update scripts

src/leaderboard/read_evals.py

@@ -77,8 +77,23 @@ class EvalResult:
             if isinstance(task.metric, str):
                 # accs = np.array([v.get(task.metric, None) for k, v in data["results"].items() if
                 #                  task.benchmark == k and isinstance(v, dict)])
-                accs = np.array([np.around(v*100, decimals=0) for k, v in data["results"].items() if task.benchmark == k])
-
+                # accs = np.array([np.around(v*100, decimals=0) for k, v in data["results"].items() if task.benchmark == k])
+                accs = []
+                import math
+                for k, v in data["results"].items():
+                    if task.benchmark == k:
+                        if isinstance(v, (int, float)) and not math.isnan(v):
+                            accs.append(np.around(v * 100, decimals=0))
+                        elif isinstance(v, list):
+                            # 处理列表中的每个数值
+                            accs.extend([np.around(x * 100, decimals=0) for x in v if
+                                         isinstance(x, (int, float)) and not math.isnan(x)])
+                        else:
+                            # 跳过 NaN 或不符合条件的值
+                            accs.append(None)
+
+                # 去掉 None 值（如果有）
+                accs = np.array([x for x in accs if x is not None])
                 # 过滤掉 None 值，确保 accs 只包含有效的数值
                 accs = accs[accs != None]