[Refactor] Cuda Graph Runner/Backend Refactor (#23906)

Co-authored-by: BBuf <1182563586@qq.com>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>

test/registered/sessions/test_session_control.py

@@ -45,7 +45,7 @@ class TestSessionControl(CustomTestCase):
                 "--attention-backend",
                 "triton",
                 "--disable-cuda-graph",
-                "--disable-piecewise-cuda-graph",
+                "--cuda-graph-backend-prefill=disabled",
             ],
         )
 

test/registered/sessions/test_session_latency.py

@@ -271,7 +271,7 @@ class TestSessionLatency(CustomTestCase):
                 "--enable-streaming-session",
                 "--mem-fraction-static",
                 "0.70",
-                "--disable-piecewise-cuda-graph",
+                "--cuda-graph-backend-prefill=disabled",
                 "--page-size",
                 "4",
             ],

test/registered/sessions/test_streaming_session_swa.py

@@ -21,7 +21,7 @@ SWA_MODEL = "openai/gpt-oss-20b"
 SWA_COMMON_ARGS = [
     "--mem-fraction-static",
     "0.70",
-    "--disable-piecewise-cuda-graph",
+    "--cuda-graph-backend-prefill=disabled",
 ]