The python course

Hello DeepLearning.AI team,

I’m currently taking Andrew Ng’s courses, including Learn Python with AI for Beginners and the Machine Learning Specialization. The content is excellent, but I’m running into a major issue that affects my ability to learn independently.

Many of the projects only run inside the provided Jupyter notebook environment. When I try to run them locally, the code breaks due to hidden helper functions, custom utilities, and strict dependency versions that aren’t documented. This makes it difficult to understand how the projects work under the hood or how to build them myself outside the course environment.

Because of this, I believe DeepLearning.AI should consider creating a complete Python course that teaches:

  • How to set up real development environments

  • How to build ML projects from scratch

  • How to manage dependencies

  • How to write training loops, data pipelines, and model classes without hidden helper files

  • How to run everything locally, not only in controlled notebooks

This would help learners become truly independent and better prepared for real-world machine learning engineering.

Thank you for your time, and I hope this feedback helps improve future courses.

Best regards,
Jacques

You have a good point. Perhaps, who knows, in the near future, there could be opportunities to incorporate something closer to real-life, practical experiences that a Machine Learning Engineer might encounter in an actual work environment.

The challenge is that these tasks can be quite vague and highly dependent on the company, the specific role, the project, and the tools and environment being used. However, I do think there may be ways to create activities that simulate some of those real-world situations and give learners a better sense of what the practical work might look like.

Just to add, and you may already be aware of this, in the last module (Module 4) of AI Python for Beginners there is information on installing Python and additional Python packages on your computer.

Also, within each programming assignment you can download any of the helper functions files by going to File → Open and then selecting the relevant files (see this post for more information Which packages and libraries should we install if we want to want to use helper functions and import print_llm_response? - #3 by lukmanaj ). It is often informative to look at how the helper functions are defined. You could also download your entire workspace including your notebook and any helper function files, and run these locally in your own environment. However, in the beginning it is easier to work in the provided environment because running locally requires, among other things, managing your own API keys (Module 4 has more info on this).

Hi Jacquescoderman,

Great point.

In the meantime, to anyone wishing to understand how to do exactly what you mention I would suggest having a look at Sebastian Raschka’s excellent book Build a Large Language Model (From Scratch).

Perhaps Sebastian could be asked to create a course at deeplearning.ai…

I really appreciate the effort of the DeepLearning.AI team to introduce, in the future, activities that simulate some of those real-world situations and give learners a better sense of what the practical work might look like.

From my own learning experience, I’ve really appreciated the simplified way in which the topics are presented. At the same time, I’ve found it challenging and valuable to personalize them based on my own case studies and interests.

Besides, I was quite impressed by the “Build with Andrew” section ( Latest build-with-andrew topics - DeepLearning.AI ). It is great to see how learners are already developing their own projects based on their backgrounds, interests, and experience!

Thank the DeepLearning.AI ’s team and this forum for creating such an engaging environment and sharing their experiences and ideas!

It’s an interesting point, but it is also worth considering that you are at the beginning of the learning process. The waters are pretty deep here: what you are asking is beyond the scope of any of the courses taught here at least that I have seen. For starters, once you step into the world of running locally, suddenly everything gets pretty complicated. How many possible local environments are there? How many versions of Windows, MacOS or linux exist out there? Quite a few, right? If they pick the latest version of Windows today, then how long will it be until there’s a newer release? So how can they structure the course to be resilient to that? And as of Windows 11, it actually supports a pretty native linux environment, if you choose to use it. So what is your definition of “local environment”?

If you want to get a sense for the level of complexity involved in setting up your own environment to run the course notebooks, here’s a thread that will get you started down that long and winding road.

A lot of people use Google Colab as their development environment. That supports Jupyter Notebooks, so it’s a bit easier to port one of the course notebooks to Colab than to a local computer. (Also note that the Terms of Use for the course materials is that they are for your personal use only: you shouldn’t plan to use them as the basis for any actual products you are thinking of creating). Here’s a thread from mentor Saif that shows how to get one of the course notebooks working on Colab.

There are tons of resources out there on the web, including the Python.Org website, StackExchange and many others, that are great for learning how to build things and solving particular problems. Once you start doing your own development, then this kind of thing never ends: there’s always some new problem. So part of what you are learning is how to find answers to this type of question under your own power. You can pretty much bet that whatever you hit has already been tripped over by hundreds or thousands of other people. “Google is your friend!” :nerd_face: