Accept prec_init as array or list #1616
Conversation
Codecov Report
@@ Coverage Diff @@
## master #1616 +/- ##
==========================================
+ Coverage 97.21% 97.31% +0.10%
==========================================
Files 68 68
Lines 2151 2157 +6
Branches 345 348 +3
==========================================
+ Hits 2091 2099 +8
+ Misses 31 30 -1
+ Partials 29 28 -1
|
There was a problem hiding this comment.
Hi @antonis19 , thanks for the PR and apologies for the delay on reviewing. Added some comments. I'd also recommend updating the type of TrainConfig.prec_init to a union type that allows all of the types we're now handling.
|
@henryre I took care of the comments you made. Now there seems to be a build error even though all the tests pass. It seems to me that these are errors related to the build configuration, or am I wrong? |
|
The build issues should be resolved on trunk. Re-running for this PR: https://github.com/snorkel-team/snorkel/pull/1616/checks?check_run_id=1764471482 |
|
@henryre seems that all the tests are passing except for codecov/patch. What would we need to do to fix this? |
There was a problem hiding this comment.
@antonis19 the codecov issue is not clear to me, so happy to bypass it once the last comment is addressed!
There was a problem hiding this comment.
🎉🎉🎉 Thanks so much for the contribution @antonis19 and your patience on the back-and-forth! Looking forward to getting this released!
## Description of proposed changes Currently, `LabelModel` only accepts a `float` or an `int` for `prec_init`. These get converted into a 1D tensor of the same value. In this pull request, `prec_init` can be passed as `np.ndarray` or as a Python list also. This way different labeling functions can get assigned different priors. ## Related issue(s) Fixes snorkel-team#1484 ## Test plan ## Checklist Need help on these? Just ask! * [ ✅ ] I have read the **CONTRIBUTING** document. * [ ✅ ] I have updated the documentation accordingly. * [ ✅ ] I have added tests to cover my changes. * [ ✅ ] I have run `tox -e complex` and/or `tox -e spark` if appropriate. * [ ✅ ] All new and existing tests passed.
Description of proposed changes
Currently,
LabelModelonly accepts afloator anintforprec_init. These get converted into a 1D tensor of the same value. In this pull request,prec_initcan be passed asnp.ndarrayor as a Python list also. This way different labeling functions can get assigned different priors.Related issue(s)
Fixes #1484
Test plan
Checklist
Need help on these? Just ask!
tox -e complexand/ortox -e sparkif appropriate.