[NPU] update nightly tests (#17952)

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: cy <chenyang08056032@163.com>

test/registered/ascend/basic_function/interface/test_ascend_enable_thinking.py

@@ -0,0 +1,197 @@
+import json
+import unittest
+
+import requests
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.ascend.test_ascend_utils import QWEN3_30B_A3B_WEIGHTS_PATH
+from sglang.test.ci.ci_register import register_npu_ci
+from sglang.test.test_utils import (
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    popen_launch_server,
+)
+
+register_npu_ci(est_time=400, suite="nightly-2-npu-a3", nightly=True)
+
+
+class TestEnableThinking(CustomTestCase):
+    """Testcase: Testing with the 'enable_thinking' feature enabled/disabled,
+                 both streaming and non-streaming input requests successful
+
+    [Test Category] Interface
+    [Test Target] /v1/chat/completions
+    """
+
+    @classmethod
+    def setUpClass(cls):
+        cls.model = QWEN3_30B_A3B_WEIGHTS_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        cls.api_key = "sk-1234"
+        cls.other_args = [
+            "--reasoning-parser",
+            "qwen3",
+            "--attention-backend",
+            "ascend",
+            "--disable-cuda-graph",
+            "--mem-fraction-static",
+            0.95,
+            "--tp",
+            2,
+        ]
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            api_key=cls.api_key,
+            other_args=cls.other_args,
+        )
+        cls.additional_chat_kwargs = {}
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_chat_completion_with_reasoning(self):
+        # Test non-streaming with "enable_thinking": True, reasoning_content should not be empty
+        client = requests.post(
+            f"{self.base_url}/v1/chat/completions",
+            headers={"Authorization": f"Bearer {self.api_key}"},
+            json={
+                "model": self.model,
+                "messages": [{"role": "user", "content": "Hello"}],
+                "temperature": 0,
+                "separate_reasoning": True,
+                "chat_template_kwargs": {"enable_thinking": True},
+                **self.additional_chat_kwargs,
+            },
+        )
+
+        self.assertEqual(client.status_code, 200, f"Failed with: {client.text}")
+        data = client.json()
+
+        self.assertIn("choices", data)
+        self.assertTrue(len(data["choices"]) > 0)
+        self.assertIn("message", data["choices"][0])
+        self.assertIn("reasoning_content", data["choices"][0]["message"])
+        self.assertIsNotNone(data["choices"][0]["message"]["reasoning_content"])
+
+    def test_chat_completion_without_reasoning(self):
+        # Test non-streaming with "enable_thinking": False, reasoning_content should be empty
+        client = requests.post(
+            f"{self.base_url}/v1/chat/completions",
+            headers={"Authorization": f"Bearer {self.api_key}"},
+            json={
+                "model": self.model,
+                "messages": [{"role": "user", "content": "Hello"}],
+                "temperature": 0,
+                "separate_reasoning": True,
+                "chat_template_kwargs": {"enable_thinking": False},
+                **self.additional_chat_kwargs,
+            },
+        )
+
+        self.assertEqual(client.status_code, 200, f"Failed with: {client.text}")
+        data = client.json()
+
+        self.assertIn("choices", data)
+        self.assertTrue(len(data["choices"]) > 0)
+        self.assertIn("message", data["choices"][0])
+
+        if "reasoning_content" in data["choices"][0]["message"]:
+            self.assertIsNone(data["choices"][0]["message"]["reasoning_content"])
+
+    def test_stream_chat_completion_with_reasoning(self):
+        # Test streaming with "enable_thinking": True, reasoning_content should not be empty
+        response = requests.post(
+            f"{self.base_url}/v1/chat/completions",
+            headers={"Authorization": f"Bearer {self.api_key}"},
+            json={
+                "model": self.model,
+                "messages": [{"role": "user", "content": "Hello"}],
+                "temperature": 0,
+                "separate_reasoning": True,
+                "stream": True,
+                "chat_template_kwargs": {"enable_thinking": True},
+                **self.additional_chat_kwargs,
+            },
+            stream=True,
+        )
+
+        self.assertEqual(response.status_code, 200, f"Failed with: {response.text}")
+
+        has_reasoning = False
+        has_content = False
+
+        print("\n=== Stream With Reasoning ===")
+        for line in response.iter_lines():
+            if line:
+                line = line.decode("utf-8")
+                if line.startswith("data:") and not line.startswith("data: [DONE]"):
+                    data = json.loads(line[6:])
+                    if "choices" in data and len(data["choices"]) > 0:
+                        delta = data["choices"][0].get("delta", {})
+
+                        if "reasoning_content" in delta and delta["reasoning_content"]:
+                            has_reasoning = True
+
+                        if "content" in delta and delta["content"]:
+                            has_content = True
+
+        self.assertTrue(
+            has_reasoning,
+            "The reasoning content is not included in the stream response",
+        )
+        self.assertTrue(
+            has_content, "The stream response does not contain normal content"
+        )
+
+    def test_stream_chat_completion_without_reasoning(self):
+        # Test streaming with "enable_thinking": False, reasoning_content should  be empty
+        response = requests.post(
+            f"{self.base_url}/v1/chat/completions",
+            headers={"Authorization": f"Bearer {self.api_key}"},
+            json={
+                "model": self.model,
+                "messages": [{"role": "user", "content": "Hello"}],
+                "temperature": 0,
+                "separate_reasoning": True,
+                "stream": True,
+                "chat_template_kwargs": {"enable_thinking": False},
+                **self.additional_chat_kwargs,
+            },
+            stream=True,
+        )
+
+        self.assertEqual(response.status_code, 200, f"Failed with: {response.text}")
+
+        has_reasoning = False
+        has_content = False
+
+        print("\n=== Stream Without Reasoning ===")
+        for line in response.iter_lines():
+            if line:
+                line = line.decode("utf-8")
+                if line.startswith("data:") and not line.startswith("data: [DONE]"):
+                    data = json.loads(line[6:])
+                    if "choices" in data and len(data["choices"]) > 0:
+                        delta = data["choices"][0].get("delta", {})
+
+                        if "reasoning_content" in delta and delta["reasoning_content"]:
+                            has_reasoning = True
+
+                        if "content" in delta and delta["content"]:
+                            has_content = True
+
+        self.assertFalse(
+            has_reasoning,
+            "The reasoning content should not be included in the stream response",
+        )
+        self.assertTrue(
+            has_content, "The stream response does not contain normal content"
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()