[feature] Add layerwise NVTX support (#11870)

python/sglang/srt/server_args.py

@@ -479,6 +479,7 @@ class ServerArgs:
     disable_cuda_graph_padding: bool = False
     enable_profile_cuda_graph: bool = False
     enable_cudagraph_gc: bool = False
+    enable_layerwise_nvtx_marker: bool = False
     enable_nccl_nvls: bool = False
     enable_symm_mem: bool = False
     disable_flashinfer_cutlass_moe_fp4_allgather: bool = False
@@ -3240,6 +3241,11 @@ class ServerArgs:
             action="store_true",
             help="Enable garbage collection during CUDA graph capture. If disabled (default), GC is frozen during capture to speed up the process.",
         )
+        parser.add_argument(
+            "--enable-layerwise-nvtx-marker",
+            action="store_true",
+            help="Enable layerwise NVTX profiling annotations for the model.",
+        )
         parser.add_argument(
             "--enable-nccl-nvls",
             action="store_true",