Oege Dijk
Oege Dijk
could you add some tests that shows this works?
> I added the requested test, and removed copy from all predict function calls. I guess all these predict functions are only predicting a single row, so maybe the cost...
> I added a test to test categorical labels, this will fail as explainer dashboard does not currently support these. I think it should be possible to reuse the titanic...
> As for the training time, I just used a RandomForestClassifier which trains in less than a second on my system. > In my "pipeline" NaN values are not passed...
Will have a closer look at the code hopefully tomorrow or this weekend. But thanks already for this contribution, let's try to get it ready and released quickly!
Yes, tests are failing as well. Seems like the output shape of the shap library has changed. Will look into it...
okay, have the fix on master. Will see if I can release tomorrow...
okay it's released: 0.4.6. I actually had to change the github pypi release mechanism, so I would appreciate it if you could let me know that it worked!
`pip install -U explainerdashboard` should install version 0.4.6 or are you using conda? (it takes about a day for the conda-forge CI to pick up the latest version and release...
Should also be on conda now. What model are you using? Can you run the classifier example from the README?