[Config] Retire get_global_server_args, and clear the deprecated flags that have a replacement (#38375)

test/manual/nightly/test_vlms_piecewise_cuda_graph.py

@@ -135,11 +135,11 @@ class TestVLMPiecewiseCudaGraph(CustomTestCase):
                 api_key=self.api_key,
                 other_args=[
                     "--trust-remote-code",
-                    "--piecewise-cuda-graph-max-tokens",
+                    "--cuda-graph-max-bs-prefill",
                     "8192",
-                    "--enforce-piecewise-cuda-graph",
+                    "--cuda-graph-backend-prefill=tc_piecewise",
                     "--tp=8",
-                    "--piecewise-cuda-graph-compiler=eager",
+                    "--cuda-graph-tc-compiler=eager",
                     "--disable-radix-cache",
                     "--log-level",
                     log_level,

test/manual/nightly/test_vlms_vit_cuda_graph.py

@@ -140,14 +140,14 @@ class TestVLMViTCudaGraph(CustomTestCase):
                 other_args=[
                     "--mm-attention-backend",
                     "fa3",
-                    "--enforce-piecewise-cuda-graph",
-                    "--piecewise-cuda-graph-max-tokens",
+                    "--cuda-graph-backend-prefill=tc_piecewise",
+                    "--cuda-graph-max-bs-prefill",
                     "8192",
                     "--chunked-prefill-size",
                     "8192",
                     "--disable-radix-cache",
                     "--disable-overlap-schedule",
-                    "--piecewise-cuda-graph-compiler",
+                    "--cuda-graph-tc-compiler",
                     "eager",
                 ],
                 env=process_env,