k-means-constrained icon indicating copy to clipboard operation
k-means-constrained copied to clipboard

Published 20 hours ago • verified publisher icon joshlk

Python package dependency issue

Open calebhallinan opened this issue 2 years ago • 4 comments

Hello! I was hoping to use your package on k-means-constrained in order to get equal cluster sizes. However, I have run into the issue where different python packages need different versions. Your package requires numpy >1.22, however numba (which is in the umap package) requires numpy <=1.21. Hence, if I would like to cluster data from umap, I am unable to do so with your package due to the numpy version requirement.

With the latest update, numba can only use numpy <=1.21 so I can't upgrade it any further.

Do you happen to know a way around this?

Thanks!

calebhallinan avatar Apr 13 '22 15:04 calebhallinan

Hi, great to hear your using the package 🙂. One way around it would be to use k-means-constrained==0.6.0 which works with numpy==1.20.3. Another way is to clone this repo and build and install the package from source with any numpy version. If the first method doesn’t work I can provide you details on how to do it

joshlk avatar Apr 13 '22 15:04 joshlk

Unfortunately, neither combination of numpy and k-means-constrained works for me.

Tried 0.6.0 and numpy 1.20.* and numpy 1.20.3 Tried 0.7.0 and numpy 1.21.* and numpy 1.21.5

Tried building with --no-binary.

Can't use 1.22 because of numba. Even more surprising to me that a couple of months ago I've managed to get 0.6.0 to work, but I don't remember which numpy version I've installed 😕

arogozhnikov avatar May 12 '22 04:05 arogozhnikov

Hi @joshlk! I'm experiencing the same issue, and any further advice on how to get around this would be greatly appreciated. Doesn't work with k-means-constrained==0.6.0 and numpy==1.20.3, k-means-constrained==0.7.0 and numpy==1.21.5 as suggested in a different thread, nor with the versions I've succesfully used earlier this year. I'm now trying to figure out how to install from the cloned repository without upgrading numpy in the process. The details on that would come in handy for me as well :)

hehomm avatar Jul 11 '22 15:07 hehomm

Hi! I'm dealing with the same issues mentioned above and would be grateful for any info on how to solve it. Thank you :)

alfredostrath avatar Jul 19 '22 15:07 alfredostrath

This is an unfortunate reality of Pythons packaging system - you can't get out of dependency hell. I'm closing this for now.

joshlk avatar Sep 09 '22 13:09 joshlk