diff --git a/paddle/fluid/framework/parallel_executor.cc b/paddle/fluid/framework/parallel_executor.cc index c3a90149a1f0daa6c5240ba2e65da41ad6f135de..67e7078fbc769dc271cf6f91bbccf59e5777eee3 100644 --- a/paddle/fluid/framework/parallel_executor.cc +++ b/paddle/fluid/framework/parallel_executor.cc @@ -155,7 +155,7 @@ struct ScaleLossGradOpHandle : public OpHandle { VLOG(3) << "2"; memory::Copy(boost::get(place_), tmp, platform::CPUPlace(), &coeff_, sizeof(float), stream); - PADDLE_ENFORCE(cudaGetLastError()); + PADDLE_ENFORCE(cudaDeviceSynchronize()); VLOG(3) << "3"; PADDLE_ENFORCE(cudaEventRecord(ev_, stream)); VLOG(3) << "4";