[tiny remove] remove torch_compile in parallel_state (#16865)

Signed-off-by: Oasis-Git <ayw.sirius19@gmail.com>

python/sglang/srt/distributed/parallel_state.py

@@ -224,7 +224,6 @@ class GroupCoordinator:
         use_message_queue_broadcaster: bool = False,
         group_name: Optional[str] = None,
         pynccl_use_current_stream: bool = False,
-        torch_compile: Optional[bool] = None,
         gloo_timeout: timedelta = timedelta(seconds=120 * 60),
     ):
         # Set group info
@@ -1333,7 +1332,6 @@ def init_model_parallel_group(
     use_mscclpp_allreduce: Optional[bool] = None,
     pynccl_use_current_stream: bool = True,
     use_torch_symm_mem_allreduce: Optional[bool] = None,
-    torch_compile: Optional[bool] = None,
 ) -> GroupCoordinator:
     if use_custom_allreduce is None:
         use_custom_allreduce = _ENABLE_CUSTOM_ALL_REDUCE
@@ -1355,7 +1353,6 @@ def init_model_parallel_group(
         use_message_queue_broadcaster=use_message_queue_broadcaster,
         group_name=group_name,
         pynccl_use_current_stream=pynccl_use_current_stream,
-        torch_compile=torch_compile,
     )
 
 
@@ -1526,7 +1523,6 @@ def initialize_model_parallel(
     pipeline_model_parallel_size: int = 1,
     backend: Optional[str] = None,
     duplicate_tp_group: bool = False,
-    torch_compile: Optional[bool] = None,
 ) -> None:
     """
     Initialize model parallel groups.
@@ -1583,7 +1579,6 @@ def initialize_model_parallel(
         ),
         group_name="tp",
         pynccl_use_current_stream=duplicate_tp_group,
-        torch_compile=torch_compile,
     )
 
     if duplicate_tp_group:
@@ -1600,7 +1595,6 @@ def initialize_model_parallel(
             ),
             group_name="pdmux_prefill_tp",
             pynccl_use_current_stream=True,
-            torch_compile=torch_compile,
         )
         if _TP.pynccl_comm:
             _TP.pynccl_comm.disabled = False

python/sglang/srt/model_executor/model_runner.py

@@ -760,7 +760,6 @@ class ModelRunner(ModelRunnerKVCacheMixin):
                 pipeline_model_parallel_size=self.pp_size,
                 expert_model_parallel_size=self.moe_ep_size,
                 duplicate_tp_group=self.server_args.enable_pdmux,
-                torch_compile=self.server_args.enable_piecewise_cuda_graph,
             )
             initialize_dp_attention(
                 server_args=self.server_args,