fix leaderboard history

Leaderboard/app.py

@@ -7,11 +7,11 @@ import numpy as np
 from utils import load_leaderboard, custom_css, load_ground_truth_references, calculate_comprehensive_metrics, parse_submission_csv
 
 # Paths
-# Paths
-LEADERBOARD_TSV = "./leaderboard.tsv"
-# Note: Keeping typo 'groud_truth.csv' as per existing file unless renamed
-GROUND_TRUTH_CSV = "groud_truth.csv" 
-ALLOWED_TEAMS_FILE = "list.txt"
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+LEADERBOARD_TSV = os.path.join(BASE_DIR, "leaderboard.tsv")
+# Fixed typo 'groud_truth.csv' -> 'ground_truth.csv'
+GROUND_TRUTH_CSV = os.path.join(BASE_DIR, "ground_truth.csv")
+ALLOWED_TEAMS_FILE = os.path.join(BASE_DIR, "list.txt")
 
 LEADERBOARD_COLUMNS = [
     "Team", "F1-score", "Precision", "Recall", "Correct Rate", 
@@ -216,12 +216,17 @@ def submit_and_update_leaderboard(team_name, submission_file):
         
         # Logic for saving history and updating file
         # Create history directory if not exists
-        history_dir = "leaderboard_history"
+        history_dir = os.path.join(BASE_DIR, "leaderboard_history")
         os.makedirs(history_dir, exist_ok=True)
         
+        # Create dated subfolder
+        date_str = datetime.now().strftime("%Y-%m-%d")
+        dated_dir = os.path.join(history_dir, date_str)
+        os.makedirs(dated_dir, exist_ok=True)
+        
         # Backup current leaderboard
-        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-        backup_path = os.path.join(history_dir, f"leaderboard_{timestamp}.tsv")
+        timestamp = datetime.now().strftime("%H%M%S")
+        backup_path = os.path.join(dated_dir, f"leaderboard_{timestamp}.tsv")
         if os.path.exists(LEADERBOARD_TSV):
             shutil.copy2(LEADERBOARD_TSV, backup_path)
         
@@ -249,76 +254,6 @@ def submit_and_update_leaderboard(team_name, submission_file):
         import traceback
         error_msg = f"Error during submission: {str(e)}\n{traceback.format_exc()}"
         return error_msg, get_latest_leaderboard()
-        
-        if error_message:
-            return error_message, get_latest_leaderboard()
-
-        if metrics is None:
-            return "Error: Metrics could not be calculated.", get_latest_leaderboard()
-        
-        # Map or Add 'Correct Rate' if present in metrics (e.g. from Accuracy)
-        if "Accuracy" in metrics:
-            metrics["Correct Rate"] = metrics["Accuracy"]
-        
-        new_f1 = metrics.get("F1-score", 0)
-        
-        # Check if team already exists
-        if team_name in current_df["Team"].values:
-            existing_row = current_df[current_df["Team"] == team_name]
-            # Use raw value not string representation if possible, assume load_leaderboard parses as float where possible
-            old_f1 = existing_row["F1-score"].values[0] if "F1-score" in existing_row.columns and pd.notna(existing_row["F1-score"].values[0]) else -1.0
-            
-            if new_f1 <= old_f1:
-                return f"Submission successful but no improvement. New F1: {new_f1:.4f} <= Old F1: {old_f1:.4f}", get_latest_leaderboard()
-            
-            # Improvement found, update existing entry
-            for metric_name, metric_value in metrics.items():
-                if metric_name in current_df.columns:
-                    current_df.loc[current_df["Team"] == team_name, metric_name] = metric_value
-        else:
-            # Add new entry with all metrics
-            new_entry_dict = {"Team": team_name}
-            new_entry_dict.update(metrics)
-            
-            # Ensure only desired columns
-            filtered_entry = {k: v for k, v in new_entry_dict.items() if k in LEADERBOARD_COLUMNS}
-            # Fill missing
-            for col in LEADERBOARD_COLUMNS:
-                if col not in filtered_entry:
-                    filtered_entry[col] = None
-            
-            new_entry = pd.DataFrame([filtered_entry])
-            current_df = pd.concat([current_df, new_entry], ignore_index=True)
-        
-        # Logic for saving history and updating file
-        # Create history directory if not exists
-        history_dir = "leaderboard_history"
-        os.makedirs(history_dir, exist_ok=True)
-        
-        # Backup current leaderboard
-        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-        backup_path = os.path.join(history_dir, f"leaderboard_{timestamp}.tsv")
-        if os.path.exists(LEADERBOARD_TSV):
-            shutil.copy2(LEADERBOARD_TSV, backup_path)
-        
-        # Enforce columns before sort and save
-        current_df = current_df[LEADERBOARD_COLUMNS]
-
-        # Sort by F1-score (descending) 
-        if "F1-score" in current_df.columns:
-            current_df = current_df.sort_values(by="F1-score", ascending=False).reset_index(drop=True)
-        
-        current_df.to_csv(LEADERBOARD_TSV, index=False)
-        
-        metrics_summary = f"Submission successful! New High Score!\n"
-        for metric_name, metric_value in metrics.items():
-            metrics_summary += f"{metric_name}: {metric_value}\n"
-        
-        return metrics_summary, get_latest_leaderboard()
-    except Exception as e:
-        import traceback
-        error_msg = f"Error during submission: {str(e)}\n{traceback.format_exc()}"
-        return error_msg, get_latest_leaderboard()
 
