diff --git a/paddle/fluid/inference/tensorrt/engine.h b/paddle/fluid/inference/tensorrt/engine.h
index c1fa94ef423f9a088ba4ae6ac6170731f6678785..e80f2e6ce56e251df57c67e4e189cb1dd80b4e21 100644
--- a/paddle/fluid/inference/tensorrt/engine.h
+++ b/paddle/fluid/inference/tensorrt/engine.h
@@ -326,6 +326,7 @@ class TensorRTEngine {
     std::unique_lock<std::mutex> lock(mutex_);
     infer_context_[predictor_id_per_thread].reset(nullptr);
     infer_context_.erase(predictor_id_per_thread);
+    cur_profile_num_ = 0;
   }
 
   nvinfer1::IHostMemory* Serialize() {