Add kimi k2 thinking to ci (#12907)

test/srt/models/test_kimi_k2_models.py

@@ -0,0 +1,68 @@
+import unittest
+from types import SimpleNamespace
+
+import requests
+
+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.test_utils import (
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    is_in_ci,
+    popen_launch_server,
+    write_github_step_summary,
+)
+
+
+class TestKimiK2Thinking(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = "moonshotai/Kimi-K2-Thinking"
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--tp",
+            "8",
+            "--trust-remote-code",
+            "--tool-call-parser",
+            "kimi_k2",
+            "--reasoning-parser",
+            "kimi_k2",
+        ]
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=other_args,
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_a_gsm8k(
+        self,
+    ):
+        requests.get(self.base_url + "/flush_cache")
+
+        args = SimpleNamespace(
+            num_shots=5,
+            data_path=None,
+            num_questions=200,
+            max_new_tokens=512,
+            parallel=128,
+            host="http://127.0.0.1",
+            port=int(self.base_url.split(":")[-1]),
+        )
+        metrics = run_eval_few_shot_gsm8k(args)
+        print(f"{metrics=}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_gsm8k (Kimi-K2-Thinking)\n" f'{metrics["accuracy"]=:.3f}\n'
+            )
+            self.assertGreater(metrics["accuracy"], 0.95)
+
+
+if __name__ == "__main__":
+    unittest.main()

test/srt/run_suite.py

@@ -168,6 +168,7 @@ suites = {
         TestFile("test_deepseek_v3_basic.py", 275),
         TestFile("test_deepseek_v3_mtp.py", 275),
         TestFile("test_disaggregation_hybrid_attention.py", 200),
+        TestFile("models/test_kimi_k2_models.py", 200),
     ],
     "per-commit-8-gpu-h20": [
         TestFile("quant/test_w4a8_deepseek_v3.py", 520),