uniquify based on HF username

evaluate.py

@@ -255,6 +255,7 @@ def _evaluate_data(filename: str, test_repo: str, split_filename: str, results_r
     results_df['submission_time'] = timestamp
     results_df['model_report'] = report
     results_df['anonymous'] = meta.participant.anonymous
+    results_df['hf_username'] = username
     safe_user = _unsafify_username(username)
     destination_path = f"results/{safe_user}_{timestamp}_results.csv"
     tmp_name = None

utils.py

@@ -28,6 +28,7 @@ def fetch_dataset_df():
         'submission_time': Value('string'),
         'model_report': Value('string'), 
         'anonymous': Value('bool'), 
+        'hf_username': Value('string')
     }
     feature_schema = Features(metric_features | other_features)
 
@@ -51,7 +52,7 @@ def fetch_dataset_df():
     # Get the most recent submission per user & endpoint
     latest = (
         df.sort_values("submission_time")
-          .drop_duplicates(subset=["Endpoint", "user"], keep="last")
+          .drop_duplicates(subset=["Endpoint", "hf_username"], keep="last") #IMPORTANT: unique on HF username not display name
           .sort_values(["Endpoint", "user"])
           .reset_index(drop=True)
     )