Revert "[feat] Add math eval to CI" (#2656)

test/srt/test_eval_accuracy_large.py

@@ -68,17 +68,6 @@ class TestEvalAccuracyLarge(unittest.TestCase):
         metrics = run_eval(args)
         self.assertGreater(metrics["score"], 0.835)
 
-    def test_math(self):
-        args = SimpleNamespace(
-            base_url=self.base_url,
-            model=self.model,
-            eval_name="math",
-            num_examples=5000,
-            num_threads=1024
-        )
-
-        metrics = run_eval(args)
-        self.assertGreaterEqual(metrics["score"], 0.519 - 0.01) # -1% to account for sampling variance
 
 if __name__ == "__main__":
     unittest.main()

test/srt/test_eval_accuracy_mini.py

@@ -37,18 +37,6 @@ class TestEvalAccuracyMini(unittest.TestCase):
         metrics = run_eval(args)
         self.assertGreaterEqual(metrics["score"], 0.65)
 
-    def test_math(self):
-        args = SimpleNamespace(
-            base_url=self.base_url,
-            model=self.model,
-            eval_name="math",
-            num_examples=64,
-            num_threads=32,
-            temperature=0.1,
-        )
 
-        metrics = run_eval(args)
-        self.assertGreaterEqual(metrics["score"], 0.519 - 0.03) # -3% to account for sampling variance
-        
 if __name__ == "__main__":
     unittest.main()