Hi, love your package. Will it receive an update so its in sync with SHAP? Preferably it would be nice if we could distill the computation and perhaps propose a PR for SHAP, since keeping up to date with shap might be infeasible.
For example when running Fasttreeshap I found a bug related to Lightgbm where if the data contained categorical values it wouldn't run: https://github.com/shap/shap/issues/2144
This is fixed in the latest shap but not in FastTreeSHAP.
Hi, love your package. Will it receive an update so its in sync with SHAP? Preferably it would be nice if we could distill the computation and perhaps propose a PR for SHAP, since keeping up to date with shap might be infeasible.
For example when running Fasttreeshap I found a bug related to Lightgbm where if the data contained categorical values it wouldn't run: https://github.com/shap/shap/issues/2144
This is fixed in the latest shap but not in FastTreeSHAP.