How do people go about making their own notebooks for the courses?

This is potentially a very stupid question, but it has been on my mind for a little while. I am quite new to the world of deep learning but am giving it my best. In accordance with Jeremy’s recommendation, I am making my own notebook and attempts at all lessons. But I am uncertain what is the best way of doing it.
I am using the Paerspace interface

Currently, I just make my own notebook in the course folder, and that is working fine.

  • I am curious how others are doing it?
  • How are you dealing with putting it on a GitHub repo?
  • What do you do if you want to experiment with your entirely own models?

Looking forward to hearing from you


Have you used git? The first step would get familiar with git commands. Then you would be get the solutions available.

I’m not saying this is how you should do things but it’s what I’m planning for v3 of the course. Personally, I found it useless simply going through the notebooks and hitting shift enter and Jeremy vehemently discouraged being lazy like that. So here’s my plan:

For each lesson:

  • First download the notebooks and data and shift enter all the cells to make sure everything is working. If not, search the forums and try to debug it on my own.
  • Next, add a cell above each cell or a group of related cells stating what is being accomplished in the following cells in my own words.
  • Then delete all of the cells containing code.
  • Then try to recreate the notebook by writing my own code and following the prompts i wrote a few steps ago.

Note I’m skipping the part about watching videos. I’ll only watch them once I have a basic understanding of the lesson and have tried to code it as well. (Besides, v3 videos aren’t out yet as of this writing) Or alternatively, i might create audio files out of them and listen during my commute.

Feel free to critique my learning plan for 2019.


Jupyter and git are separate things and there is no integration. Who would need integration? I just do it manually
If you wish to read more, look here:

If you want to seriously learn about version control, I suggest a passionate and informative book

I like this a lot. One thing I’m a bit torn on with the fill-in-the-blank approach is that, from a blank page, it’s easy to forget steps. For example, if one of the steps I write down is “create columns for each part of the date”, I think I’ll be able to recall pretty quickly what I need to do, but from a blank page, I could see myself forgetting to do that. I guess there are a few options:

  • Ignore this issue and remember that you’ll have your notes in front of you when you’re implementing this for a project
  • Test yourself separately on remembering the steps,m
  • As a follow up step, try implementing from a blank page

Would love to get your thoughts!

Great question.
What I did with cats and dogs:
I found a similar Kaggle competition, Histopatological Cancer Detection
I copied some lines, which I could understand
I changed the dataset to kaggle one
I run simplest net
Then I copied more lines, trying to understand, I added resnet50 (I liked the name)
I repeated last step.
After two days:
I am 121 (ai-kido moonlighters - I try to involve friends, skittish boys) from 449 teams
This is crazy good course!
I never had any contact with ML or AI before. I have a gaming notebook, a leftover from earlier project with i7, 32GB RAM, GTX1070
I am hooked, I learn and test new things every day…