[Main][Bugfix]Avoid using the fusion operator in the MOE model (#3834)

### What this PR does / why we need it?
The current MatmulReduceScatter operator experiences performance
degradation in small-shape scenarios, so it determines whether to use
this operator by judging the size of the shape.

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.11.0rc3
- vLLM main:
vllm-project/vllm@releases/v0.11.1

---------

Signed-off-by: ZYang6263 <[email protected]>

Author: ZYang6263

vllm_ascend/ascend_forward_context.py

@@ -113,13 +113,16 @@ def set_ascend_forward_context(
         # Currently, it is an empirical value. In normal scenarios, if the concurrency exceeds this threshold,
         # the performance benefits can be maximized. Conversely, if the concurrency is below the threshold,
         # the performance may degrade due to the switching of communication methods.
+        mmrs_fusion = True
         if is_moe_model(vllm_config):
             sp_enabled = enable_sp(vllm_config) and \
                 tp_world_size > 1 and num_tokens is not None
+            mmrs_fusion = False
         else:
             sp_enabled = enable_sp(vllm_config) and \
                 tp_world_size > 1 and \
                 num_tokens is not None and num_tokens > 1000
+        forward_context.mmrs_fusion = mmrs_fusion
 
         if sp_enabled:
             pad_size = (tp_world_size -

vllm_ascend/ops/linear_op.py

@@ -382,8 +382,10 @@ def matmul_and_reduce(self, input_parallel: torch.Tensor,
         try:
             forward_context = get_forward_context()
             sp_enabled = forward_context.sp_enabled
+            mmrs_fusion = forward_context.mmrs_fusion
         except AssertionError:
             sp_enabled = False
+            mmrs_fusion = False
 
         x = input_parallel
 
@@ -409,8 +411,9 @@ def matmul_and_reduce(self, input_parallel: torch.Tensor,
                                                    quant_per_tensor)
 
         # For unquant
-        if isinstance(self.layer.quant_method, UnquantizedLinearMethod
-                      ) and torch.version.cann.startswith("8.3"):
+        if mmrs_fusion and isinstance(
+                self.layer.quant_method, UnquantizedLinearMethod
+        ) and torch.version.cann.startswith("8.3"):
             output = torch_npu.npu_mm_reduce_scatter_base(
                 x,
                 self.layer.weight.t(),
@@ -423,10 +426,11 @@ def matmul_and_reduce(self, input_parallel: torch.Tensor,
             if bias_ is not None:
                 output.add_(bias_)
         # For w8a8 quant
-        elif (isinstance(self.layer.quant_method, AscendLinearMethod)
-              and isinstance(self.layer.quant_method.quant_method,
-                             AscendW8A8LinearMethod)
-              ) and torch.version.cann.startswith("8.3"):
+        elif mmrs_fusion and (
+                isinstance(self.layer.quant_method, AscendLinearMethod)
+                and isinstance(self.layer.quant_method.quant_method,
+                               AscendW8A8LinearMethod)
+        ) and torch.version.cann.startswith("8.3"):
             if x.dtype != torch.int8:
                 x_quant = quant_per_tensor(
                     x, self.layer.aclnn_input_scale_reciprocal,