Pain points, how should I take this course?

Hi everyone,

So I finished the 1st lesson (vid, kaggle nb and book chapter 1), however I am stuck on how to actually apply what I have learned on the things I want to do.
For example, I wanted to use tabular_learner on the kaggle house price prediction competition. But I don’t understand what I’m supposed to give as input to the path parameter in TabularDataLoaders.
Why do we even need a path though if we’re already sending the location of the csv file to the function call?

More importantly, how am I supposed to overcome this? Am I supposed to read the documentation of the Path class, which I did do, but failed miserable lol (and dive deeper into the many stuff I don’t understand). For example,
path = Path(‘bird_or_not’).
dest = (path/‘forest’) #what does this mean?
Path is a class right, how can an object of a class have a /‘string’ after it?

I mean, I don’t mind trying to figure it out from the documentation, or googling around…but I guess I would like to have some confirmation whether the best way to take the course would be to go to the next lesson, or try to figure this out first?

I’m sorry if my tone sounds negative, I got a little frustrated :sweat_smile:


Focus on area that is interesting for U and code that is important for U. Think how to use ML, AI. There is no need to read all the book or take the video lesson at once. Practice coding and implementing those methods. Don’t make my mistake and don’t deep dive too long under higher level API. Look into azure or gcp that all ML algorithms, neural nets, even Vertex (graph neural nets) are easy to reach by few clicks. So, base is some understanding and knowledge to what kinds of problems these methods are applicable.

Does info in Ch 9 help?
fastbook/09_tabular.ipynb at master · fastai/fastbook · GitHub

1 Like

Welcome and Hi,

In my opinion, this is the most important chapter. In not much into data science/data mining, but the understanding of the term of the distances of embedded features (entities) will grant you in better awareness of Stable Diffusion (cosine similarity is used in those type models on embeddings vectors). This only one example. This is only my taste/opinion. :slight_smile: BTW: My approach is coming back to previous chapters and try to do suggested exercises. I also do to myself some dropout and learn other things like bash or sing. Karpathy YT channel is also interesting. Take the best for U from the bests. :slight_smile:

1 Like