columns visibility

.gitignore

@@ -0,0 +1 @@
+venv

app.py

@@ -23,12 +23,15 @@ def load_data(file_path):
 # Function to style the DataFrame
 @st.cache_data
 def style_dataframe(df: pd.DataFrame):
-    df[RESULTS_COLUMN_NAME] = df.apply(lambda row: [row[SENTIMENT_COLUMN_NAME], row[UNDERSTANDING_COLUMN_NAME], row[PHRASEOLOGY_COLUMN_NAME]], axis=1)
-    # Insert the new column after the 'Average' column
-    cols = list(df.columns)
-    cols.insert(cols.index(AVERAGE_COLUMN_NAME) + 1, cols.pop(cols.index(RESULTS_COLUMN_NAME)))
-    df = df[cols]
-    # Create a color ramp using Seaborn
+    # Sprawdzamy, czy kolumna Average istnieje w ramce danych
+    if AVERAGE_COLUMN_NAME in df.columns:
+        # Jeśli wszystkie potrzebne kolumny są dostępne, tworzymy Results
+        if all(col in df.columns for col in [SENTIMENT_COLUMN_NAME, UNDERSTANDING_COLUMN_NAME, PHRASEOLOGY_COLUMN_NAME]):
+            df[RESULTS_COLUMN_NAME] = df.apply(lambda row: [row[SENTIMENT_COLUMN_NAME], row[UNDERSTANDING_COLUMN_NAME], row[PHRASEOLOGY_COLUMN_NAME]], axis=1)
+            # Insert the new column after the 'Average' column
+            cols = list(df.columns)
+            cols.insert(cols.index(AVERAGE_COLUMN_NAME) + 1, cols.pop(cols.index(RESULTS_COLUMN_NAME)))
+            df = df[cols]
     return df
 
 def styler(df: pd.DataFrame):
@@ -149,7 +152,7 @@ with tab1:
 
     # Closing filters in a expander
     with st.expander("Filtering benchmark data", icon='🔍'):
-        # Filtering data, e.g. slider for params, average score, etc.
+        # Filtry powinny być dostępne tylko dla kolumn, które są w danych
         col_filter_params, col_filter_average, col_filter_sentiment, col_filter_understanding, col_filter_phraseology = st.columns(5, gap='medium')
     
         with col_filter_params:
@@ -173,21 +176,29 @@ with tab1:
                 )
             ]
 
-        with col_filter_average:
-            average_slider = st.slider("Average score", step=0.1, min_value=0.0, max_value=5.0, value=(0.0, 5.0))
-            data = data[(data[AVERAGE_COLUMN_NAME] >= average_slider[0]) & (data[AVERAGE_COLUMN_NAME] <= average_slider[1])]
-
-        with col_filter_sentiment:
-            sentiment_slider = st.slider("Sentiment score", step=0.1, min_value=0.0, max_value=5.0, value=(0.0, 5.0))
-            data = data[(data[SENTIMENT_COLUMN_NAME] >= sentiment_slider[0]) & (data[SENTIMENT_COLUMN_NAME] <= sentiment_slider[1])]
-
-        with col_filter_understanding:
-            understanding_slider = st.slider("Understanding score", step=0.1, min_value=0.0, max_value=5.0, value=(0.0, 5.0))
-            data = data[(data[UNDERSTANDING_COLUMN_NAME] >= understanding_slider[0]) & (data[UNDERSTANDING_COLUMN_NAME] <= understanding_slider[1])]
-
-        with col_filter_phraseology:
-            phraseology_slider = st.slider("Phraseology score", step=0.1, min_value=0.0, max_value=5.0, value=(0.0, 5.0))
-            data = data[(data[PHRASEOLOGY_COLUMN_NAME] >= phraseology_slider[0]) & (data[PHRASEOLOGY_COLUMN_NAME] <= phraseology_slider[1])]
+        # Sprawdzamy, czy kolumna Average jest dostępna przed dodaniem filtra
+        if AVERAGE_COLUMN_NAME in data.columns:
+            with col_filter_average:
+                average_slider = st.slider("Average score", step=0.1, min_value=0.0, max_value=5.0, value=(0.0, 5.0))
+                data = data[(data[AVERAGE_COLUMN_NAME] >= average_slider[0]) & (data[AVERAGE_COLUMN_NAME] <= average_slider[1])]
+
+        # Sprawdzamy, czy kolumna Sentiment jest dostępna przed dodaniem filtra
+        if SENTIMENT_COLUMN_NAME in data.columns:
+            with col_filter_sentiment:
+                sentiment_slider = st.slider("Sentiment score", step=0.1, min_value=0.0, max_value=5.0, value=(0.0, 5.0))
+                data = data[(data[SENTIMENT_COLUMN_NAME] >= sentiment_slider[0]) & (data[SENTIMENT_COLUMN_NAME] <= sentiment_slider[1])]
+
+        # Sprawdzamy, czy kolumna Understanding jest dostępna przed dodaniem filtra
+        if UNDERSTANDING_COLUMN_NAME in data.columns:
+            with col_filter_understanding:
+                understanding_slider = st.slider("Understanding score", step=0.1, min_value=0.0, max_value=5.0, value=(0.0, 5.0))
+                data = data[(data[UNDERSTANDING_COLUMN_NAME] >= understanding_slider[0]) & (data[UNDERSTANDING_COLUMN_NAME] <= understanding_slider[1])]
+
+        # Sprawdzamy, czy kolumna Phraseology jest dostępna przed dodaniem filtra
+        if PHRASEOLOGY_COLUMN_NAME in data.columns:
+            with col_filter_phraseology:
+                phraseology_slider = st.slider("Phraseology score", step=0.1, min_value=0.0, max_value=5.0, value=(0.0, 5.0))
+                data = data[(data[PHRASEOLOGY_COLUMN_NAME] >= phraseology_slider[0]) & (data[PHRASEOLOGY_COLUMN_NAME] <= phraseology_slider[1])]
 
