clean redundant code in previous PR (#12957)

python/sglang/srt/models/glm4_moe.py

@@ -61,7 +61,6 @@ from sglang.srt.layers.moe import (
 )
 from sglang.srt.layers.moe.ep_moe.layer import get_moe_impl_class
 from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoE
-from sglang.srt.layers.moe.kt_ep_wrapper import KTEPWrapperMethod
 from sglang.srt.layers.moe.topk import TopK
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
 from sglang.srt.layers.quantization.fp8_kernel import is_fp8_fnuz
@@ -455,42 +454,6 @@ class Glm4MoeSparseMoeBlock(nn.Module):
         else:
             return self.forward_deepep(hidden_states, forward_batch)
 
-    def forward_normal_dual_stream(
-        self,
-        hidden_states: torch.Tensor,
-        should_allreduce_fusion: bool = False,
-        use_reduce_scatter: bool = False,
-    ) -> torch.Tensor:
-
-        current_stream = torch.cuda.current_stream()
-        self.alt_stream.wait_stream(current_stream)
-        shared_output = self._forward_shared_experts(hidden_states)
-
-        with torch.cuda.stream(self.alt_stream):
-            # router_logits: (num_tokens, n_experts)
-            router_logits = self.gate(hidden_states)
-            topk_output = self.topk(hidden_states, router_logits)
-            final_hidden_states = self.experts(hidden_states, topk_output)
-            if not _is_cuda or isinstance(self.experts.quant_method, KTEPWrapperMethod):
-                final_hidden_states *= self.routed_scaling_factor
-
-        current_stream.wait_stream(self.alt_stream)
-        with use_symmetric_memory(
-            parallel_state.get_tp_group(), disabled=not is_allocation_symmetric()
-        ):
-            final_hidden_states_out = torch.empty_like(final_hidden_states)
-
-        torch.add(final_hidden_states, shared_output, out=final_hidden_states_out)
-        final_hidden_states = final_hidden_states_out
-        if (
-            self.tp_size > 1
-            and not should_allreduce_fusion
-            and not use_reduce_scatter
-            and not should_use_flashinfer_cutlass_moe_fp4_allgather()
-        ):
-            final_hidden_states = tensor_model_parallel_all_reduce(final_hidden_states)
-        return final_hidden_states
-
     def forward_normal(
         self,
         hidden_states: torch.Tensor,