upgrade reward math verification

Author: YeAnbang

applications/ColossalChat/coati/distributed/reward/reward_fn.py

@@ -1,4 +1,5 @@
 import torch
+from math_verify import parse, verify
 
 from .reward_utils import extract_boxed_solution, extract_solution, validate_response_structure
 
@@ -35,11 +36,7 @@ def math_reward_fn(input_ids, gt_answer, response_idx, **kwargs):
         format_acc += 1
 
     # Check answer accuracy, answer is considered correct if the answer is correct and the format is valid
-    if (
-        format_valid
-        and final_answer is not None
-        and gt_answer.strip().replace(" ", "").lower() == final_answer.strip().replace(" ", "").lower()
-    ):
+    if format_valid and final_answer is not None and verify(parse(gt_answer.strip()), parse(final_answer.strip())):
         ans_acc += 1
         reward += acc_score
 
@@ -91,7 +88,7 @@ def boxed_math_reward_fn(input_ids, gt_answer, response_idx, **kwargs):
         reward += format_score
 
     # Check answer accuracy, answer is considered correct if the answer is correct and the format is valid
-    if format_valid and final_answer is not None and gt_answer.strip().lower() == final_answer.strip().lower():
+    if format_valid and final_answer is not None and verify(parse(gt_answer.strip()), parse(final_answer.strip())):
         ans_acc += 1
         reward += acc_score