[MoE][Refactor] Make select_experts a non-static method (#29067)

Signed-off-by: Bill Nell <bnell@redhat.com>

vllm/model_executor/layers/quantization/bitsandbytes.py

@@ -495,7 +495,7 @@ class BitsAndBytesMoEMethod(FusedMoEMethodBase):
 
     def apply(
         self,
-        layer: torch.nn.Module,
+        layer: FusedMoE,
         x: torch.Tensor,
         router_logits: torch.Tensor,
         top_k: int,
@@ -518,25 +518,11 @@ class BitsAndBytesMoEMethod(FusedMoEMethodBase):
     ) -> torch.Tensor | tuple[torch.Tensor, torch.Tensor]:
         from vllm.model_executor.layers.fused_moe import fused_experts
 
-        if enable_eplb:
-            raise NotImplementedError(
-                "EPLB not supported for `BitsAndBytesMoEMethod` yet."
-            )
-
-        topk_weights, topk_ids, _ = FusedMoE.select_experts(
+        topk_weights, topk_ids, _ = layer.select_experts(
             hidden_states=x,
             router_logits=router_logits,
-            use_grouped_topk=use_grouped_topk,
-            top_k=top_k,
-            renormalize=renormalize,
-            topk_group=topk_group,
-            num_expert_group=num_expert_group,
-            custom_routing_function=custom_routing_function,
-            scoring_func=scoring_func,
-            routed_scaling_factor=routed_scaling_factor,
-            e_score_correction_bias=e_score_correction_bias,
-            indices_type=self.topk_indices_dtype,
         )
+        # TODO(bnell): Do these need to be called on the hot path?
         if self.quant_config.load_in_8bit:
             w13, w2 = self._apply_8bit_dequant(layer)
         else: