[Bug Fix] Fix naive_block_assignment always defaulting to False due to arg misalignment (#33848)

Signed-off-by: Runkai Tao <rt572@physics.rutgers.edu>

vllm/lora/layers/fused_moe.py

@@ -219,7 +219,7 @@ class FusedMoEWithLoRA(BaseLayerWithLoRA):
                     self.max_loras,
                     self.adapter_enabled,
                     expert_map,
-                    naive_block_assignment,
+                    naive_block_assignment=naive_block_assignment,
                 )
 
                 moe_state_dict["sorted_token_ids_lora"] = sorted_token_ids_lora

vllm/lora/punica_wrapper/punica_base.py

@@ -458,6 +458,7 @@ class PunicaWrapperBase(PunicaWrapperABC):
         adapter_enabled: torch.Tensor,
         expert_map: torch.Tensor | None = None,
         pad_sorted_ids: bool = False,
+        naive_block_assignment: bool = False,
     ) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
         """
         Aligns tokens and experts into block-sized chunks for LoRA-based