Closed boyleconnor closed 1 year ago
I should mention that I'm getting a bunch of MyPy issues that I don't think are related to my code:
Not sure what to do about these
@boyleconnor Do you get those MyPy errors when running make mypy
? I'm not able to reproduce with that command locally.
Aha, it happens when I upgrade to 0.982
. 0.942
(my original version) and 0.950
(locked) are too old to get the error. It looks like those log_experiment
warnings are spurious, because the arguments it's complaining about should have optional type and are being assigned from an optional type, but I'm not sure why they're happening.
@boyleconnor Just #type: ignore
them. mypy is improved beyond some of ADAM's original code base and sometimes our usage of inheritance still causes mypy issues. Given the time sink it would probably be to correctly update everything across the library its not worth it.
I defined and added three new top-level nodes to the ADAM perception graph,
size
,relative_size
, andrelative_distance
. I also changed one of the feature extraction steps of ADAM to output adict
/object
for therelative_size
feature instead of thelist
/array
(which was harder to read as I originally coded it; you had to infer the relevantobject_name
instead of it being stored explicitly).This PR is definitely not done yet, as it doesn't handle the
touching
feature, but I want to get eyes on this ASAP.I also wanted to hear from @lichtefeld and/or @spigo900 whether there's anything downstream of this that I need to change to make sure that these new nodes in the perception graph are getting ingested properly.