修改train_model_grpo.py文件-开启vLLM 观察能否解决损失率值0并且无变化问题

src/train_model_grpo_v1.py

@@ -90,7 +90,7 @@ class ModelTrainer:
         )
 
         torch.cuda.empty_cache()
-
+        print("self.config.learning_rate=",self.config.learning_rate)
         training_args = GRPOConfig(
             use_vllm=self.config.use_vllm,
             learning_rate=self.config.learning_rate,