several quality-of-life improvements #293
Conversation
|
|
||
| # The inclination prior is a sine truncated between 0, and pi/2. | ||
| self.priors['inc'] = dist.truncsine() | ||
| self.priors['inc'] = dist.truncsine() |
There was a problem hiding this comment.
I think for changing the order of things here you'll need to check that it doesn't influence the sampling.
I remember having problems with that at some point, where I thought using dictionaries should have solved this, but apparently it didn't.
Just so we aren't sampling nurot_e when we think it's something else like eps_g.
| self.Fpp(num_g, nu_p, nu_g, self.obs['dnu'][0], DPi1), lmbda=lmbda) | ||
| return a, b | ||
|
|
||
| num_p, num_g = jax.lax.fori_loop(0, self.rootiter, _body, (nu_p, nu_g)) |
There was a problem hiding this comment.
Does this improve the compile-time?
We of course also need to verify that results don't change.
There was a problem hiding this comment.
this significantly improves compile times because of no loop unrolling, and allows parameterisation by number of iterations without recompilation. this also is (as far as I can tell) the preferred pattern for for loops.
…ents of a modeID.results object
…e step of manual edits for priors)
Uh oh!
There was an error while loading. Please reload this page.