[#860] Adding Spurious Correlation feature #1140
Conversation
cleanlab/datalab/datalab.py
Outdated
| @@ -635,3 +636,64 @@ def load(path: str, data: Optional[Dataset] = None) -> "Datalab": | |||
| load_message = f"Datalab loaded from folder: {path}" | |||
| print(load_message) | |||
| return datalab | |||
|
|
|||
| def _spurious_correlation( | |||
| self, properties_of_interest: Optional[List[str]] = None | |||
There was a problem hiding this comment.
For now, let's omit this argument in Datalab._spurious_correlation().
Remember to remove the parameter in the docstring as well.
| odd_aspect_ratio_score 0.900000 | ||
| """ | ||
| try: | ||
| issues = self.get_issues() |
There was a problem hiding this comment.
Please add a validation step that ensures that the issues dataframe has all the relevant (image-specific) scores.
If it doesn't an error with a helpful message should be raised.
There was a problem hiding this comment.
Added a validation step here to cjeck all vision/image issues are present in the correlations dataframe
There was a problem hiding this comment.
To be clear, the issues dataframe should be validated, not the correlations_df.
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #1140 +/- ##
==========================================
- Coverage 95.90% 95.82% -0.08%
==========================================
Files 81 82 +1
Lines 6050 6102 +52
Branches 996 1071 +75
==========================================
+ Hits 5802 5847 +45
- Misses 148 151 +3
- Partials 100 104 +4 ☔ View full report in Codecov by Sentry. |
Co-authored-by: Pratham Savaliya <no-reply>
…rrelation() method
…ISION_ISSUES and removed optional parameter in method and docstring
…d vision scores in transformed datasets
allincowell
force-pushed
the
spurious_correlation
branch
from
193543f to
2b01057
Compare
Summary
spurious_correlation.pymodule incleanlab/datalab/internallocation._spurious_correlationinDatalabclass that uses an instance ofSpuriousCorrelationsclass.Links to Relevant Issues or Conversations