updated pos_label in all callbacks to deal with multiple workers

TODO.md

@@ -7,15 +7,19 @@
 - add multiclass confusion matrix
 - individual trees: highlight selected tree
 - Add feature names to waterfall plot
+- fix percentages difference bug lift plot vs classification plot
 
 ### Regression plots:
 
 ## Explainers:
+- add `get_metrics_dict()` function
 
 ## Dashboard:
+- FIX NOT LOADING UNTIL CLICKED ON CONTRIBUTIONS TAB BUG!
 - add option for vertical contributions?
 - reformat contributions table
 - add final prediction to contributions table
+- 
 
 ## Methods:
 

explainerdashboard/dashboard_tabs/contributions_tab.py

@@ -264,6 +264,7 @@ def display_value(contributions_size):
     def update_output_div(index, topx, pos_label):
         if index is None:
             raise PreventUpdate
+        explainer.pos_label = pos_label #needed in case of multiple workers
         int_idx = explainer.get_int_idx(index)
         if explainer.is_classifier:
             def display_probas(pred_probas_raw, labels, round=2):
@@ -308,4 +309,5 @@ def update_pdp_col(clickData):
          Input('label-store', 'data')]
     )
     def update_pdp_graph(idx, col, pos_label):
+        explainer.pos_label = pos_label #needed in case of multiple workers
         return explainer.plot_pdp(col, idx, sample=100)

explainerdashboard/dashboard_tabs/model_summary_tab.py

@@ -126,6 +126,7 @@ def register_callbacks(self, app, **kwargs):
             [State('tabs', 'value')]
         )
         def update_importances(tablesize, cats, permutation_shap, pos_label, tab): 
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             return self.explainer.plot_importances(
                         type=permutation_shap, topx=tablesize, cats=cats)
 
@@ -304,6 +305,7 @@ def update_div_visibility(bins_or_quantiles):
             [State('tabs', 'value')],
         )
         def update_precision_graph(percentage, cutoff, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             return self.explainer.plot_lift_curve(cutoff=cutoff, percentage=percentage)
 
         @app.callback(
@@ -317,6 +319,7 @@ def update_precision_graph(percentage, cutoff, pos_label, tab):
             [State('tabs', 'value')],
         )
         def update_precision_graph(bin_size, quantiles, bins, cutoff, multiclass, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             if bins=='bin_size':
                 return self.explainer.plot_precision(
                     bin_size=bin_size, cutoff=cutoff, multiclass=multiclass)
@@ -333,6 +336,7 @@ def update_precision_graph(bin_size, quantiles, bins, cutoff, multiclass, pos_la
             [State('tabs', 'value')],
         )
         def update_precision_graph(percentage, cutoff, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             return self.explainer.plot_classification(cutoff=cutoff, percentage=percentage)
 
         @app.callback(
@@ -343,6 +347,7 @@ def update_precision_graph(percentage, cutoff, pos_label, tab):
             [State('tabs', 'value')],
         )
         def update_precision_graph(cutoff, percentage, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             return self.explainer.plot_confusion_matrix(
                                 cutoff=cutoff, normalized=percentage)
 
@@ -353,6 +358,7 @@ def update_precision_graph(cutoff, percentage, pos_label, tab):
              Input('tabs', 'value')],
         )
         def update_precision_graph(cutoff, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             return self.explainer.plot_roc_auc(cutoff=cutoff)
 
         @app.callback(
@@ -362,6 +368,7 @@ def update_precision_graph(cutoff, pos_label, tab):
             [State('tabs', 'value')],
         )
         def update_precision_graph(cutoff, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             return self.explainer.plot_pr_auc(cutoff=cutoff)
 
