diff --git a/paddle/operators/math/detail/lstm_gpu_kernel.h b/paddle/operators/math/detail/lstm_gpu_kernel.h
index e07655eaac2bc2731a3e52cdd7405828bbf9ac3c..1781460c351eedebdc766284f549db061c8874b2 100644
--- a/paddle/operators/math/detail/lstm_gpu_kernel.h
+++ b/paddle/operators/math/detail/lstm_gpu_kernel.h
@@ -248,7 +248,7 @@ void gpu_lstm_backward(const platform::DeviceContext& context, Op op,
   cudaStreamSynchronize(stream);
   // TODO(qingqing): Add cuda error check for each kernel.
   cudaError_t err = cudaGetLastError();
-  PADDLE_ENFORCE_EQ(err, cudaGetErrorString(err));
+  PADDLE_ENFORCE(err, cudaGetErrorString(err));
 }
 
 }  // namespace detail