quality of life improvements related to importing mmda types, predictors, etc. #141
Conversation
kyleclo
changed the title
from mmda.types|predictors|parsers import * possible
| import PysbdSentenceBoundaryPredictor | ||
| __all__.append('PysbdSentenceBoundaryPredictor') | ||
|
|
||
| with necessary(["layoutparser", "torch", "torchvision", "effdet"], soft=True) as PYTORCH_AVAILABLE: |
There was a problem hiding this comment.
It would be nice to not have to define this in three places:
- setup extras_require
- required_backends on the predictor itself
- here
But I don't have a good suggestion.
There was a problem hiding this comment.
yeah :/ ive created an issue here: #144
i need to think a bit more how this would be resolved. a reasonable stop-gap could be defining a top-level config that all 3 locations imports from; reserve for next PR
There was a problem hiding this comment.
I think this all makes sense. You'll want to bump setup.py version +0.0.1 again because I think @geli-gel beat you to the punch with v0.0.36.
Putting the api types into ai2_internal also seems like a good move but note it will be breaking for the parsers/models still operationalized in spp instead of TIMO. Should be relatively easy to fix if/when we upgrade them though, I think.
from mmda.types|predictors|parsers import *possible