Update app.py

app.py

@@ -8,56 +8,30 @@ import re
 
 from huggingface_hub import login
 
-
 token = os.environ.get("HG_TOKEN")
 login(token)
 
-print("Loading dataset...")
 try:
     dataset = load_dataset("sudoping01/bambara-speech-recognition-benchmark", name="default")["eval"]
     references = {row["id"]: row["text"] for row in dataset}
-    print(f"Successfully loaded dataset with {len(references)} samples")
 except Exception as e:
-    print(f"Error loading dataset: {str(e)}")
     references = {}
-    print("WARNING: Using empty references dictionary due to dataset loading error")
 
-# Initialize leaderboard file with consistent column names
+
 leaderboard_file = "leaderboard.csv"
 if not os.path.exists(leaderboard_file):
-    # Create with Model_Name consistently
     pd.DataFrame(columns=["Model_Name", "WER", "CER", "Combined_Score", "timestamp"]).to_csv(leaderboard_file, index=False)
-    print("Created new leaderboard file")
-    
-    # Add example entries for first-time visitors
-    example_data = [
-        ["Example Model 1", 0.35, 0.20, 0.305, "2023-01-01 00:00:00"],
-        ["Example Model 2", 0.40, 0.18, 0.334, "2023-01-02 00:00:00"],
-        ["Example Model 3", 0.32, 0.25, 0.299, "2023-01-03 00:00:00"]
-    ]
-    example_df = pd.DataFrame(
-        example_data, 
-        columns=["Model_Name", "WER", "CER", "Combined_Score", "timestamp"]
-    )
-    example_df.to_csv(leaderboard_file, index=False)
-    print("Added example data to empty leaderboard for demonstration")
 else:
-    # Load existing leaderboard
     leaderboard_df = pd.read_csv(leaderboard_file)
     
-    # Rename column if needed for consistency
-    if "submitter" in leaderboard_df.columns and "Model_Name" not in leaderboard_df.columns:
-        leaderboard_df = leaderboard_df.rename(columns={"submitter": "Model_Name"})
-        leaderboard_df.to_csv(leaderboard_file, index=False)
-        print("Renamed 'submitter' column to 'Model_Name' for consistency")
+
+    # if "submitter" in leaderboard_df.columns and "Model_Name" not in leaderboard_df.columns:
+    #     leaderboard_df = leaderboard_df.rename(columns={"submitter": "Model_Name"})
+    #     leaderboard_df.to_csv(leaderboard_file, index=False)
     
-    # Add Combined_Score column if it doesn't exist
     if "Combined_Score" not in leaderboard_df.columns:
-        leaderboard_df["Combined_Score"] = leaderboard_df["WER"] * 0.7 + leaderboard_df["CER"] * 0.3 # WER 70% and CER 30%
+        leaderboard_df["Combined_Score"] = leaderboard_df["WER"] * 0.7 + leaderboard_df["CER"] * 0.3
         leaderboard_df.to_csv(leaderboard_file, index=False)
-        print("Added Combined_Score column to existing leaderboard")
-    
-    print(f"Loaded existing leaderboard with {len(leaderboard_df)} entries")
 
 def normalize_text(text):
     """
@@ -89,9 +63,7 @@ def calculate_metrics(predictions_df):
         reference = normalize_text(references[id_val])
         hypothesis = normalize_text(row["text"])
         
-        
         if not reference or not hypothesis:
-            print(f"Warning: Empty reference or hypothesis for ID {id_val}")
             continue
             
         reference_words = reference.split()
@@ -99,18 +71,15 @@ def calculate_metrics(predictions_df):
         reference_chars = list(reference)
         
         try:
-
             sample_wer = wer(reference, hypothesis)
             sample_cer = cer(reference, hypothesis)
             
             sample_wer = min(sample_wer, 2.0)  
             sample_cer = min(sample_cer, 2.0)  
             
-         
             total_ref_words += len(reference_words)
             total_ref_chars += len(reference_chars)
             
-                
             results.append({
                 "id": id_val,
                 "reference": reference,
@@ -120,13 +89,13 @@ def calculate_metrics(predictions_df):
                 "wer": sample_wer,
                 "cer": sample_cer
             })
-        except Exception as e:
-            print(f"Error calculating metrics for ID {id_val}: {str(e)}")
+        except Exception:
+            pass
     
     if not results:
         raise ValueError("No valid samples for WER/CER calculation")
         
-    # Calculate standard average metrics
+
     avg_wer = sum(item["wer"] for item in results) / len(results)
     avg_cer = sum(item["cer"] for item in results) / len(results)
     
