Unify GSM8K eval path to Chat API for regression CI readiness (#21667)

test/manual/test_torch_flex_attention_backend.py

@@ -7,7 +7,7 @@ import unittest
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree
-from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
+from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -30,17 +30,17 @@ class TestTorchFlexAttnBackend(CustomTestCase):
 
         try:
             args = SimpleNamespace(
+                base_url=base_url,
+                eval_name="gsm8k",
+                api="completion",
+                max_tokens=512,
+                num_examples=100,
+                num_threads=10,
                 num_shots=8,
-                data_path=None,
-                num_questions=100,
-                parallel=10,
-                max_new_tokens=512,
-                host="http://127.0.0.1",
-                port=int(base_url.split(":")[-1]),
             )
-            metrics = run_eval_few_shot_gsm8k(args)
+            metrics = run_eval(args)
             print(f"{metrics=}")
-            self.assertGreater(metrics["accuracy"], 0.62)
+            self.assertGreater(metrics["score"], 0.62)
         finally:
             kill_process_tree(process.pid)