Fix FK computation whenmur != muf

[Radonirinaunimi]

This error was introduced after an EKO update. Before that change, the scale $\mu_R^2$ was correctly used to compute $\alpha_s$:

pineko/src/pineko/evolve.py

Lines 400 to 408 in e88a4a2

	ren_grid2 = xir * xir * mur2_grid
	alphas_values = [
	4.0
	* np.pi
	* sc.a_s(
	mur2,
	)
	for mur2 in ren_grid2
	]

And this is the reason why people managed to compute FK tables for H1 Jets before (see eg. vp-report).

This PR adds a quick fix to this issue (which is a bit ad-hoc). Ideally, I'd say, mur2grid(mur2, nf) should be part of the operator cards but this would require an update of EKO which is not ideal. At least with this fix, the EKO can be correctly convolve with the Grid.

TODO:

• Fix the CLI and Benchmark to account for the theory ID in order to fetch the corresponding theory card

[felixhekhorn]

The mention commit fixes another bug: for (n)FONLL to correctly play out one has to use the correct $n_f$ for $\alpha_s$ (it is an explicit dependency - see Eq.(27) in 1001.2312) - see also #83 . I understand this was not done carefully enough

[scarlehoff]

Before merging this, please check that it can indeed by used for herajet and that it works for a few other datasets (beyond the ones in the tests) for theory 4.1.

Possibly with N3LO DIS grids.
Does this fix by chance also the NMC issues?

[scarlehoff]

List of possible bugs fixed by this:

• HERAjet @kamillaurent
• HERA @ NNLO @andrpie (but actually it was just version differences)
• NMC @ NNLO (this was fixed by not optimizing)

[Radonirinaunimi]

I did check HERAjet a few days back and it works. The other bug (related to NaN) will be fixed by NNPDF/pineappl#367 which will be propagated into pineko once released.

[Radonirinaunimi]

This now completely solves all the issues we've been seeing. Before merging, I'd like to double check again the NMC, could someone please provide the grids and EKOs for one of the NMC datasets?

[scarlehoff]

Should this be merged? Was this ever tested?

[Radonirinaunimi]

Should this be merged? Was this ever tested?

Well, in principle yes (checked for different datasets). I did ask for the NMC grids and EKOs back then but haven't received them ¯\(°°)_/¯

[scarlehoff]

I'll merge it then, the problem left with NMC seem to be at the grid level NNPDF/yadism#344, if after we fix that we still have issue we can reopen.

[scarlehoff]

@felixhekhorn there's still a changes requested by you, are you happy with the changes that were done?