Open Jorge-Alda opened 1 year ago
A bit of dependency hell:
smelli
's obstable_sm
only works correctly with Python 3.11. Otherwise, it complains that the arrays are not C-contiguous, probably related to numpy
.shap
's dependencies, llvmlite
(why does it even need a compilator) only can be installed via pip
with Python 3.9.Luckily, we don't use them at the same time. So I will do the first part with Python 3.11, and when all the training points are computed, will switch to 3.9 for the machine learning part.
The culprit was in fact scipy
's version 1.10. Going back to 1.9.3 solves the issue, and we can do everything in Python 3.9.
Update: The fit now includes new values for $R_{D^*}$, angular observables in $B^0\to D^*\ell\nu and more updated observables.
Recalc with the newest $R{K^{(*)}}$ and $R{D^{(*)}}$ experimental measurements and a couple of improvements:
smelli
: https://github.com/Jorge-Alda/smelli/tree/cacheRGE