mean + std

app.py

@@ -162,7 +162,10 @@ def prepare_for_group_plotting(metric, top_k, direction: PARTITION_OPTIONS, rege
     else:
         keys = heapq.nsmallest(top_k, means, key=means.get)
 
-    return [(key, means[key]) for key in keys]
+    
+    means = [means[key] for key in keys]
+    stds = [metric[key].standard_deviation for key in keys]
+    return keys, means, stds
 
 
 def set_alpha(color, alpha):
@@ -237,11 +240,15 @@ def plot_bars(
     y = []
 
     for i, (name, histogram) in enumerate(progress.tqdm(data.items(), total=len(data), desc="Plotting...")):
-        histogram_prepared = prepare_for_group_plotting(histogram, top_k, direction, regex, rounding)
-        x, y = zip(*histogram_prepared)
+        x, y, stds = prepare_for_group_plotting(histogram, top_k, direction, regex, rounding)
 
-        fig.add_trace(go.Bar(x=x, y=y, name=name, marker=dict(color=set_alpha(px.colors.qualitative.Plotly[i % len(px.colors.qualitative.Plotly)], 0.5))))
-    
+        fig.add_trace(go.Bar(
+            x=x, 
+            y=y, 
+            name=f"{name} Mean", 
+            marker=dict(color=set_alpha(px.colors.qualitative.Plotly[i % len(px.colors.qualitative.Plotly)], 0.5)),
+            error_y=dict(type='data', array=stds, visible=True)
+        ))
 
     fig.update_layout(
         title=f"Bar Plots for {metric_name}",
@@ -533,7 +540,7 @@ The data might not be 100% representative, due to the sampling and optimistic me
         if not new_dsts:
             return gr.update(value=list(selected_runs))
         dst_union = new_dsts.union(selected_runs or [])
-        return gr.update(value=list(dst_union))
+        return gr.update(value=sorted(list(dst_union)))
 
     regex_button.click(
         fn=update_datasets_with_regex,