Update src/leaderboard/read_evals.py

src/leaderboard/read_evals.py

@@ -68,27 +68,27 @@ class EvalResult:
             if architectures:
                 architecture = ";".join(architectures)
 
-        # Extract results available in this file (some results are split in several files)
-        results = {}
-        for task in Tasks:
-            task = task.value
-
-            # We average all scores of a given metric (not all metrics are present in all files)
-            accs = np.array([v for k, v in data[task.benchmark].items() if task.metric == k])
-            if accs.size == 0 or any([acc is None for acc in accs]):
-                continue
-
-            mean_acc = np.mean(accs) * 100.0
-            results[task.benchmark] = mean_acc
-        # if "scores" not in data:
-        #     raise KeyError(f"'scores' key not found in JSON file: {json_filepath}")
-
-        # scores = data["scores"]
+        # # Extract results available in this file (some results are split in several files)
         # results = {}
         # for task in Tasks:
         #     task = task.value
-        #     if task.metric in scores:
-        #         results[task.benchmark] = Decimal(scores[task.metric])
+
+        #     # We average all scores of a given metric (not all metrics are present in all files)
+        #     accs = np.array([v for k, v in data[task.benchmark].items() if task.metric == k])
+        #     if accs.size == 0 or any([acc is None for acc in accs]):
+        #         continue
+
+        #     mean_acc = np.mean(accs) * 100.0
+        #     results[task.benchmark] = mean_acc
+        if "scores" not in data:
+            raise KeyError(f"'scores' key not found in JSON file: {json_filepath}")
+
+        scores = data["scores"]
+        results = {}
+        for task in Tasks:
+            task = task.value
+            if task.metric in scores:
+                results[task.benchmark] = Decimal(scores[task.metric])
 
         return self(
             eval_name=result_key,