Greetings from your nice fit 🤖 !
I have good news for you, I just finished my tasks:

• Fit Name: NNBOT-ef07731e1-2022-02-10
• Fit Report: https://vp.nnpdf.science/ASUKIbIHSPObSN3r9QEvXg==
• Fit Data: https://data.nnpdf.science/fits/NNBOT-ef07731e1-2022-02-10.tar.gz

Check the report carefully, and please buy me a ☕ , or better, a GPU 😉!

In the last commit the ThPredictions are no longer using the NNPDFResult class and instead use the StatsResult. Everyting seems to work fine (I'll send the fit bot again now). Since now the central value is computed by taking the average over the result of the members (instead of taking the result of the central replica as the central value) the regression tests have slightly changed. I've uploaded in this comment the diff of the differences for future reference. They are mostly order 1e-3 with some getting to 1e-2.

I will finish what's missing from this PR (mainly making sure that the grid_values are never used as arrays but always as instances of the Stats class) and then will mark this PR for review. This will also make the xplotting grid to be consistent with other grids like Lumi1dGrid for which grid_values is already a Stats instance.

The change of the MC PDF to include by default all replicas (and then being the job of the StatsClass to decide what to do with them) I will do in a separate PR pointing to this one because although is in principle a smaller change (if I did everything correctly it should only be a small change on the behaviour of the Stats and PDF classes) but that could have many unforeseeable consequences.

differences_from_da3069c_to_1fe212432869d.tar.gz

Good to be rid of the nnpdfresult abomination!

Greetings from your nice fit 🤖 !
I have good news for you, I just finished my tasks:

• Fit Name: NNBOT-2d5809e12-2022-02-10
• Fit Report: https://vp.nnpdf.science/5rdA1fW-TFeW9BWsWtwKBQ==
• Fit Data: https://data.nnpdf.science/fits/NNBOT-2d5809e12-2022-02-10.tar.gz

Check the report carefully, and please buy me a ☕ , or better, a GPU 😉!

Greetings from your nice fit 🤖 !
I have good news for you, I just finished my tasks:

• Fit Name: NNBOT-9a3e29c9d-2022-02-11
• Fit Report: https://vp.nnpdf.science/3h4wnx7yQ0aqfCuSbYEvxQ==
• Fit Data: https://data.nnpdf.science/fits/NNBOT-9a3e29c9d-2022-02-11.tar.gz

Check the report carefully, and please buy me a ☕ , or better, a GPU 😉!

This is ready for review.

At this point the raw rawdata or data is used wherever the _rawdata was used, so the results are expected to be equivalent to whatever we got before this PR. Both the tests and the fitbot seems to confirm that.

In order to continue (and deal with #1516) I will open today another PR pointing to this one where instead of the raw data, the error_members property/method is used wherever it makes sense. The reason for making it another PR is because, as said in #1516, the code is very inconsistent in many places so it can potentially break havoc and I prefer not to increase the entropy of this (large enough) PR.

Apart from the things discussed above I think we should merge this to make some progress.

Apart from the things discussed above I think we should merge this to make some progress.

If you want small things like those (that do not break anything) you can add as a TODO to the third PR, #1522 and I'll do them there.