         # Extract unique provider names from the "Model" column
         providers = data["Model"].apply(lambda x: x.split('/')[0].lower()).unique()
@@ -195,42 +206,84 @@ with tab1:
         # Filter data based on selected providers
         data = data[data["Model"].apply(lambda x: x.split('/')[0].lower()).isin(selected_providers)]
 
+    # Dodanie filtra widocznych kolumn - bez kolumny Results, która jest tworzona później
+    base_columns = ["Model", "Params", AVERAGE_COLUMN_NAME, 
+                  SENTIMENT_COLUMN_NAME, UNDERSTANDING_COLUMN_NAME, PHRASEOLOGY_COLUMN_NAME]
+    default_columns = base_columns.copy()
+    selected_columns = st.multiselect("Visible columns", base_columns, default=default_columns)
+
     ### Display data
+    # Przygotowujemy pełne dane ze wszystkimi kolumnami
     styled_df_show = style_dataframe(data)
     styled_df_show = styler(styled_df_show)
 
-    st.data_editor(styled_df_show, column_config={
+    # Dostosowanie column_config do wybranych kolumn
+    column_config = {
         "Model": st.column_config.TextColumn("Model", help="Model name", width="large"),
-        "Params": st.column_config.NumberColumn("Params [B]"),
-        AVERAGE_COLUMN_NAME: st.column_config.NumberColumn(AVERAGE_COLUMN_NAME),
-        RESULTS_COLUMN_NAME: st.column_config.BarChartColumn(
+    }
+    
+    # Ustawiamy konfigurację dla wszystkich kolumn
+    if "Params" in styled_df_show.columns:
+        column_config["Params"] = st.column_config.NumberColumn("Params [B]") if "Params" in selected_columns else None
+    
+    if AVERAGE_COLUMN_NAME in styled_df_show.columns:
+        column_config[AVERAGE_COLUMN_NAME] = st.column_config.NumberColumn(AVERAGE_COLUMN_NAME) if AVERAGE_COLUMN_NAME in selected_columns else None
+    
+    if RESULTS_COLUMN_NAME in styled_df_show.columns:
+        # Results pokazujemy tylko jeśli Average jest wybrane
+        column_config[RESULTS_COLUMN_NAME] = st.column_config.BarChartColumn(
             "Bar chart of results", help="Summary of the results of each task",
-            y_min=0,y_max=5,),
-        SENTIMENT_COLUMN_NAME: st.column_config.NumberColumn(SENTIMENT_COLUMN_NAME, help='Ability to analyze sentiment'),
-        UNDERSTANDING_COLUMN_NAME: st.column_config.NumberColumn(UNDERSTANDING_COLUMN_NAME, help='Ability to understand language'),
-        PHRASEOLOGY_COLUMN_NAME: st.column_config.NumberColumn(PHRASEOLOGY_COLUMN_NAME, help='Ability to understand phraseological compounds'),
-        }, hide_index=True, disabled=True, height=500)
-
-    # Add selection for models and create a bar chart for selected models using the AVERAGE_COLUMN_NAME, SENTIMENT_COLUMN_NAME, PHRASEOLOGY_COLUMN_NAME, UNDERSTANDING_COLUMN_NAME
-    # Add default selection of 3 best models from AVERAGE_COLUMN_NAME and 1 best model with "Bielik" in Model column
-    default_models = list(data.sort_values(AVERAGE_COLUMN_NAME, ascending=False)['Model'].head(3))
-    bielik_model = data[data['Model'].str.contains('Bielik')].sort_values(AVERAGE_COLUMN_NAME, ascending=False)['Model'].iloc[0]
-    if bielik_model not in default_models:
-        default_models.append(bielik_model)
-    selected_models = st.multiselect("Select models to compare", data["Model"].unique(), default=default_models)
-    selected_data = data[data["Model"].isin(selected_models)]
-    categories = [AVERAGE_COLUMN_NAME, SENTIMENT_COLUMN_NAME, PHRASEOLOGY_COLUMN_NAME, UNDERSTANDING_COLUMN_NAME]
-
-    if selected_models:
+            y_min=0, y_max=5) if AVERAGE_COLUMN_NAME in selected_columns else None
+    
