[6/N] MoE Refactor: Cleanup MoE-related configs (#8849)

python/sglang/srt/layers/communicator.py

@@ -17,7 +17,7 @@ from enum import Enum, auto
 from functools import partial
 from typing import Dict, Optional
 
-import torch.distributed
+import torch
 
 from sglang.srt.distributed import (
     get_tensor_model_parallel_world_size,
@@ -35,6 +35,7 @@ from sglang.srt.layers.dp_attention import (
     get_global_dp_buffer,
     get_local_dp_buffer,
 )
+from sglang.srt.layers.moe import get_moe_a2a_backend
 from sglang.srt.layers.utils import is_sm100_supported
 from sglang.srt.managers.schedule_batch import global_server_args_dict
 from sglang.srt.model_executor.forward_batch_info import ForwardBatch
@@ -111,7 +112,7 @@ class LayerScatterModes:
         if context.is_layer_sparse:
             return (
                 ScatterMode.SCATTERED
-                if not global_server_args_dict["moe_a2a_backend"].is_standard()
+                if not get_moe_a2a_backend().is_none()
                 else ScatterMode.FULL
             )
         else: