search

andreufont / LyaCosmoParams

Notes on cosmological parameters and Lyman alpha simulations
2 stars 2 forks source link

modified post-processing code to be able to choose axis #107

Open andreufont opened 4 years ago

andreufont commented 4 years ago

Adds option to specify the axis along which we want to extract the skewers and measure p1d.

It also includes book-keeping changes in extract_skewers.py and snapshot_admin.py related to issue #98.

When using axis=None, the code reproduces the current p1d (with the book-keeping changes mentioned above).

Before merging this, we should wait for GriddedSpectra in fake_spectra to be updated: https://github.com/sbird/fake_spectra/issues/42

Chris-Pedersen commented 3 years ago

Hi @andreufont I'm happy to pick this up, have you tested it works with Simeon's master on fake_spectra? We could switch to using that instead of my modified fake_spectra branch. As far as I understand it temperature rescalings are possible on master, but I haven't experimented with using the RateNetwork class to do this, and we would also have to modify the existing scripts to work with the RateNetwork syntax.

The other thing that would change if we switch branches is the fit range. Currently when I compare the TDR fits between my and Simeon's branches, there's a ~4% difference, likely due to differences in the fit ranges and method.

So I guess we either:

Switch to master and:

  1. Test the RateNetwork temperature rescalings, and rework our scripts to use them
  2. Decide on whether to recalculate the T0 fits using Simeon's master, or modify the fit range to be consistent with our previous values

Stick with my modified branch and update it to include the multiple-axis commits that you added to fake_spectra master.

Either way I don't think we need to store the p1d along each axis as part of the emulator data, only the combined p1d.

Lets discuss this tomorrow!

andreufont commented 3 years ago

Yes, master branch in fake_spectra works for set_axis, I've used it extensively in Sherwood sims.

Good point about rescalings and fitting ranges, let's chat later today.