Practical Deep Learning for Coders review
fast.ai · fast-ai
Our fast.ai Practical Deep Learning review: Practical Deep Learning for Coders is the best free deep learning course for people who can already code and want to build models first. Jeremy Howard teaches 9 lessons of about 90 minutes each, and you deploy your own model by the end of lesson 2. It costs nothing, and there is no certificate.
What does it take?
| Difficulty | 4 out of 5 |
|---|---|
| Hours | About 13.5 hours |
| Format | Video, Code notebooks, Projects |
| Maths needed | School-level maths |
| Coding needed | Yes, you will write code |
| Hardware needed | Free cloud GPU (Google Colab or Kaggle Notebooks) |
| Certificate | No certificate |
| Free to audit | Yes |
| Pace | Self-paced |
| Last updated by the provider | Not checked yet |
| Teacher | Jeremy Howard (fast.ai) |
| Edition | 2022 part 1, recorded at the University of Queensland |
| Lessons | 9, each about 90 minutes |
| Prerequisite | 1 year of coding, preferably Python, plus high school maths |
| Cost | Free |
| Certificate | None |
| Software | PyTorch, fastai, Hugging Face Transformers, Gradio |
Who should take it, and who should skip it?
Take it if you are
- Coders with 1 year of experience, preferably in Python, who want to apply deep learning
- Learners who prefer building a working model first and learning the theory after
- Anyone with high school maths: the course teaches the calculus and linear algebra it needs
Skip it if you are
- Anyone who cannot code yet: fast.ai lists coding as the one prerequisite
- Learners who want a certificate: the course offers none
- Anyone who wants a current recording: this is the 2022 edition
What will you be able to do after it?
- Build and deploy your own deep learning model, on data you collect, by the end of lesson 2
- Train models for computer vision, NLP, tabular data and collaborative filtering
- Use random forests and regression alongside neural networks
- Deploy a model as a web app
- Write stochastic gradient descent and a full training loop from scratch
What's in each module?
| Module | What it covers |
|---|---|
| Lesson 1: Getting started | The first lesson, with hands-on notebook exercises, mostly on Kaggle. |
| Lesson 2: Deployment | By the end you have built and deployed your own model on data you collect. |
| Lesson 3: Neural net foundations | Neural net foundations. The course teaches the calculus and linear algebra you need. |
| Lesson 4: Natural language (NLP) | Natural language processing. |
| Lesson 5: From-scratch model | A from-scratch model. |
| Lesson 6: Random forests | Random forests. |
| Lesson 7: Collaborative filtering | Collaborative filtering. |
| Lesson 8: Convolutions (CNNs) | Convolutions, also known as CNNs. |
| Bonus: Data ethics | A bonus lesson on data ethics. |
What we liked, and what we didn't
What we liked
- It costs nothing, and the book it is based on is free online
- Every lesson has hands-on exercises, mostly in notebooks on Kaggle
- You deploy your own model by the end of lesson 2
- It needs no university maths: the course teaches the calculus and linear algebra you need
- Videos are captioned, and help is on forums.fast.ai
What we didn't
- It is the 2022 part 1 recording, so it predates later releases
- There is no certificate
- 13.5 hours of video understates the time: fast.ai says the notebooks are the point of the course
- fast.ai advises against training on your own computer unless you can handle GPU drivers
- You need to code already: fast.ai asks for about a year of experience
What's the catch?
This is the 2022 recording, and the course offers no certificate. The 9 lessons run about 13.5 hours of video, but the real time is well above that, because fast.ai says that if you do not work through the notebooks yourself you will not get nearly as much out of the course. It also assumes you can already code. Fast.ai describes a separate part 2 of more than 30 hours, Deep Learning Foundations to Stable Diffusion, for more experienced students, so treat this course as the first step and not the whole subject.
Is it still current?
The course page was checked on 5 October 2026. It lists Practical Deep Learning for Coders 2022 part 1, recorded at the University of Queensland, and uses PyTorch, fastai, Hugging Face Transformers and Gradio. The course is based on fast.ai's book, which is freely available online, and each lesson goes with book chapters that are all available as Jupyter notebooks.
What should you take before it?
- Machine Learning Crash Course
Google's Machine Learning Crash Course has 12 modules and takes about 13.5 hours by Google's own timings. It is free, with quiz badges and no certificate.
- CS50's Introduction to AI with Python
CS50's Introduction to AI with Python runs 7 weeks with 12 projects, costs nothing and gives a free CS50 Certificate if you score 70% on every project.
What should you take after it?
- Hugging Face LLM Course
Hugging Face LLM Course has 12 chapters of 6 to 8 hours each, costs nothing and has no certificate. You learn LLMs with the Transformers library.
- Neural Networks: Zero to Hero
Neural Networks: Zero to Hero is 8 free YouTube videos, about 14.5 hours, where Andrej Karpathy builds a GPT from scratch in code. No certificate.
How we rated it: 9.1 out of 10
| Criterion | Score out of 10 |
|---|---|
| Clarity | 8 |
| Hands-on practice | 10 |
| Up to date | 8 |
| Depth | 8 |
| Truly free | 10 |
Questions people ask
Is Practical Deep Learning for Coders free?
Yes. fast.ai describes it as a free course, and the book it is based on is freely available online. The course offers no certificate.
How long does Practical Deep Learning for Coders take?
There are 9 lessons of about 90 minutes each, so about 13.5 hours of video. Fast.ai says you need to do the notebook exercises too, so plan for well more.
What do you need before Practical Deep Learning for Coders?
You need to know how to code, and a year of experience is enough, preferably in Python. You also need high school maths. No university maths is needed.
Do you need a GPU for Practical Deep Learning for Coders?
You do not need special hardware. The course mainly uses Kaggle Notebooks and Paperspace Gradient, and fast.ai strongly suggests not training on your own computer.
Which edition is Practical Deep Learning for Coders?
The course page lists the 2022 part 1 edition, recorded at the University of Queensland. A separate part 2 of more than 30 hours, Deep Learning Foundations to Stable Diffusion, is for more experienced students.
Who teaches Practical Deep Learning for Coders?
Jeremy Howard teaches it. He co-founded fast.ai with Rachel Thomas, leads development of the fastai library, and is a former President and Chief Scientist of Kaggle.
So should you take it?
Take Practical Deep Learning for Coders if you can code and want to build and deploy models before you study the theory. It scores 9.1: hands-on is 10 because every lesson has notebook exercises, and truly free is 10 because it costs nothing, while clarity, up to date and depth are 8. It is the 2022 recording and offers no certificate, so check the lesson list against what you need.
Where to go next
- Deep learning, hands-on
A learning path that includes Practical Deep Learning for Coders.
- Become an ML engineer
A learning path that includes Practical Deep Learning for Coders.
- fast.ai: our platform review
What else it offers, and what costs money.
- CS50's Introduction to AI with Python
CS50's Introduction to AI with Python runs 7 weeks with 12 projects, costs nothing and gives a free CS50 Certificate if you score 70% on every project.
- Machine Learning Specialization
The Machine Learning Specialization has 3 courses and 95 hours of Andrew Ng's lessons. Coursera's FAQ says you cannot take it for free, so check the cost first.
Last checked : edition, teacher, lesson list, prerequisites, hardware, software and certificate.
Not sure this is the one?
Answer two questions and we'll point you to the course that fits where you are today.