank was computed

Rank No. 144 of 500 comes from a composite of 59.9: observable signals 24.1 of 55 points plus editorial review 35.8 of 45 points.

All 11 inputs
InputPoints
Observable signals24.1/55
Reach (Amazon ratings volume)10.1/16.5
Consensus across best-of lists5.5/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted7.3/8.3
Amazon sales rank1.2/5.5
Recency0.0/2.8
Editorial review35.8/45
Author authority9.6/11.3
Influence10.8/13.5
Practical value7.7/9.0
Readability3.6/4.5
Currency4.1/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A proven code-first route into deep learning for working programmers, though its fastai-specific code has aged faster than its teaching.

Strengths

  • Runnable notebooks for every chapter
  • Top-down teaching that builds intuition quickly
  • Rebuilds core components from scratch

Keep in mind

Its language chapters center on recurrent networks rather than transformers, and code tied to the fastai library may need updates for current versions.

Key takeaways

  1. Start by training a working model, then study the theory behind it; top-down learning keeps motivation high and grounds abstractions in results.
  2. Transfer learning from pretrained models lets small teams reach strong accuracy with modest data and a single GPU.
  3. Practical skills such as finding a learning rate, augmenting data and reading validation loss matter more day to day than advanced mathematics.
  4. Practitioners carry ethical responsibility for deployed models, including feedback loops, bias and misuse, and should plan for these risks before release.

Notable facts

  • The companion fastbook notebook repository on GitHub has more than 25,000 stars. Source (opens in a new tab)
  • Peter Norvig, then Google's Director of Research, called it one of the best sources for a programmer to become proficient in deep learning. Source (opens in a new tab)

About the author

Jeremy Howard and Sylvain Gugger

Jeremy Howard co-founded fast.ai with Rachel Thomas, created the fastai library, co-created the ULMFiT language-model method and is founding CEO of Answer.AI; he was earlier president and chief scientist of Kaggle and founder of the medical AI company Enlitic. Sylvain Gugger worked at fast.ai on the library, then became a core maintainer of Hugging Face Transformers and creator of its Accelerate library.

Questions

Is Deep Learning for Coders with fastai and PyTorch still worth reading in 2026?
For learning how to train models, yes. Its lessons on transfer learning, learning rates, overfitting, data augmentation and model evaluation still apply in any framework. Its weaknesses come with age: the language chapters use recurrent networks rather than transformers, and some fastai code may need updating for current library versions. Pair it with a newer text on large language models for current architectures.
Do I need a math background to read Deep Learning for Coders?
No advanced math is required. The book is written for programmers comfortable with Python, and it introduces the necessary calculus and linear algebra through code as each concept arises. Readers start by training working models and study the mechanics in detail later. Some familiarity with array programming and Jupyter notebooks makes the early chapters easier, but the authors explain most tools when they first appear.
How does the book relate to the fast.ai course?
The book is based on fast.ai's free online deep learning course, and its chapters are available as Jupyter notebooks in the public fastbook repository on GitHub. Many readers work through the notebooks alongside the course videos. The printed edition offers polished prose in one structured sequence, which some learners prefer over video lessons, and it remains a convenient reference after the course ends.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.