Closed wbenoit26 closed 1 week ago
This is awesome and long overdue. I'll take a look at the details in a bit, but it may make sense to just add the cosmology
module to the utils
library, especially since it's a single module with limited functionality currently. Will also keep our dependencies from getting too bloated. I'll leave it up to you though.
I agree about the cosmology lib. I don't foresee us having many more functions that would fit in there. I'll move it into utils
The testing finished up and the SV looks bad: https://ldas-jobs.ligo-wa.caltech.edu/~william.benoit/pycbc_waveform_sv.html. The training curves look good, and some quick checks of the testing waveforms look good, so I'm not sure what the issue is. I'll look into it some more.
@EthanMarx I'm thinking that this PR might be a good place to start maintaining a dev
branch
@EthanMarx I'm going to open up a test PR so I can debug some of the container building issues that are happening here. I may also close this PR and create a new one that's a little more focused.
Yeah - splitting this up into smaller PRs would be great. What's the core of the issue? Can you reproduce the container build failure locally?
No, everything works locally, which is a challenge. Seems kind of sporadic, too - the data container build failed once out of nowhere on here. But yeah, I'll split this up.
This PR changes our waveform generation method from
bilby
topycbc
. Along the way, it also:cosmology
library to store whatever cosmological calculations we need. In particular, calculation of astrophysical volume now lives here, and there's now aDEFAULT_COSMOLOGY
variable for standardization purposes.rejection.py
to generate waveforms using thefrom_parameter
functionalitypycbc
organizes theirParameterList
stheta_jn
parameter distribution inpriors.py
toinclination
to match whatpycbc
expectsphi_jl
andphi_12
parameter distributions to be distributions over individual azimuthal angles of each object to allow for easier conversion to spin parameters thatpycbc
expectsTo-do:
Training and validation waveform generation ran, and testing waveforms are running now. The training run is here.
@EthanMarx there's still some work to be done on this, but I'd be interested in getting your feedback on the structure.