search

ratt-ru / QuartiCal

CubiCal, but with greater power.
MIT License
7 stars 4 forks source link

WIP Gain smoothing (do not merge) #311

Open landmanbester opened 7 months ago

landmanbester commented 7 months ago

This is an attempt to smooth gain solutions using the NIFTy library. It is still rather experimental because it will only parallelise with the processes scheduler and is only really meant for bandpass smoothing at this stage. There is still a bit of work before this will be ready to be used in anger

JSKenyon commented 5 months ago

The v0.2.1-dev branch has been merged. This PR will need to be updated accordingly. Just shout if you need any help - I know it is irritating.

JSKenyon commented 5 months ago

Did a little digging - I think that this branch changes the function signature of add_calibration_graph. This will need to be taken into account in the test fixtures.

bennahugo commented 5 months ago

mmm this will make quartical very hard to install on the range of machines we have - is it possible we can keep nifty as an optional dependency?

landmanbester commented 5 months ago

Yeah the signature has changed because I added the bypass_solver option. This is useful when smoothing and transferring gains because you don't have to crunch through all the vis data. I still need to modify the testing suite to account for this but at least I can pip install the branch again. @bennahugo I have just temporarily disabled 3.9 support because the dev branch of nifty I am depending on doesn't support it. I fully intent to support it before attempting to merge these changes