Issue with Numpy and installing fastai into jupyter notebook

Part 1 (2019) Advanced (Part 1 v3)
1

Windows 11 with anaconda, running jupyter notebook through microsoft edge. Attempting to install fastai and use it to test something for work purposes but I am unable to get past this point.

Currently, no matter what version of numpy i have installed it says i have 2.02 installed inside of my notebook environment. I actually went into my environment folder and it stated I didn’t have any numpy installed until I used bash to install it through powershell.

Now, when i go to install fastai it tells me the following error, which means the dependencies of fast ai itself are literally not able to run.

“ERROR: pip’s dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
numba 0.59.1 requires numpy<1.27,>=1.22, but you have numpy 2.0.2 which is incompatible.
pywavelets 1.5.0 requires numpy<2.0,>=1.22.4, but you have numpy 2.0.2 which is incompatible.
streamlit 1.32.0 requires numpy<2,>=1.19.3, but you have numpy 2.0.2 which is incompatible.”

Please help, i was pretty excited to be playing around with this cool tool , especially the tabular, and I can’t even get it to work at all.

2

I have attempted to reset the kernel, install Numpy 1.26, install numpy 2.0, and both seem to not resolve the issue i’m getting with the following code:

from fastai.tabular.all import *

dls = TabularDataLoaders.from_csv('/Desktop/TrendforAI.csv', y_names="Replace",
    cat_names = ['Result'],
    cont_names = ['Time Stamp'],
    procs = [Categorify, FillMissing, Normalize])

learn = tabular_learner(dls, metrics=accuracy)

Please advise, This has been a pretty frustrating experience so far given that this doesn’t seem to just “work” out of box.

3

Is tabular learn broke? I’ve tried installing all dependencies but I keep on getting a Import Error.

 A module that was compiled using NumPy 1.x cannot be run in
NumPy 2.1.2 as it may crash. To support both 1.x and 2.x
versions of NumPy, modules must be compiled with NumPy 2.0.
Some module may need to rebuild instead e.g. with 'pybind11>=2.12'.

If you are a user of the module, the easiest solution will be to
downgrade to 'numpy<2' or try to upgrade the affected module.
We expect that some modules will need time to support NumPy 2.

Followed by a list of import dependency errors , all linked to pandas, I upgraded pandas and that didn’t seem to improve anything. I am unsure what is the cause of this grief, but when one imports and follows instructions one shouldn’t be left scratching your head. Maybe its my lack of experience with python, but this seems like something I cannot get myself out of because a different issue happens on kaggle notebook.

4

Hi.
Unless you specially want to run local you can try Google Colab with Google Drive. Both are free and fastai is available. There are pay option as well.
Regards Conwyn

5

Hey Conwyn,

Thanks, I would much rather run it locally, is there any way to resolve the issue, or , any particular way to get this set up so I can try out the tabular learn?