update get_expert_load return type

Author: lt

vllm_ascend/worker/model_runner_v1.py

@@ -1587,7 +1587,7 @@ def profile_run(self) -> None:
         self.encoder_cache.clear()
         gc.collect()
 
-    def do_get_expert_load(self) -> str:
+    def do_get_expert_load(self) ->  torch.Tensor:
         return self.eplb_updator.get_expert_load()
 
     def do_update_expert_load_statistical_period(self, num_expert_load_gather: int, num_iterations: int):

vllm_ascend/worker/worker_v1.py

@@ -209,7 +209,7 @@ def compile_or_warm_up_model(self) -> None:
         # the model initialization and profiling.
         set_random_seed(self.model_config.seed)
 
-    def get_expert_load(self) -> str:
+    def get_expert_load(self) -> torch.Tensor:
         """ todo 一共几个worker"""
         moe_load = self.model_runner.do_get_expert_load()
         return moe_load