Feature Request: Importance scores for Models
madprogramer opened this issue · comments
Thanks for your patience on this @madprogramer! Our current approach for this is to make it as approachable as possible to extend vetiver with customized endpoints.
I have a demo here that shows how to:
- store fairness metrics as model metadata
- serve an explainability score at a new customized endpoint (using DALEXtra)
We also have some materials on adding a new customized endpoint from our workshop at posit::conf last year. Check out:
See also #68 with plans for more advanced documentation, including extending vetiver with new endpoints.