Alan Melo Clappis

Try to install the lib without openmp: Set the property use_openmp in setup.py to False and install again. https://github.com/lyst/lightfm/blob/1f3f35d2813393eb269073150fdf762ff19f0b42/setup.py#L147 This could make the train process more slow, but should work

> I usually do this to get feature importance. > > ```python > vals= np.abs(shap_values).mean(0) > feature_importance = pd.DataFrame(list(zip(X_train.columns,vals)),columns=['col_name','feature_importance_vals']) > feature_importance.sort_values(by=['feature_importance_vals'],ascending=False,inplace=True) > feature_importance.head() > ``` I think the shap_values format...

It will be a nice feature, @dyaacov

Same issue here, the .deb fix the problem. Otherwise, the paste looks slow. I think that the problem is the new ubuntu (20.04). CopyQ has the same problem. https://github.com/hluk/CopyQ/issues/1382

@warrickguy yeah. https://github.com/CristianHenzel/ClipIt/releases/download/v1.4.5/clipit_1.4.5_amd64_indicator.deb

seems it only happens when early stopping is activated

maybe a workaround is change line 306 of feature_selector to train_features, valid_features, train_labels, valid_labels = train_test_split(features, labels, test_size = 0.15, **stratify=labels**)

I stopped use this extension because this bug too.