search

ML4GW / aframev2

Detecting binary black hole mergers in LIGO with neural networks
MIT License
5 stars 16 forks source link

Pickleable pastro #249

Closed EthanMarx closed 1 month ago

EthanMarx commented 1 month ago

@wbenoit26 Probably a little annoying to see this reversion to the class based implementation but I realized it was necessary for a couple reasons:

  1. Our functional approach was slick, but we couldn't pickle the fit models due to the use of local function definitions
  2. In the online deployment, we won't want to refit every time we restart, so being able to pickle the fit model is necessary
  3. This class based approach makes it a bit easier to implement different fitting procedures
wbenoit26 commented 1 month ago

This looks great, and makes a ton of sense. I think we should remove the ability to downsample the background because doing so generally produces inaccurate results.

EthanMarx commented 1 month ago

Okay great glad you agree - yeah I can remove that behavior until we figure it out - was passing None to that variable anyway