From c775c31cb30f0f6c32e742fb4d5fbed5be384573 Mon Sep 17 00:00:00 2001
From: Liangsheng Yin <hnyls2002@gmail.com>
Date: Sun, 12 Apr 2026 20:26:54 -0700
Subject: [PATCH] Simplify test_chunked_prefill; remove redundant tests
 (#22652)

---
 .../scheduler/test_chunked_prefill.py         | 35 ------------
 .../scheduler/test_mixed_chunked_prefill.py   | 55 +++++++++++++++++++
 2 files changed, 55 insertions(+), 35 deletions(-)
 delete mode 100644 test/registered/scheduler/test_chunked_prefill.py
 create mode 100644 test/registered/scheduler/test_mixed_chunked_prefill.py

diff --git a/test/registered/scheduler/test_chunked_prefill.py b/test/registered/scheduler/test_chunked_prefill.py
deleted file mode 100644
index a4257dedb..000000000
--- a/test/registered/scheduler/test_chunked_prefill.py
+++ /dev/null
@@ -1,35 +0,0 @@
-"""
-python3 -m unittest test_chunked_prefill.TestChunkedPrefill.test_mixed_chunked_prefill_without_radix_cache
-"""
-
-import unittest
-
-from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
-from sglang.test.test_utils import CustomTestCase, run_mmlu_test, run_mulit_request_test
-
-register_cuda_ci(est_time=360, suite="stage-b-test-1-gpu-small")
-register_amd_ci(est_time=312, suite="stage-b-test-1-gpu-small-amd")
-
-
-class TestChunkedPrefill(CustomTestCase):
-    def test_chunked_prefill(self):
-        run_mmlu_test(disable_radix_cache=False, enable_mixed_chunk=False)
-
-    def test_mixed_chunked_prefill(self):
-        run_mmlu_test(disable_radix_cache=False, enable_mixed_chunk=True)
-
-    def test_chunked_prefill_without_radix_cache(self):
-        run_mmlu_test(disable_radix_cache=True, enable_mixed_chunk=False)
-
-    def test_mixed_chunked_prefill_without_radix_cache(self):
-        run_mmlu_test(disable_radix_cache=True, enable_mixed_chunk=True)
-
-    def test_mixed_chunked_prefill_multi_requests(self):
-        run_mulit_request_test(
-            enable_mixed_chunk=True,
-            chunked_prefill_size=2048,
-        )
-
-
-if __name__ == "__main__":
-    unittest.main()
diff --git a/test/registered/scheduler/test_mixed_chunked_prefill.py b/test/registered/scheduler/test_mixed_chunked_prefill.py
new file mode 100644
index 000000000..895943394
--- /dev/null
+++ b/test/registered/scheduler/test_mixed_chunked_prefill.py
@@ -0,0 +1,55 @@
+import unittest
+
+from sglang.srt.environ import envs
+from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+from sglang.test.kits.eval_accuracy_kit import GSM8KMixin
+from sglang.test.test_utils import (
+    DEFAULT_MODEL_NAME_FOR_TEST,
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    popen_launch_server,
+)
+
+register_cuda_ci(est_time=180, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=180, suite="stage-b-test-1-gpu-small-amd")
+
+
+class TestMixedChunkedPrefill(GSM8KMixin, CustomTestCase):
+    model = DEFAULT_MODEL_NAME_FOR_TEST
+    base_url = DEFAULT_URL_FOR_TEST
+    gsm8k_accuracy_thres = 0.62
+
+    extra_args = [
+        "--enable-mixed-chunk",
+        "--chunked-prefill-size",
+        "32",
+    ]
+
+    @classmethod
+    def setUpClass(cls):
+        with envs.SGLANG_ENABLE_STRICT_MEM_CHECK_DURING_BUSY.override(2):
+            cls.process = popen_launch_server(
+                cls.model,
+                cls.base_url,
+                timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+                other_args=cls.extra_args,
+            )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+
+class TestMixedChunkedPrefillNoRadixCache(TestMixedChunkedPrefill):
+    extra_args = [
+        "--enable-mixed-chunk",
+        "--chunked-prefill-size",
+        "32",
+        "--disable-radix-cache",
+    ]
+
+
+if __name__ == "__main__":
+    unittest.main()