From cd65be98dfade5bb03d4a3bea0260df11d53f87b Mon Sep 17 00:00:00 2001
From: Baizhou Zhang <sobereddiezhang@gmail.com>
Date: Thu, 28 May 2026 14:34:41 -0700
Subject: [PATCH] [CP] Add back qwen 30b test (#26604)

---
 test/registered/cp/test_qwen3_30b.py | 132 +++++++++++++++++++++++++++
 1 file changed, 132 insertions(+)
 create mode 100644 test/registered/cp/test_qwen3_30b.py

diff --git a/test/registered/cp/test_qwen3_30b.py b/test/registered/cp/test_qwen3_30b.py
new file mode 100644
index 000000000..e21dafb5c
--- /dev/null
+++ b/test/registered/cp/test_qwen3_30b.py
@@ -0,0 +1,132 @@
+import unittest
+from types import SimpleNamespace
+
+from sglang.test.ci.ci_register import register_cuda_ci
+from sglang.test.run_eval import run_eval
+from sglang.test.test_utils import (
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    kill_process_tree,
+    popen_launch_server,
+)
+
+register_cuda_ci(est_time=261, stage="extra-b", runner_config="4-gpu-h100")
+
+QWEN3_30B_MODEL_PATH = "Qwen/Qwen3-30B-A3B-FP8"
+
+GSM8K_BASELINE_ACCURACY = 0.85
+
+
+class TestQwen330B(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = QWEN3_30B_MODEL_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=[
+                "--tp-size",
+                "4",
+                "--moe-dp-size",
+                "2",
+                "--ep-size",
+                "2",
+                "--attn-cp-size",
+                "2",
+                "--enable-prefill-context-parallel",
+                "--cuda-graph-max-bs",
+                "32",
+                "--max-running-requests",
+                "32",
+                "--trust-remote-code",
+                "--disable-piecewise-cuda-graph",
+                "--model-loader-extra-config",
+                '{"enable_multithread_load": true, "num_threads": 64}',
+            ],
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_gsm8k(self):
+        args = SimpleNamespace(
+            model=self.model,
+            eval_name="gsm8k",
+            num_shots=5,
+            num_examples=200,
+            max_tokens=16000,
+            num_threads=128,
+            repeat=1,
+            temperature=0.6,
+            top_p=0.95,
+            top_k=20,
+            base_url=self.base_url,
+            host="http://127.0.0.1",
+            port=int(self.base_url.split(":")[-1]),
+        )
+        metrics = run_eval(args)
+        print(f"{metrics=}")
+        self.assertGreaterEqual(metrics["score"], GSM8K_BASELINE_ACCURACY)
+
+
+class TestQwen330BCP(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = QWEN3_30B_MODEL_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=[
+                "--tp-size",
+                "4",
+                "--moe-dp-size",
+                "1",
+                "--ep-size",
+                "4",
+                "--attn-cp-size",
+                "2",
+                "--enable-prefill-context-parallel",
+                "--cuda-graph-max-bs",
+                "32",
+                "--max-running-requests",
+                "32",
+                "--trust-remote-code",
+                "--disable-piecewise-cuda-graph",
+                "--model-loader-extra-config",
+                '{"enable_multithread_load": true, "num_threads": 64}',
+            ],
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_gsm8k(self):
+        args = SimpleNamespace(
+            model=self.model,
+            eval_name="gsm8k",
+            num_shots=5,
+            num_examples=200,
+            max_tokens=16000,
+            num_threads=128,
+            repeat=1,
+            temperature=0.6,
+            top_p=0.95,
+            top_k=20,
+            base_url=self.base_url,
+            host="http://127.0.0.1",
+            port=int(self.base_url.split(":")[-1]),
+        )
+        metrics = run_eval(args)
+        print(f"{metrics=}")
+        self.assertGreaterEqual(metrics["score"], GSM8K_BASELINE_ACCURACY)
+
+
+if __name__ == "__main__":
+    unittest.main()