Releases: marcotcr/lime
Releases · marcotcr/lime
drop python2 support
Custom masks in LimeText, fixed bug in LimeTabular
- Added mask_string parameter to lime_text, allow user to control how tokens are masked.
- Fixed bug in truncnorm sampling where min == max
fixes, sparse support in LimeTabular
- Added sparse support for LimeTabularExplainer (thanks @imatiach-msft)
- Changed undiscretize function for LimeTabularExplainer, now using truncated normal
- Minor fixes: re.split in python3.7, submodular pick on non-binary tasks, others
small stuff
- LimeTabularExplainer accepts statistics rather than a dataset now
- various small fixes
Submodular pick, custom tokenization
0.1.1.32 version
Char level explanations, sampling around instance for continuous features in Tabular
Allow users to customize segmentation function, minor fixes
0.1.1.26 version
Random state, ascii bug
Minor fixes.
fixing imports
EntropyDiscretizer did not work due to imports.
fixing bugs
- bug in LimeImage
- bug where predict proba doesn't show