move model instantiation inside loop

[jgallowa07]

This addresses #92

[jgallowa07]

Seems that the tests are failing due to

Run nbqa ruff .
/usr/share/miniconda/envs/dms-vep-pipeline-3/bin/python3.11: No module named ruff
Error: Process completed with exit code 1.

Not exactly sure why this is happening. Looking into it now

Update
I see you're working with the same issue now and decided to remove the notebook linting. I'll wait till you merge #94 - then rebase this branch.

[jbloom]

I am going to look at your pulls soon (was behind)---but yes, the failures were because ruff is now supposed to run directly on notebooks rather than via nbqa as was done prior to the pull request I just made.

[jgallowa07]

Great, seems to be working now. Thanks to @Caleb-Carr for getting me to do this - and apologies for not making this patch a priority over the rest of #86. I realize this is non-trivial so it's worth mentioning to others who are looking at these results. The main problem this bug introduces is that the chosen lasso_shift average model would essentially represent the training of all lasso_shifts preceding it.

I am going to look at your pulls soon (was behind) --

No rush on my end, I'm in a similar boat - #91 is still a WIP so nothing to look at currently - later this week hopefully.

[jbloom]

Makes sense. I'm not sure if the prior behavior was a bug strictly speaking (that is also a way to do it, right?) but I'm fine with this change.