How accurate are the videos to 2020

While I am practicing to code simultaneously watching the videos I am having little trouble with missing functions or libraries. Could @jeremy please tell me, if there is some kind of a work around or a documentation that we can look up. Please and thank you.

1 Like

Hi Surya, you might have already found this but if not, this has been helpful for me in the last day or so as I attempt something similar to you

I started this course in 2020 and just finished it, here is what I did. I downloaded the repository from github, which also contain the fastai 0.7 library (sources and documentations) and I put it in the same directory as the course notebooks or your practice notebooks, then you should be able to import everything without changing the code (I only tweaked some code in the library which gave some errors twice as I remember). the vast majority of course can be completed on your personal computer, and if you have a Nvidia GPU you can complete the whole course on our computer.

1 Like

Thank you so much @mmohamed93 I have recently started, so I will keep in mind.

Thank you so much @brettmoreton Yeah this helped me.

So I am doing the ML course and in lesson 4(video lesson 8) I couldn’t get
fastai.torch import’ so I followed all the instructions and I still get ModuleNotFoundError: No module named ‘fastai.torch_imports’

I would really appreciate if you could point me to an instruction manual or a step by step process which would help me complete the ML course, and I plan on doing the DL course after.

I am running windows 10, with a GPU (Nvidia 1050)

Nvm I uninstalled and reinstalled everything. it works fine now. used git clone for fastai this time

I have a question regarding what you wrote about completing the course on personal computer. Do we ever use gpu for this course? and if so for which models? I am still working with Random Forest and when I use my computer, even when I write “conda activate fastai” I don’t see any gpu being used.

I am trying to use google colab to finish this course. However, I am not able to install or import fastai.tabular into colab
I’m getting this error
ERROR: Could not find a version that satisfies the requirement fastai.tabular (from versions: none)
ERROR: No matching distribution found for fastai.tabular
any help?

Could you please give the link of the repositiory that you downloaded ?
Also does it support the intro to ml course ?

@hamzautd7 just make sure the that the fastai 0.7 library you have cloned from github or downloaded is in the same directory as the notenook you are running, make sure you have Pytorch and Cuda installed.

1 Like you can download the repositiry and it includes the fastai 0.7 library (named fastai) it should cover everything you need for the intro to machine learning course.

for the Random Forest no GPU is used only in neural network you will start using the GPU (Pytorch and CUDA)

Thanks !

Is there anything like this, but using native fastai v1 functions?

I am especially struggling, on how to create the categorized (without re-implenting proc_df, train_cats…) dataframe. If I understand correctly, in v1 the TabularDataBunch somehow substitutes the dataframe with the categorical columns. understand, instead of editing the existing dataframe, v1 uses the class TabularDataBunch should be used. According to the documentation:

procs = [FillMissing, Categorify, Normalize]
data = TabularDataBunch.from_df(path, df, dep_var, valid_idx=valid_idx, procs=procs, cat_names=cat_names)

But how do I now get the df and y for the random forest?

m = RandomForestRegressor(n_jobs=-1), y)
m.score(df, y)

I think that the solution is already in these lines, but I am not quite getting it:

(cat_x,cont_x),y = next(iter(data.train_dl))
for o in (cat_x, cont_x, y): print(to_np(o[:5]))

EDIT: Whenever using next(iter(data.train_dl)), my tensors are only of size 64. What’s the reason for that? I’d expect the size to be the size of my initial dataframe.

Hi Mohammed:

I opened up but i dont see the lessons that ere taught as part of Intro to ML course in 2018. I see that they used lessons1.ipynb et for Random Forest implementaton for Bluebook Buldozer’s usecase. Can you please point me to the corret page if you have info on that.

I think its been moved here:

1 Like

Thanks, Morgan.

1 Like