Update docstrings in validphys with new datasets names

[kamillaurent]

I am still not sure when to add "variant": "legacy" and "frac": 0.75 when switching names.

[scarlehoff]

In general, if you are not doing a fit, then the fraction is not needed (since it is the training/validation split)

RE legacy, use the variant that is used in the 4.1 runcard. If, for whataver reason, it cannot be used (e.g. it is an example that only runs for older theories and data for some reason) then use legacy.

[kamillaurent]

Regarding datasets that are not in the dataset_names.yml (i.e. NTVNUDMNFe from validphys2/src/validphys/n3fit_data.py), should I remove them or should I change them with the closest name (in this case NTVNUDMNFe_dw_ite)?

[scarlehoff]

should I remove them or should I change them with the closest name (in this case NTVNUDMNFe_dw_ite)?

I'd say the closest one.

But again, if it is an exmaple, you can choose a different dataset and make sure that the example is showing the same.

[scarlehoff]

Hi @kamillaurent is this complete?

[kamillaurent]

I don't spot any old name in the docstrings, so I believe this is complete @scarlehoff

[kamillaurent]

I don't spot any old name in the docstrings, so I believe this is complete @scarlehoff

[scarlehoff]

here I think NMC was acting more as a noun referring to the NMC collaboration itself. But it's ok, it reads good either way

[scarlehoff]

Here, instead of changing this manually you should run the code above and update the info (you will see that you cannot use pseudodata_test_fit_n3fit and have to use an updated version such as pseudodata_test_fit_n3fit_260518), which has 15 replicas instead of 10 and which has ATLAS_Z0_7TEV_46FB_CC-Y as first dataset.

[scarlehoff]

Same here. Instead of updating the results of the code, you should update the fit instead.

A recent one you can use is: nnpdf40-like_t0_sampling, and below instead of experiments which doesn't exist anymore, you need to use dataset_inputs = fit.as_input()['dataset_inputs']