Add Support for Page Size greater than 1 for Flashinfer MLA Backend (#8593)

Signed-off-by: Pavani Majety <pmajety@nvidia.com>

test/srt/test_mla_flashinfer.py

@@ -120,5 +120,49 @@ class TestFlashinferMLAMTP(CustomTestCase):
         self.assertGreater(avg_spec_accept_length, 2.5)
 
 
+class TestFlashinferMLAPageSize16(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = "deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct"
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = ["--trust-remote-code"]
+        if torch.cuda.is_available() and torch.version.cuda:
+            other_args.extend(
+                [
+                    "--cuda-graph-max-bs",
+                    "4",
+                    "--attention-backend",
+                    "flashinfer",
+                    "--page-size",
+                    "16",
+                ]
+            )
+        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_gsm8k(self):
+        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(metrics)
+
+        self.assertGreater(metrics["accuracy"], 0.615)
+
+
 if __name__ == "__main__":
     unittest.main()