Making n3fit use python covmats

[siranipour]

Btw, do we need to add use_t0: true somewhere in the fit runcard?

[Zaharid]

That is in n3fit_exec.py. That said, maybe we want to iterate the t0s in the runcards? @enocera

[siranipour]

Maybe we should remove the explicit node and make the provider take in the t0 covmat always?

[wilsonmr]

Maybe we should remove the explicit node and make the provider take in the t0 covmat always?

Yeah do this

[wilsonmr]

That is in n3fit_exec.py. That said, maybe we want to iterate the t0s in the runcards? @enocera

I think the current way of setting t0 pdf in runcard is clearer than having it being iterated under the hood, I'm also not even sure what that would mean in fact.

[wilsonmr]

these PRs are looking very simple (which is a good thing!)

[Zaharid]

This is failing with

[CRITICAL]: Bug in n3fit ocurred. Please report it.

Traceback (most recent call last):

  File "$PREFIX/lib/python3.7/site-packages/n3fit/scripts/n3fit_exec.py", line 227, in run

    super().run()

  File "$PREFIX/lib/python3.7/site-packages/validphys/app.py", line 151, in run

    super().run()

  File "$PREFIX/lib/python3.7/site-packages/reportengine/app.py", line 374, in run

    rb.execute_sequential()

  File "$PREFIX/lib/python3.7/site-packages/reportengine/resourcebuilder.py", line 168, in execute_sequential

    perform_final=self.perform_final)

  File "$PREFIX/lib/python3.7/site-packages/reportengine/resourcebuilder.py", line 175, in get_result

    fres =  function(**kwdict)

  File "$PREFIX/lib/python3.7/site-packages/validphys/covmats.py", line 239, in dataset_t0_predictions

    return central_predictions(dataset, t0set).to_numpy().squeeze()

  File "$PREFIX/lib/python3.7/site-packages/validphys/convolution.py", line 189, in central_predictions

    return _predictions(dataset, pdf, central_fk_predictions)

  File "$PREFIX/lib/python3.7/site-packages/validphys/convolution.py", line 124, in _predictions

    fkfunc(load_fktable(fk).with_cuts(cuts), pdf) for fk in dataset.fkspecs

  File "$PREFIX/lib/python3.7/site-packages/validphys/convolution.py", line 124, in <listcomp>

    fkfunc(load_fktable(fk).with_cuts(cuts), pdf) for fk in dataset.fkspecs

  File "$PREFIX/lib/python3.7/site-packages/validphys/convolution.py", line 262, in central_fk_predictions

    return central_dis_predictions(loaded_fk, pdf)

  File "$PREFIX/lib/python3.7/site-packages/validphys/convolution.py", line 396, in central_dis_predictions

    return _gv_dis_predictions(loaded_fk, gv)

  File "$PREFIX/lib/python3.7/site-packages/validphys/convolution.py", line 337, in _gv_dis_predictions

    gv = gvfunc(qmat=[Q], vmat=FK_FLAVOURS[fm], xmat=xgrid).squeeze(-1)

  File "$PREFIX/lib/python3.7/site-packages/validphys/pdfbases.py", line 292, in central_grid_values

    return self.apply_grid_values(func, vmat, xmat, qmat)

  File "$PREFIX/lib/python3.7/site-packages/validphys/pdfbases.py", line 406, in apply_grid_values

    gv = func(flmat, xmat, qmat)

  File "$PREFIX/lib/python3.7/site-packages/validphys/gridvalues.py", line 121, in central_grid_values

    return _grid_values(pdf.load_t0(), flmat, xmat, qmat)

AttributeError: 'NoneType' object has no attribute 'load_t0'

[siranipour]

@Zaharid the reason for that crash is because use_t0 is set to False in the regression fit runcard:

nnpdf/n3fit/src/n3fit/tests/regressions/quickcard.yml

Line 22 in fcb8f08

use_t0 : False

We decided to add an explicit_node to dispatch on use_t0

[github-actions]

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

• Fit Name: NNBOT-fbfc6b7b0-2021-06-28
• Fit Report: https://vp.nnpdf.science/y_xnzq8pRTu2ON-mJGWY7Q==
• Fit Data: https://data.nnpdf.science/fits/NNBOT-fbfc6b7b0-2021-06-28.tar.gz

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

[siranipour]

Are we happy to merge this

[scarlehoff]

This seems ok to me (and both bots pass)