100 Days of ML Code: A 22K-Star Chinese Tutorial — Read This Before You Start

When I first searched for Chinese machine learning tutorials on GitHub, this project practically jumped off the screen — 22K stars, titled “100 Days of ML Code,” and entirely in Chinese. As someone who gets headaches from English documentation, I immediately hit Star and cloned it.

What This Project Actually Is

In simple terms, it is a 100-day structured learning plan for machine learning. Each day covers one concept with a Jupyter Notebook, starting from basic data preprocessing and gradually moving into deep learning. Created by the MLEveryday community, the entire project is written in Chinese, which makes it incredibly approachable for beginners in China.

The coverage is quite broad: supervised learning (linear regression, logistic regression, SVM, decision trees), unsupervised learning (K-Means, hierarchical clustering), deep learning basics (CNN, RNN), plus data visualization and feature engineering. Each topic comes with code and an infographic, which gives it a very systematic feel.

How It Actually Works in Practice

I started from Day 1 and the first two weeks went smoothly. If you have decent Python fundamentals, you basically follow the notebooks line by line and watch the outputs. The code quality is decent — it runs, and the comments are sufficient.

But around Day 30, I started hitting issues. Some libraries are outdated. For example, TensorFlow 1.x syntax is used, and if you install the latest version, things break immediately. Also, some infographics are unreadable on mobile — you need a desktop screen to zoom in.

Quick Start

If you still want to give it a try, getting started is simple:

git clone https://github.com/MLEveryday/100-Days-Of-ML-Code.git
cd 100-Days-Of-ML-Code
pip install -r requirements.txt
jupyter notebook

Then open the notebook in the Day1 folder and follow along.

Pros

• Gradual progression: One topic per day, so you do not get overwhelmed
• Chinese-friendly: All comments and explanations are in Chinese, great if English is a struggle
• Runnable code: Most notebooks work out of the box without hunting for datasets
• Broad coverage: From traditional ML to DL basics, you get a solid foundation after 100 days

Cons

• No longer maintained: Last updated in 2022, so some library versions are outdated
• Shallow deep learning: CNN and RNN sections are introductory only — you will need more resources to go deeper
• Poor infographic experience: Those info graphics are nearly unusable on phones
• No end-to-end projects: All 100 days are isolated concepts without a complete project

Final Thoughts

This project is best for absolute beginners who want a structured entry into machine learning. If English documentation intimidates you, this is a solid starting point. But if you already have some ML background or want job-ready skills, you will need much more after these 100 days.

My recommendation: treat it as a roadmap, not a destination. After completing the 100 days, you will have a good overview of machine learning. Then pick a direction — recommendation systems, computer vision, NLP — and go deeper.

If you are starting in 2024 or later, check dependency versions first. Some code may need small tweaks to run. But the core concepts themselves remain timeless.

---

About the Author

Liudingyu is a full-stack developer and heavy GitHub user. With 900+ starred repos over the past 3 years, this site only covers tools I’ve actually used or deeply researched.

📧 Found a great tool to recommend? Email [email protected]