+    if SENTIMENT_COLUMN_NAME in styled_df_show.columns:
+        column_config[SENTIMENT_COLUMN_NAME] = st.column_config.NumberColumn(SENTIMENT_COLUMN_NAME, help='Ability to analyze sentiment') if SENTIMENT_COLUMN_NAME in selected_columns else None
+    
+    if UNDERSTANDING_COLUMN_NAME in styled_df_show.columns:
+        column_config[UNDERSTANDING_COLUMN_NAME] = st.column_config.NumberColumn(UNDERSTANDING_COLUMN_NAME, help='Ability to understand language') if UNDERSTANDING_COLUMN_NAME in selected_columns else None
+    
+    if PHRASEOLOGY_COLUMN_NAME in styled_df_show.columns:
+        column_config[PHRASEOLOGY_COLUMN_NAME] = st.column_config.NumberColumn(PHRASEOLOGY_COLUMN_NAME, help='Ability to understand phraseological compounds') if PHRASEOLOGY_COLUMN_NAME in selected_columns else None
+    
+    st.data_editor(styled_df_show, column_config=column_config, hide_index=True, disabled=True, height=500)
+
+    # Załaduj oryginalne dane do filtrowania i sortowania
+    original_data = load_data('data.json')
+    
+    # Używamy oryginalnych danych do wyboru modeli i tworzenia wykresu
+    try:
+        default_models = list(data.sort_values(AVERAGE_COLUMN_NAME, ascending=False)['Model'].head(3))
+        if 'Bielik' in ' '.join(data['Model'].tolist()):
+            bielik_model = data[data['Model'].str.contains('Bielik')].sort_values(AVERAGE_COLUMN_NAME, ascending=False)['Model'].iloc[0]
+            if bielik_model not in default_models:
+                default_models.append(bielik_model)
+    except (KeyError, ValueError):
+        # Jeśli Average nie jest dostępne, używamy pierwszych 3 modeli bez sortowania
+        default_models = list(data['Model'].head(3))
+        if 'Bielik' in ' '.join(data['Model'].tolist()):
+            try:
+                bielik_model = data[data['Model'].str.contains('Bielik')]['Model'].iloc[0]
+                if bielik_model not in default_models:
+                    default_models.append(bielik_model)
+            except (IndexError, KeyError):
+                pass
+    
+    selected_models = st.multiselect("Select models to compare", original_data["Model"].unique(), default=default_models)
+    selected_data = original_data[original_data["Model"].isin(selected_models)]
+    
+    # Dostosowanie kategorii do dostępnych kolumn
+    available_categories = [col for col in [AVERAGE_COLUMN_NAME, SENTIMENT_COLUMN_NAME, PHRASEOLOGY_COLUMN_NAME, UNDERSTANDING_COLUMN_NAME] 
+                           if col in selected_data.columns]
+    
+    if selected_models and available_categories:
         # Kolorki do wyboru:
         # colors = px.colors.sample_colorscale("viridis", len(selected_models)+1)
         colors = px.colors.qualitative.G10[:len(selected_models)]
 
         fig_bars = go.Figure()
         for model, color in zip(selected_models, colors):
-            values = selected_data[selected_data['Model'] == model][categories].values.flatten().tolist()
+            values = selected_data[selected_data['Model'] == model][available_categories].values.flatten().tolist()
             fig_bars.add_trace(go.Bar(
-                x=categories,
+                x=available_categories,
                 y=values,
                 name=model,
                 marker_color=color
@@ -246,6 +299,8 @@ with tab1:
         )
         fig_bars.update_yaxes(range=[0, 5.1])
         st.plotly_chart(fig_bars)
+    elif selected_models:
+        st.warning("Brak dostępnych kategorii do porównania. Upewnij się, że wybrano przynajmniej jedną z kolumn: Average, Sentiment, Phraseology, Language understanding.")
 
 
 ### Zakładka 2 --> Opis