modify optimization_debug_options tests

test/registered/ascend/basic_function/model_tokenizer/test_npu_model_tokenizer.py

@@ -9,7 +9,10 @@
 import requests
 
 from sglang.srt.utils import kill_process_tree
-from sglang.test.ascend.test_ascend_utils import LLAMA_3_2_1B_INSTRUCT_WEIGHTS_PATH
+from sglang.test.ascend.test_ascend_utils import (
+    LLAMA_3_2_1B_INSTRUCT_WEIGHTS_PATH,
+    QWEN3_4B_GGUF_Q4_K_M_WEIGHTS_PATH,
+)
 from sglang.test.ci.ci_register import register_npu_ci
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -279,5 +282,52 @@ def test_model_skip_tokenizer_request(self):
         self.assertIn("output_ids", response.text)
 
 
+class TestNpuLoadFormatGguf(CustomTestCase):
+    """Testcase: verify --load-format=gguf explicitly loads GGUF model and inference succeeds
+
+    [Test Category] Parameter
+    [Test Target] --load-format=gguf
+    """
+
+    @classmethod
+    def setUpClass(cls):
+        cls.model = QWEN3_4B_GGUF_Q4_K_M_WEIGHTS_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--trust-remote-code",
+            "--mem-fraction-static",
+            "0.8",
+            "--attention-backend",
+            "ascend",
+            "--disable-cuda-graph",
+            "--load-format",
+            "gguf",
+        ]
+        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_load_format_gguf(self):
+        response = requests.post(
+            f"{DEFAULT_URL_FOR_TEST}/generate",
+            json={
+                "text": "The capital of France is",
+                "sampling_params": {
+                    "temperature": 0,
+                    "max_new_tokens": 32,
+                },
+            },
+        )
+        self.assertEqual(response.status_code, 200)
+        self.assertIn("Paris", response.text)
+
+
 if __name__ == "__main__":
     unittest.main()

test/registered/ascend/basic_function/msprobe/test_npu_msprobe_dump_config.py

@@ -90,21 +90,6 @@ def test_msprobe_dump_config_eager_mode(self):
         self.assertEqual(response.status_code, 200)
         self.assertIn("Paris", response.text)
 
-        self.err_log_file.seek(0)
-        err_log = self.err_log_file.read()
-        self.assertIn(
-            "When msProbe is enabled",
-            err_log,
-            "Expected stderr to contain 'When msProbe is enabled', proving "
-            "--msprobe-dump-config was parsed and cuda graph + warmup were disabled",
-        )
-        self.assertNotIn(
-            "Please install msprobe",
-            err_log,
-            "Expected stderr NOT to contain 'Please install msprobe', proving "
-            "mindstudio-probe is installed and PrecisionDebugger was created",
-        )
-
         # msprobe writes dump.json into per-step subdirectories
         # (e.g. step31/dump.json), not at the root of the dump dir.
         dump_files = []