Closed zkbt closed 2 years ago
Oh, sorry, I have my atom
set up to automatically run black
on anything I save, so it made a bunch of fussy automatic formatting changes.
Looks good! How does it work if I tried A = B + C - D? Does it iteratively figure out how to combine them?
I also had a thought which is that I liked having the names in the dictionary I passed to CombinedModel so that I could do clever stuff later with plotting where I could separate out the individual components. I don't know whether it's possible with this because you don't supply a name for each model - it gets called 'left' or 'right' if I've understood correctly
Looks good! How does it work if I tried A = B + C - D? Does it iteratively figure out how to combine them?
It didn't last week, but I think that in principle it could with slightly more careful parameter management. I can take a look!
I also had a thought which is that I liked having the names in the dictionary I passed to CombinedModel so that I could do clever stuff later with plotting where I could separate out the individual components. I don't know whether it's possible with this because you don't supply a name for each model - it gets called 'left' or 'right' if I've understood correctly
That's a really good point that it would be extremely helpful to have names for the model components of a CombinedModel
. One other option (which I wound up implementing for Rainbow
s to be able to compare them with MultiRainbow
) would be to give each Model a .name
attribute (that could have some reasonable default for each class) that could be used to store these instead?
I've adapted this into the systematics_model branch with some changes to include naming. Not sure what's best to do with this pull request now?
I tried out a thing on
try-combined-model-operations
to switch from the user interacting directly with.attach_models
to just using mathematical operations (+-*/). It seems to work?!