No longer able to install updated SciPy after update to ArcPro 3.4.2

[EricSimpson96]

Hello,

My workplace has recently updated us to ArcGIS Pro 3.4.2 from 3.2.x. We have some custom Python tools that perform various functions and rely partially on the SciPy package. 

I have created a new cloned environment (trying to reload my previous cloned environment threw an error stating it was not a valid conda environment) and activated it in an attempt to update from SciPy v. 1.9.x to 1.13.0, which is the version we were able to run our tools on in our previous version of ArcGIS Pro. Attempting to install SciPy without specifying the version I need also defaults to an attempt to install 1.9.x rather than the most recent version.

Trying to force the installation of the relevant version using command prompt gave a dependency error with MKL, and it seems that the version of MKL that is tied to ArcGIS Pro in this version is not compatible with newer versions of SciPy despite the fact that it worked previously. The specific dependency error that led to this conclusion is 

"arcpy-base 3.4.* is installable and it requires mkl 2022.2.1 arcgispro_19755"

which leads me to believe that I cannot update mkl without crippling my other ArcGIS Pro functions. Does anyone have a workaround or potential other method for updating my packages?

Thank you!

[DanPatterson]

If your scripts don't require arcpy or anything from ArcToolbox, you can "create" your own environment and install whatever you want.  I have one separate from the arcgispro-py3 environment running python 3.13, numpy 2.2, scipy 1.15.2  etc etc

you can try, with sketchy success a

conda install xyzxy --no-pin

in a cloned environment, but sometimes the dependencies run too deep and can't be resolved, so a separate environment with no arc-anything may be necessary.