[Bugfix] EPLB verfiy

Author: shenchuxiaofugui

vllm_ascend/ops/common_fused_moe.py

@@ -172,6 +172,7 @@ def __init__(self, *args, **kwargs):
         self.expert_map_path = ascend_config.expert_map_path
         self.global_redundant_expert_num = ascend_config.init_redundancy_expert
         self.global_num_experts = num_experts + self.global_redundant_expert_num
+        init_eplb_enable = False
         if self.custom_routing_function is None and self.e_score_correction_bias is not None:
             vllm_config = get_current_vllm_config()
             self.e_score_correction_bias.data = self.e_score_correction_bias.data.to(
@@ -191,6 +192,7 @@ def __init__(self, *args, **kwargs):
                         self.moe_instance_id, self.ep_rank))
                 self.log2phy = self.expert_load_balancer.get_rank_log2phy_map(
                     self.moe_instance_id, self.ep_rank).npu()
+                init_eplb_enable = True
             except Exception as e:
                 logger.warning(
                     f"Init expert map of mtp/eagle when using sample.{e}")
@@ -236,8 +238,7 @@ def __init__(self, *args, **kwargs):
             self.moe_load = torch.zeros(local_num_experts,
                                         dtype=torch.int64).npu()
 
-        eplb_enable = self.dynamic_eplb or (self.expert_map_path is not None)
-        if eplb_enable and (not hasattr(self.quant_method, "quant_method") or
+        if init_eplb_enable and (not hasattr(self.quant_method, "quant_method") or
                             not isinstance(self.quant_method.quant_method,
                                            AscendW8A8DynamicFusedMoEMethod)):
             raise ValueError("Eplb supports only w8a8_dynamic quantization.")