fixes post rebase

Signed-off-by: Patryk Saffer <[email protected]>

Author: Patryk999

vllm/model_executor/layers/fused_moe/layer.py

@@ -1627,12 +1627,11 @@ def select_experts(
         if self.enable_eplb:
             topk_ids = eplb_map_to_physical_and_record(
                 topk_ids=topk_ids,
-                expert_load_view=expert_load_view,
-                logical_to_physical_map=logical_to_physical_map,
-                eplb_static=eplb_static,
-                logical_replica_count=logical_replica_count,
-                indices_type=indices_type, # TODO
                 expert_load_view=self.expert_load_view,
+                logical_to_physical_map=self.logical_to_physical_map,
+                logical_replica_count=self.logical_replica_count,
+                eplb_static=self.eplb_static,
+                indices_type=indices_type
             )
 
         if (indices_type is not None) and topk_ids.dtype != indices_type:

vllm/v1/worker/gpu_model_runner.py

@@ -3613,7 +3613,7 @@ def load_model(self, eep_scale_up: bool = False) -> None:
             if self.parallel_config.eplb_config.load_path is not None:
                 self.eplb_state.rearrange(self.model, load_initial_load_window=True)
                 if self.parallel_config.eplb_config.static:
-                    self.eplb_state = None # TODO FIX
+                    self.eplb_state = None
             if self.eplb_state.is_async:
                 self.eplb_state.start_async_loop(rank_mapping=rank_mapping)