Shap: predict() got an unexpected keyword argument 'validate_features'

[gaush123]

Using
Xgboost version: 0.7.post3

SHAP version: 0.28.5

TypeError Traceback (most recent call last)
in
15 explainer = shap.TreeExplainer(model)
16
---> 17 shap_values = explainer.shap_values(X)
18
19 # visualize the first prediction's explanation (use matplotlib=True to avoid Javascript)

~/anaconda3/envs/propensity/lib/python3.7/site-packages/shap/explainers/tree.py in shap_values(self, X, y, tree_limit, approximate)
177 phi = self.model.original_model.predict(
178 X, ntree_limit=tree_limit, pred_contribs=True,
--> 179 approx_contribs=approximate, validate_features=False
180 )
181

TypeError: predict() got an unexpected keyword argument 'validate_features'

[sararob]

I'm also seeing this error, running shap in Colab with xgboost version 0.7 and shap 0.28.5. Here's my model

model = xgb.train({'objective': 'binary:logistic'}, dmatrix)

I'm able to run the code to create the explainer without errors:

explainer = shap.TreeExplainer(model)

But when I run: shap_values = explainer.shap_values(np.array(x_test[:10])) I get the same error that @gaush123 ran into. Let me know if I can provide more details.

[SaadAhmed96]

I guess you need to create a DMatrix for the test set as well or you can also try X_test.as_matrix() to get the shap values. Let me know if this works.

[gaush123]

Thanks @SaadAhmed96 it worked.
changing
shap_values = explainer.shap_values(X)
to
shap_values = explainer.shap_values(X.as_matrix())
worked

[ahujak]

I'm also running into the same error. I am using shap in Colab with xgboost version 0.7 and shap 0.28.5.
I used the suggestion of changing X to X.as_matrix() but that does not work either. Is there anything else that I can do?

[AshwinB-hat]

shap: 0.28.5
xgboost: 0.7.post4
Same issue. The example given in the read of shap for xgboost gives same error. X.as_matrix() does not help.

[SaadAhmed96]

Can you post the full example here so that I can look at it? Maybe something else is causing issues.

[AshwinB-hat]

print(shap.__version__)
print(xgb.__version__)
shap.initjs()

X,y = shap.datasets.boston()
model = xgb.train({"learning_rate": 0.01}, xgb.DMatrix(X, label=y), 100)

explainer = shap.TreeExplainer(model)
shap_values = explainer.shap_values(X.as_matrix())

shap.force_plot(explainer.expected_value, shap_values[0,:], X.iloc[0,:])

This is the same from the example README. using a Dmatrix or normal X without as_matrix() gives the same error.

## Output

0.28.5
0.7.post4

TypeError Traceback (most recent call last)

in ()
10 # (same syntax works for LightGBM, CatBoost, and scikit-learn models)
11 explainer = shap.TreeExplainer(model)
---> 12 shap_values = explainer.shap_values(X.as_matrix())
13
14 # visualize the first prediction's explanation (use matplotlib=True to avoid Javascript)

/usr/local/lib/python3.6/dist-packages/shap/explainers/tree.py in shap_values(self, X, y, tree_limit, approximate)
177 phi = self.model.original_model.predict(
178 X, ntree_limit=tree_limit, pred_contribs=True,
--> 179 approx_contribs=approximate, validate_features=False
180 )
181

TypeError: predict() got an unexpected keyword argument 'validate_features'

[SaadAhmed96]

Thanks for posting this example. It works fine on my system. May be its because of the xgboost version you are having issues. Try updating it.

[AshwinB-hat]

Yes thanks @SaadAhmed96 that solved the issue.

[joshw66]

I'm experiencing the same error on the example above. Shap 0.28.5 and Xgboost 0.82:

X,y = shap.datasets.boston()
model = xgb.train({"learning_rate": 0.01}, xgb.DMatrix(X, label=y), 100)

explainer = shap.TreeExplainer(model)
shap_values = explainer.shap_values(X.as_matrix())

shap.force_plot(explainer.expected_value, shap_values[0,:], X.iloc[0,:])

TypeError: predict() got an unexpected keyword argument 'validate_features'

[slundberg]

@joshw66 that's surprising. Could you share the full output (along with xgboost.__version__ just to be sure)

[joshw66]

@slundberg sorry eventually realised it was an issue with installing most recent version of xgboost, worked fine once I'd resoved that. Thanks for getting back anyway

[xinzhou1023]

Is there any update on this issue? I'm having the same issue with xgboost '0.7.post3' and shap '0.29.3'. @slundberg

[Primestone]

it seems xgboost 0.7 version have no 'validate_features' but '_validate_features'