 def create_ui():
     with gr.Blocks(theme=gr.themes.Soft(text_size=gr.themes.sizes.text_md), css=custom_css) as demo:

Leaderboard/utils.py

@@ -226,6 +226,9 @@ def calculate_comprehensive_metrics(submission_file_obj, references):
     metrics = {}
     error_message = None
     
+    if not references:
+        return None, "Error: Ground truth references could not be loaded. Please ensure 'ground_truth.csv' exists locally or HF_TOKEN is set to access 'IqraEval/QuranMB.v2.labels'."
+
     temp_dir = tempfile.mkdtemp()
     try:
         temp_aligned_dir = os.path.join(temp_dir, "aligned")

app.py

@@ -7,11 +7,11 @@ import numpy as np
 from utils import load_leaderboard, custom_css, load_ground_truth_references, calculate_comprehensive_metrics, parse_submission_csv
 
 # Paths
-# Paths
-LEADERBOARD_TSV = "./leaderboard.tsv"
-# Note: Keeping typo 'groud_truth.csv' as per existing file unless renamed
-GROUND_TRUTH_CSV = "groud_truth.csv" 
-ALLOWED_TEAMS_FILE = "list.txt"
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+LEADERBOARD_TSV = os.path.join(BASE_DIR, "leaderboard.tsv")
+# Fixed typo 'groud_truth.csv' -> 'ground_truth.csv'
+GROUND_TRUTH_CSV = os.path.join(BASE_DIR, "ground_truth.csv")
+ALLOWED_TEAMS_FILE = os.path.join(BASE_DIR, "list.txt")
 
 LEADERBOARD_COLUMNS = [
     "Team", "F1-score", "Precision", "Recall", "Correct Rate", 
@@ -216,12 +216,17 @@ def submit_and_update_leaderboard(team_name, submission_file):
         
         # Logic for saving history and updating file
         # Create history directory if not exists
-        history_dir = "leaderboard_history"
+        history_dir = os.path.join(BASE_DIR, "leaderboard_history")
         os.makedirs(history_dir, exist_ok=True)
         
+        # Create dated subfolder
+        date_str = datetime.now().strftime("%Y-%m-%d")
+        dated_dir = os.path.join(history_dir, date_str)
+        os.makedirs(dated_dir, exist_ok=True)
+        
         # Backup current leaderboard
-        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-        backup_path = os.path.join(history_dir, f"leaderboard_{timestamp}.tsv")
+        timestamp = datetime.now().strftime("%H%M%S")
+        backup_path = os.path.join(dated_dir, f"leaderboard_{timestamp}.tsv")
         if os.path.exists(LEADERBOARD_TSV):
             shutil.copy2(LEADERBOARD_TSV, backup_path)
         
