wbalmer / backtracks

Python package to fit relative astrometry with background star motion tracks.
https://backtracks.readthedocs.io/
BSD 3-Clause "New" or "Revised" License
6 stars 2 forks source link

Check SOURCE_ID and source_id #56

Closed tomasstolker closed 3 months ago

tomasstolker commented 3 months ago
gotten commented 3 months ago

I get the same error as the CI on a fresh install of your PR.

gotten commented 3 months ago

ndim also includes the parameters of the host star. so it is bigger than the amount of labels. so the if statement that checks the dims and appends a label uses a value that wasnt expected.

ndim is a weird leftover from the way we ran the initial code.

if i recall

a robust fix is to revamp the ndim system. for now we need to change the number in the if statement that appends a parallax to the labels. I think we kind of assume in the end that we use "11" dimensions in almost all fits. I'm running a test now with ndim==4 replaced by ndim ==10 and ndim==5 replaced by ndim==11.

tomasstolker commented 3 months ago

Right, I agree, it requires a more complete implementation! Let me try to fix this case and I will leave a more robust implementation for someone else, since I don't know the code sufficiently well for that.

gotten commented 3 months ago

Still crashes?. So label is defined by us as something that is 4 or 5 long. And cornerplot looks at the size of samples (extracted from dynesty results) for the size of the for loop that also goes over the labels. So there can be a mismatch there.

tomasstolker commented 3 months ago

I couldn't find an easy fix, so reverted those changes!