@@ -142,7 +111,7 @@ def update_ranking(method):
         current_lb = pd.read_csv(leaderboard_file)
         
         if "Combined_Score" not in current_lb.columns:
-            current_lb["Combined_Score"] = current_lb["WER"] * 0.7 + current_lb["CER"] * 0.3 # 70% for WER
+            current_lb["Combined_Score"] = current_lb["WER"] * 0.7 + current_lb["CER"] * 0.3
         
         if method == "WER Only":
             return current_lb.sort_values("WER")
@@ -150,15 +119,12 @@ def update_ranking(method):
             return current_lb.sort_values("CER")
         else:  # Combined Score
             return current_lb.sort_values("Combined_Score")
-    except Exception as e:
-        print(f"Error updating ranking: {str(e)}")
-        # Return empty dataframe if something goes wrong
+    except Exception:
         return pd.DataFrame(columns=["Model_Name", "WER", "CER", "Combined_Score", "timestamp"])
 
 def process_submission(model_name, csv_file):
     try:
         df = pd.read_csv(csv_file)
-        print(f"Processing submission from {model_name} with {len(df)} rows")
         
         if len(df) == 0:
             return "Error: Uploaded CSV is empty.", None
@@ -170,7 +136,7 @@ def process_submission(model_name, csv_file):
             dup_ids = df[df["id"].duplicated()]["id"].unique()
             return f"Error: Duplicate IDs found: {', '.join(map(str, dup_ids[:5]))}", None
 
-        # Check if IDs match the reference dataset
+
         missing_ids = set(references.keys()) - set(df["id"])
         extra_ids = set(df["id"]) - set(references.keys())
         
@@ -180,7 +146,7 @@ def process_submission(model_name, csv_file):
         if extra_ids:
             return f"Error: Found {len(extra_ids)} extra IDs not in reference dataset. First few extra: {', '.join(map(str, list(extra_ids)[:5]))}", None
         
-        # Calculate WER and CER
+
         try:
             avg_wer, avg_cer, weighted_wer, weighted_cer, detailed_results = calculate_metrics(df)
             
@@ -194,7 +160,7 @@ def process_submission(model_name, csv_file):
         leaderboard = pd.read_csv(leaderboard_file)
         timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
         
-        #  (70% WER, 30% CER)
+        # Calculate combined score (70% WER, 30% CER)
         combined_score = avg_wer * 0.7 + avg_cer * 0.3
         
         new_entry = pd.DataFrame(
@@ -210,23 +176,6 @@ def process_submission(model_name, csv_file):
     except Exception as e:
         return f"Error processing submission: {str(e)}", None
 
-# Make sure we have at least some data for first-time visitors
-if os.path.exists(leaderboard_file):
-    leaderboard_df = pd.read_csv(leaderboard_file)
-    if len(leaderboard_df) == 0:
-        # Add example entries if leaderboard is empty
-        example_data = [
-            ["Example Model 1", 0.35, 0.20, 0.305, "2023-01-01 00:00:00"],
-            ["Example Model 2", 0.40, 0.18, 0.334, "2023-01-02 00:00:00"],
-            ["Example Model 3", 0.32, 0.25, 0.299, "2023-01-03 00:00:00"]
-        ]
-        example_df = pd.DataFrame(
-            example_data, 
-            columns=["Model_Name", "WER", "CER", "Combined_Score", "timestamp"]
-        )
-        example_df.to_csv(leaderboard_file, index=False)
-        print("Added example data to empty leaderboard for demonstration")
-
 with gr.Blocks(title="Bambara ASR Leaderboard") as demo:
     gr.Markdown(
         """
@@ -240,17 +189,13 @@ with gr.Blocks(title="Bambara ASR Leaderboard") as demo:
     with gr.Tabs() as tabs:
         with gr.TabItem("🏅 Current Rankings"):
             try:
-                # Load and make sure we have current leaderboard data
                 current_leaderboard = pd.read_csv(leaderboard_file)
                 
                 if "Combined_Score" not in current_leaderboard.columns:
                     current_leaderboard["Combined_Score"] = current_leaderboard["WER"] * 0.7 + current_leaderboard["CER"] * 0.3
                 
-                # Sort by combined score
                 current_leaderboard = current_leaderboard.sort_values("Combined_Score")
-            except Exception as e:
-                print(f"Error loading leaderboard: {str(e)}")
-                # Create empty dataframe if we can't load the file
+            except Exception:
                 current_leaderboard = pd.DataFrame(columns=["Model_Name", "WER", "CER", "Combined_Score", "timestamp"])
             
             gr.Markdown("### Current ASR Model Rankings")
@@ -310,7 +255,5 @@ with gr.Blocks(title="Bambara ASR Leaderboard") as demo:
                 outputs=[output_msg, leaderboard_display]
             )
 
-print("Starting Bambara ASR Leaderboard app...")
-
 if __name__ == "__main__":
     demo.launch(share=True)