Not working: Explainability tab on ML model; partial dependence, subpop analysis, and individual exp
When trying to do either partial dependence plots, subpopulation analysis or individual prediction explanations on any of the classification(0/1) algorithms I ran it gives me the following error:
External code failed
Failed to run post training computation : <class 'ValueError'> : Number of features of the model must match the input. Model n_features is 25 and input n_features is 24
Could anyone help me fix this issue, either in DataIKU or give me a solution that I can still run these things but then manually add it somewhere in a script/notebook? Thanks in advance!
Answers
-
Hello,
Could you give us more information about your model?
- Is it an MLflow model, or did you use the lab?
- When trying to use explainability tabs, are you doing it directly from the model page, or are you doing this from a model evaluation?
-
Hi there @AurelienL
, Thanks for the reply!1. So far I am running all the different models on slightly different datasets via the Lab.
2. I am in the visual analyses tab, where I then click on one of the trained models, for example "Random Forest" and then from that page I select either of the explainability tabs. Variable importance shows me a correct histogram, but the other three analysis/explanation tabs do not work when trying to run.
Hope you can help me further, as I think this must be something that should be fixable. Everything else seems fine, getting confusion matrix and other metrics as one could expect.
Thanks in advance!
-
From what I understand, what you are doing should work, could you open a support ticket so that we can investigate this further?
