sync from github

open-moe-llm-leaderboard-gh/src/backend/envs.py

@@ -58,6 +58,7 @@ class Tasks(Enum):
     # task20 = Task("race", "acc", "RACE", 0)
     task21 = Task("mmlu", "acc", "MMLU", 5)
     task22 = Task("gsm8k_custom", "em", "GSM8K", 5)
+    task23 = Task("gsm8k_cot", "em", "GSM8K", 8)
 
 
 EVAL_REQUESTS_PATH_BACKEND = os.path.join(CACHE_PATH, "eval-queue-bk")

open-moe-llm-leaderboard-gh/src/display/utils.py

@@ -122,8 +122,8 @@ for task in Tasks:
         continue
     # auto_eval_column_dict.append([f"{task.name}_prefilling_time", ColumnContent, ColumnContent(f"{task.value.col_name} {PREs}", "number", False, hidden=True)])
     auto_eval_column_dict.append([f"{task.name}_decoding_throughput", ColumnContent, ColumnContent(f"{task.value.col_name} {TS}", "number", True, hidden=True)])
-    auto_eval_column_dict.append([f"{task.name}_gpu_mbu", ColumnContent, ColumnContent(f"{task.value.col_name} {MBU}", "number", True, hidden=True)])
-    auto_eval_column_dict.append([f"{task.name}_gpu_mfu", ColumnContent, ColumnContent(f"{task.value.col_name} {MFU}", "number", True, hidden=True)])
+    auto_eval_column_dict.append([f"{task.name}_mbu", ColumnContent, ColumnContent(f"{task.value.col_name} {MBU}", "number", True, hidden=True)])
+    auto_eval_column_dict.append([f"{task.name}_mfu", ColumnContent, ColumnContent(f"{task.value.col_name} {MFU}", "number", True, hidden=True)])
 
 
 # Model information