Update app.py

app.py

@@ -16,28 +16,40 @@ def cached_model_info(api, model):
 
 def convert_markdown_table_to_dataframe(md_content):
     cleaned_content = re.sub(r'\|\s*$', '', re.sub(r'^\|\s*', '', md_content, flags=re.MULTILINE), flags=re.MULTILINE)
-    df = pd.read_csv(StringIO(cleaned_content), sep="\|", engine='python', skipinitialspace=True)
-    df.columns = df.columns.str.strip()
+    df = pd.read_csv(StringIO(cleaned_content), sep="\|", engine='python')
+    df.columns = [col.strip() for col in df.columns]
+
+    # Initialize empty columns for URL and Model if not already present
+    if 'Model' not in df.columns:
+        df['Model'] = None
+    if 'URL' not in df.columns:
+        df['URL'] = None
+
+    model_link_pattern = r'\[(.*?)\]\((.*?)\)'
+    for index, row in df.iterrows():
+        match = re.search(model_link_pattern, row['Model'])
+        if match:
+            df.at[index, 'Model'] = match.group(1)
+            df.at[index, 'URL'] = match.group(2)
 
-    model_link_pattern = r'\[(.*?)\]\((.*?)\)\s*\[.*?\]\(.*?\)'
-    # Correctly process 'Model' column to extract URLs and model names
-    df['URL'] = df['Model'].apply(lambda x: re.search(model_link_pattern, x).group(2) if re.search(model_link_pattern, x) else None)
-    df['Model'] = df['Model'].apply(lambda x: re.sub(model_link_pattern, r'\1', x))
     return df
 
 def get_and_update_model_info(df):
     api = HfApi()
+    df['Likes'] = -1  # Default value for Likes
+    df['Tags'] = ''   # Default value for Tags
     for index, row in df.iterrows():
-        model_info = cached_model_info(api, row['Model'].strip())
-        if model_info:
-            df.at[index, 'Likes'] = model_info.likes
-            df.at[index, 'Tags'] = ', '.join(model_info.tags)
-        else:
-            df.at[index, 'Likes'] = -1
-            df.at[index, 'Tags'] = ''
+        if pd.notnull(row['Model']):
+            model_info = cached_model_info(api, row['Model'].strip())
+            if model_info:
+                df.at[index, 'Likes'] = model_info.likes
+                df.at[index, 'Tags'] = ', '.join(model_info.tags)
     return df
 
 def calculate_highest_combined_score(data, column):
+    # Ensure the column exists and has numeric data
+    if column not in data.columns or not pd.api.types.is_numeric_dtype(data[column]):
+        return column, {}
     scores = data[column].dropna().tolist()
     models = data['Model'].tolist()
     top_combinations = {r: [] for r in range(2, 7)}
@@ -51,8 +63,8 @@ def calculate_highest_combined_score(data, column):
 
 def display_highest_combined_scores(data, score_columns):
     for column in score_columns:
-        if column in data.columns:
-            _, top_combinations = calculate_highest_combined_score(data, column)
+        _, top_combinations = calculate_highest_combined_score(data, column)
+        if top_combinations:
             st.subheader(f"Top Combinations for {column}")
             for r, combinations in top_combinations.items():
                 st.write(f"**Number of Models: {r}**")
@@ -60,6 +72,7 @@ def display_highest_combined_scores(data, score_columns):
                     st.write(f"Score: {score}, Models: {', '.join(combination)}")
 
 
+
 # Function to create bar chart for a given category
 def create_bar_chart(df, category):
     """Create and display a bar chart for a given category."""
@@ -88,16 +101,18 @@ def create_bar_chart(df, category):
 def main():
     st.title("Model Leaderboard")
     st.markdown("Displaying top combinations of models based on scores.")
-    # Placeholder for actual markdown content
-    content = """Your markdown content here"""
+    content = """Your markdown content here"""  # Placeholder for actual markdown content
+
     if content:
         df = convert_markdown_table_to_dataframe(content)
         df = get_and_update_model_info(df)
         score_columns = ['Average', 'AGIEval', 'GPT4All', 'TruthfulQA', 'Bigbench']
         for col in score_columns:
-            df[col] = pd.to_numeric(df[col], errors='coerce')
+            if col in df.columns:
+                df[col] = pd.to_numeric(df[col], errors='coerce')
         display_highest_combined_scores(df, score_columns)
 
+
     # Create tabs for leaderboard and about section
     content = create_yall()
     tab1, tab2 = st.tabs(["🏆 Leaderboard", "📝 About"])