version v0.1

src/populate.py

@@ -14,15 +14,35 @@ def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchm
     all_data_json = [v.to_dict() for v in raw_data]
 
     df = pd.DataFrame.from_records(all_data_json)
-    df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
 
-    # Add ranking based on average score
-    df[AutoEvalColumn.rank.name] = range(1, len(df) + 1)
+    # If empty (e.g., first run on HF Spaces), return empty frame with expected columns
+    if df.empty:
+        return pd.DataFrame(columns=cols)
 
+    # Ensure Average exists; compute from available benchmark columns if missing
+    average_col = AutoEvalColumn.average.name
+    present_bench_cols = [c for c in benchmark_cols if c in df.columns]
+    if average_col not in df.columns and len(present_bench_cols) > 0:
+        df[average_col] = df[present_bench_cols].mean(axis=1)
+
+    # Sort if Average present
+    if average_col in df.columns:
+        df = df.sort_values(by=[average_col], ascending=False)
+
+    # Add rank if Average present
+    if average_col in df.columns:
+        df[AutoEvalColumn.rank.name] = range(1, len(df) + 1)
+
+    # Add any missing columns as NA and reorder
+    for c in cols:
+        if c not in df.columns:
+            df[c] = pd.NA
     df = df[cols].round(decimals=2)
 
-    # filter out if any of the benchmarks have not been produced
-    df = df[has_no_nan_values(df, benchmark_cols)]
+    # Filter NaNs only on existing benchmark columns
+    safe_benchmark_cols = [c for c in benchmark_cols if c in df.columns]
+    if len(safe_benchmark_cols) > 0:
+        df = df[has_no_nan_values(df, safe_benchmark_cols)]
     return df