fixes from initial installation

bd-j / forcepho

Generative modeling galaxy photometry for JWST

https://forcepho.readthedocs.io

19 stars 4 forks source link

fixes from initial installation #44

Closed lgarrison closed 4 years ago

lgarrison commented 4 years ago

A few fixes from my initial setup:

Fix the littlemcmc _sample_one_chain import to _sample for the latest littlemcmc
Add chain=0, random_seed=random_seed to match the import change
Make some of the scripts executable
Change the jitclass import to avoid numba warning

@bd-j I am very unsure if chain=0 is the right value to use with the new _sample function. Here is the source: https://github.com/eigenfoo/littlemcmc/blob/main/littlemcmc/sampling.py#L402

bd-j commented 4 years ago

I think chain=0 is fine with the default NDArray backend (other backends seem to be overridden anway) and seems to only be used for displaying the progressbar at the moment. The sampling wrappers in littlemcmc look like they are getting rapidly more complex and aren't as stable as I'd hoped...

lgarrison commented 4 years ago

Yeah, that's too bad. I guess these are "internal" functions we're dealing with, though. Should we go ahead and merge these changes, or do you want to think about what to do about littlemcmc first?

bd-j commented 4 years ago

Is it working for you with these changes? If so go ahead and merge. I'm putting together install instructions and I think it's possible to specify a particular commit of littlemcmc.

bd-j commented 4 years ago

for reference I've been using f4053158863b5417780a1b on odyssey

lgarrison commented 4 years ago

Sounds good. We could also just pin to version 0.2.0, which was released 3 days ago, it looks like: https://github.com/eigenfoo/littlemcmc/releases/tag/v0.2.0