-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fixing the grouping and the thcovmat flag #1588
Conversation
@scarlehoff as you can see I have only changed |
We only want to order by fitting group when we are fitting. I'm not sure this is a good solution actually, I think the "metadata group" should be equal to the "fitting group" during a fit. Is there a reason why they should be different? |
The point is that by default |
Then I think the right thing is for metadata group to default (in the case of a fit) to the fitting group. |
nnpdf/validphys2/src/validphys/config.py Line 1581 in 05602d6
This is the point in which the default data_grouping is stated. Do we want to change this? |
Yes. Indeed, I can even see that @wilsonmr left a comment many moons ago predicting this day would come :P |
FWIW while the while "data keyword" grouping system was great in both allowing complex new uses of the code and maintaining compatibility, I think the resulting trade off in complexity is not favorable to us anymore, in particular since we don't have to support an |
Do you like the new solution more? |
Greetings from your nice fit 🤖 !
Check the report carefully, and please buy me a ☕ , or better, a GPU 😉! |
The grouping used in the thcovmat fit is broken because
groups_dataset_inputs_loaded_cd_with_cuts
does not usegroup_dataset_inputs_by_fitting_group
and therefore is not grouped as the thcovmat (even if the groupings are similar so sometimes it happens to be correct by chance).Moreover, the thcovmat group is used also when both
use_thcovmat_in_sampling
anduse_thcovmat_in_fitting
are false.