modify code according to review

Author: wangxinxin08

paddle/fluid/inference/tensorrt/plugin/deformable_conv_op_plugin.cu

@@ -210,8 +210,8 @@ int DeformableConvPlugin::enqueue(int batch_size, const void* const* inputs,
 #if CUDA_ARCH_FP16_SUPPORTED(__CUDA_ARCH__)
  enqueue_impl<half>(batch_size, inputs, outputs, workspace, stream);
 #else
- VLOG(1) << "Current CUDA arch dose not support fp16, so deformable conv "
-  "trt plugin will do nothing while data type is fp16";
+ PADDLE_THROW(platform::errors::InvalidArgument(
+ "Current CUDA arch dose not support fp16. Please use fp32 instead."));
 #endif
  } else {
  PADDLE_THROW(platform::errors::InvalidArgument(