         @app.callback(
@@ -452,13 +459,13 @@ def layout(self):
 
 
     def register_callbacks(self, app, **kwargs):
-
         @app.callback(
             Output('model-summary', 'children'),
             [Input('label-store', 'data')],
             [State('tabs', 'value')]
         )
         def update_model_summary(pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             rmse = np.round(np.sqrt(mean_squared_error(self.explainer.y, self.explainer.preds)), 2)
             mae = np.round(mean_absolute_error(self.explainer.y, self.explainer.preds), 2)
             r2 = np.round(r2_score(self.explainer.y, self.explainer.preds), 2)
@@ -480,6 +487,7 @@ def update_model_summary(pos_label, tab):
             [State('tabs', 'value')]
         )
         def update_predicted_vs_actual_graph(logs, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             return self.explainer.plot_predicted_vs_actual(logs=logs)
 
         @app.callback(
@@ -490,6 +498,7 @@ def update_predicted_vs_actual_graph(logs, pos_label, tab):
             [State('tabs', 'value')],
         )
         def update_residuals_graph(pred_or_actual, ratio, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             vs_actual = pred_or_actual=='vs_actual'
             return self.explainer.plot_residuals(vs_actual=vs_actual, ratio=ratio)
 
@@ -501,6 +510,7 @@ def update_residuals_graph(pred_or_actual, ratio, pos_label, tab):
             [State('tabs', 'value')],
         )
         def update_residuals_graph(col, ratio, pos_label, tab):
+            self.explainer.pos_label = pos_label #needed in case of multiple workers
             return self.explainer.plot_residuals_vs_feature(col, ratio=ratio, dropna=True)
 
 

explainerdashboard/dashboard_tabs/shadow_trees_tab.py

@@ -88,7 +88,6 @@ def shadow_trees_layout(explainer, round=2, **kwargs):
 
 
 def shadow_trees_callbacks(explainer, app, round=2, **kwargs):
-
     @app.callback(
         [Output('tree-basevalue', 'children'),
          Output('tree-predictions-table', 'columns'),
@@ -129,6 +128,7 @@ def display_click_data(clickData, idx, tab):
         [State('tabs', 'value')]
     )
     def update_tree_graph(index, pos_label, tab):
+        explainer.pos_label=pos_label
         if index is not None:
             return explainer.plot_trees(index, round=round)
         return {}

explainerdashboard/dashboard_tabs/shap_dependence_tab.py

@@ -134,6 +134,7 @@ def shap_dependence_callbacks(explainer, app, **kwargs):
          Input('label-store', 'data')],
         [State('tabs', 'value')])
     def update_dependence_shap_scatter_graph(summary_type, cats, depth, pos_label, tab):
+        explainer.pos_label = pos_label #needed in case of multiple workers
         ctx = dash.callback_context
         if ctx.triggered:
             if depth is None: depth = 10
@@ -199,6 +200,7 @@ def set_color_col_dropdown(col, cats):
         [State('dependence-col', 'value'),
          State('dependence-group-categoricals', 'checked')])
     def update_dependence_graph(color_col, idx, pos_label, col, cats):
+        explainer.pos_label = pos_label #needed in case of multiple workers
         if color_col is not None:
             return explainer.plot_shap_dependence(
                         col, color_col, highlight_idx=idx, cats=cats)

explainerdashboard/dashboard_tabs/shap_interactions_tab.py

@@ -168,6 +168,7 @@ def update_col_options(cats, col, tab):
         [State('interaction-group-categoricals', 'checked')])
     def update_interaction_scatter_graph(summary_type, col, depth, pos_label, cats):
         if col is not None:
+            explainer.pos_label = pos_label #needed in case of multiple workers
             if depth is None: 
                 depth = len(explainer.columns_ranked(cats))-1 
             if summary_type=='aggregate':
@@ -205,6 +206,7 @@ def display_scatter_click_data(clickdata):
          State('interaction-group-categoricals', 'checked')])
     def update_dependence_graph(interact_col, index, pos_label, col, cats):
         if interact_col is not None:
+            explainer.pos_label = pos_label #needed in case of multiple workers
             return (explainer.plot_shap_interaction_dependence(
                         col, interact_col, highlight_idx=index, cats=cats),
                     explainer.plot_shap_interaction_dependence(