fix!: use es.FormFactor in _RelativisticBreitWigner

[redeboer]

Small fix for a bug that was forgotten to be changed in #134 (the change was correctly implemented for _NonDynamic

tensorwaves/src/tensorwaves/physics/helicity_formalism/amplitude.py

Lines 456 to 461 in c35f270

	class _NonDynamic:
	def __init__(
	self,
	dynamics_props: DynamicsProperties,
	form_factor: Optional[es.FormFactor] = None,
	) -> None:

Not sure why this wasn't spotted before. Long-term, this is better addressed through #162.

To-do list

• Correctly import form factor from AmplitudeModel for a relativistic BW
• Update test values for the new model (deterministic test and callback test)
• Revisit fit example of the example workflow

[redeboer]

@spflueger It seems that the pre-factor for relativistic Breit-Wigners was converted incorrectly form the AmplitudeModel so far. This bug fix causes some of the tests to break again, but the model also looks completely different. Compare the models in the workflow notebooks prior to optimizing:

v0.1.3 (see here)

this branch (generated locally with c35f270)

Is this the expected behaviour when adding a form-factor?

[redeboer]

#187 (comment) So actually there are three separate problems:

1. the check on whether the correct formalism is used was faulty (fixed by fix: throw invalid L error correctly #188)
2. the form factor was imported incorrectly from the es.AmplitudeModel (this PR fix!: use es.FormFactor in _RelativisticBreitWigner #187).
3. there is a factor missing in the relativistic BW without form factor:
   

   tensorwaves/src/tensorwaves/physics/helicity_formalism/amplitude.py

   Lines 421 to 425 in c35f270

   	return relativistic_breit_wigner(
   	dataset[self._dynamics_props.inv_mass_name],
   	self._dynamics_props.resonance_mass,
   	self._dynamics_props.resonance_width,
   	)

   
   should have mass and width, like here
   

   tensorwaves/src/tensorwaves/physics/helicity_formalism/amplitude.py

   Lines 451 to 453 in c35f270

   	return relativistic_breit_wigner(
   	inv_mass_squared, mass0, width
   	) * atfi.complex(mass0 * gamma0 * atfi.sqrt(ff2), atfi.const(0.0))

These three problems can be addressed by separate PRs (more transparent). This PR will need to fix some of the test values and rethink the example parameter values of the workflow notebook.

[redeboer]

22d5eba fixes the tests, apart form test_callbacks.py. That test has an additional problem: it checks whether the parameters that the optimizer produces after the optimization is finished are equal to the 'latest' parameters stored to the traceback. However, (1) the callbacks by default store write the latest parameters every 10 steps (fixed by 59ee43a) and (2) the values are not written on the last step. So the "latest" parameters in the files that the callbacks (YAML and CSV) dump are actually not the latest parameters.

I can fix this tests quickly, but we may have to think of a fix here (@spflueger)?

[codecov]

Codecov Report

Merging #187 (db3330f) into master (574fe16) will decrease coverage by 1.42%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master     #187      +/-   ##
==========================================
- Coverage   83.66%   82.24%   -1.43%     
==========================================
  Files          13       13              
  Lines         704      704              
  Branches       97       97              
==========================================
- Hits          589      579      -10     
- Misses         83       90       +7     
- Partials       32       35       +3     

Flag	Coverage Δ
unittests	82.24% <0.00%> (-1.43%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...ensorwaves/physics/helicity_formalism/amplitude.py	75.37% <0.00%> (-0.38%)	⬇️
src/tensorwaves/data/generate.py	87.03% <0.00%> (-9.26%)	⬇️
src/tensorwaves/optimizer/callbacks.py	72.16% <0.00%> (-4.13%)	⬇️