loganamorrison / hazma Goto Github PK

I'm getting build errors on python 3.5 when the system tries to install scikit-image. I get the following:

Running scikit-image-0.16.2/setup.py -q bdist_egg --dist-dir /tmp/easy_install-mn54hdhq/scikit-image-0.16.2/egg-dist-tmp-k530fmby
Python 3.5.6 detected.

scikit-image 0.16+ supports only Python 3.6 and above.

from TravisCI. I think we are only using scikit-image for contour-finding. Either we should remove support for python < 3.6 or write our own contour-finding algorithm. The latter isn't too difficult. The Marching Squares algorithm (https://en.wikipedia.org/wiki/Marching_squares )
is fairly straight-forward. We'd just want to make achieve similar performance as scikit-image.

According to these docs, the assertAlmostEqual(a, b) function, which essentially all of our tests use, checks round(a-b, 7) == 0. This is the wrong thing to do since we want relative errors. For example,

round(1e-20 - 5e-21, 5) == 0
>>> True

We should instead switch over to something like numpy.testing.assert_allclose(), which looks at relative errors.

Currently, it looks like we just set the kinetic decoupling temperature to 10^-4. We should compute a more rigorous value for the kinetic decoupling temperature.

I think it would be better to isolate the Hazma code and tutorials into separate repositories. It seems like it would be nice to have a standalone HazmaTutorials repository which contains all Jupyter notebooks (similar to DifferentialEquations.jl and its corresponding DiffEqTutorials.jl repository.)

Also, it doesn't seem like we need the manual. I think the paper + docs should be good enough. @adam-coogan, what do you think?

We should add submodules inside the scalar and vector mediator classes to compute the relic density of the dark matter in these frameworks.

Installing hazma in a clean virtualenv fails since setup.py imports numpy.

Since March 2023, the scikit-ci project has been archived.

To move forward, we suggest to:

• remove the scikit-ci.yml file
• switch to using cibuildwheel to support building the project across platforms and CI services.
• remove files like setup.py and standardize on using pyproject.toml along with the scikit-build and cmake. See https://scikit-build.readthedocs.io

References:

• scikit-build/scikit-ci#53

rambo.generate_energy_histograms is returning energies less than the particles masses. For example, given a final state of ['muon', 'muon', 'charged_kaon'], a muon energy of 93.3445621755 is produced and kaon energies of 77.6570608521, 176.525412004, ect. are produced.