@@ -249,76 +254,6 @@ def submit_and_update_leaderboard(team_name, submission_file):
         import traceback
         error_msg = f"Error during submission: {str(e)}\n{traceback.format_exc()}"
         return error_msg, get_latest_leaderboard()
-        
-        if error_message:
-            return error_message, get_latest_leaderboard()
-
-        if metrics is None:
-            return "Error: Metrics could not be calculated.", get_latest_leaderboard()
-        
-        # Map or Add 'Correct Rate' if present in metrics (e.g. from Accuracy)
-        if "Accuracy" in metrics:
-            metrics["Correct Rate"] = metrics["Accuracy"]
-        
-        new_f1 = metrics.get("F1-score", 0)
-        
-        # Check if team already exists
-        if team_name in current_df["Team"].values:
-            existing_row = current_df[current_df["Team"] == team_name]
-            # Use raw value not string representation if possible, assume load_leaderboard parses as float where possible
-            old_f1 = existing_row["F1-score"].values[0] if "F1-score" in existing_row.columns and pd.notna(existing_row["F1-score"].values[0]) else -1.0
-            
-            if new_f1 <= old_f1:
-                return f"Submission successful but no improvement. New F1: {new_f1:.4f} <= Old F1: {old_f1:.4f}", get_latest_leaderboard()
-            
-            # Improvement found, update existing entry
-            for metric_name, metric_value in metrics.items():
-                if metric_name in current_df.columns:
-                    current_df.loc[current_df["Team"] == team_name, metric_name] = metric_value
-        else:
-            # Add new entry with all metrics
-            new_entry_dict = {"Team": team_name}
-            new_entry_dict.update(metrics)
-            
-            # Ensure only desired columns
-            filtered_entry = {k: v for k, v in new_entry_dict.items() if k in LEADERBOARD_COLUMNS}
-            # Fill missing
-            for col in LEADERBOARD_COLUMNS:
-                if col not in filtered_entry:
-                    filtered_entry[col] = None
-            
-            new_entry = pd.DataFrame([filtered_entry])
-            current_df = pd.concat([current_df, new_entry], ignore_index=True)
-        
-        # Logic for saving history and updating file
-        # Create history directory if not exists
-        history_dir = "leaderboard_history"
-        os.makedirs(history_dir, exist_ok=True)
-        
-        # Backup current leaderboard
-        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-        backup_path = os.path.join(history_dir, f"leaderboard_{timestamp}.tsv")
-        if os.path.exists(LEADERBOARD_TSV):
-            shutil.copy2(LEADERBOARD_TSV, backup_path)
-        
-        # Enforce columns before sort and save
-        current_df = current_df[LEADERBOARD_COLUMNS]
-
-        # Sort by F1-score (descending) 
-        if "F1-score" in current_df.columns:
-            current_df = current_df.sort_values(by="F1-score", ascending=False).reset_index(drop=True)
-        
-        current_df.to_csv(LEADERBOARD_TSV, index=False)
-        
-        metrics_summary = f"Submission successful! New High Score!\n"
-        for metric_name, metric_value in metrics.items():
-            metrics_summary += f"{metric_name}: {metric_value}\n"
-        
-        return metrics_summary, get_latest_leaderboard()
-    except Exception as e:
-        import traceback
-        error_msg = f"Error during submission: {str(e)}\n{traceback.format_exc()}"
-        return error_msg, get_latest_leaderboard()
 
 def create_ui():
     with gr.Blocks(theme=gr.themes.Soft(text_size=gr.themes.sizes.text_md), css=custom_css) as demo:

utils.py

@@ -226,6 +226,9 @@ def calculate_comprehensive_metrics(submission_file_obj, references):
     metrics = {}
     error_message = None
     
+    if not references:
+        return None, "Error: Ground truth references could not be loaded. Please ensure 'ground_truth.csv' exists locally or HF_TOKEN is set to access 'IqraEval/QuranMB.v2.labels'."
+
     temp_dir = tempfile.mkdtemp()
     try:
         temp_aligned_dir = os.path.join(temp_dir, "aligned")