Problem installing genomatnn conda environment from .yml file

[romieuj]

Hello,

I'm trying to install the conda environment which is in the yml file on the gith but it doesn't seem to work "Solving environment" doesn't resolve.

I guess some package versions don't seem to be available anymore in 2023, for example what version of stdpopsim are you using? Or do you have an updated version of the environment?

I tried with stdpopsim version 0.1 but after installing genomatnn the tool doesn't seem to recognize msprime even though it exists in the environment (the package loads with python on the command terminal).
genomatnn.txt

the txt file contains the conda environment that I tried to build to run genomatnn (with stdpopsim=0.1) but which doesn't work either.

Thanks in advance!

Ps: I don't understand, I managed to install genomatnn in 2022 just by replacing the numpy version with 1.20.0.

[grahamgower]

I've just now pinned numpy in the environment file (numpy=1.19) and then installed following the instructions in the genomatnn README.md. The tests ran successfully for me.

My guess is that you have some local python packages installed under $HOME/.local/ which are interfering. Unfortunately, conda environments are not adequately isolated, so having incompatible versions of a package installed as a local "user" package often causes problems (and the conda folks have known this for a decade; e.g. see conda/conda#448, conda/conda#7173). So, I recommend to first remove the folder $HOME/.local/lib/python3.8, and then do a reinstall of genomatnn from scratch, following the README.md instructions.

PS: genomatnn uses a very specific version of stdpopsim - the selection fork on my github. These changes were (finally) merged into stdpopsim, but additional changes were made on top of these that are not compatible with genomatnn (and the subsequent stdpopsim release was ~2 years after I stopped coding on genomatnn). But, the genomatnn installation instructions should ensure that the correct version from my fork is installed/used appropriately.