fix compile

Author: Yang Yang

paddle/operators/fill_constant_batch_size_like_op.cc

@@ -78,7 +78,5 @@ REGISTER_OP_WITHOUT_GRADIENT(fill_constant_batch_size_like,
  ops::FillConstantBatchSizeLikeOpMaker);
 REGISTER_OP_CPU_KERNEL(
  fill_constant_batch_size_like,
- ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::CPUPlace, float>);
-REGISTER_OP_CPU_KERNEL(
- fill_constant_batch_size_like,
+ ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::CPUPlace, float>,
  ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::CPUPlace, double>);

paddle/operators/fill_constant_batch_size_like_op.cu

@@ -19,7 +19,5 @@
 namespace ops = paddle::operators;
 REGISTER_OP_GPU_KERNEL(
  fill_constant_batch_size_like,
- ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::GPUPlace, float>);
-REGISTER_OP_GPU_KERNEL(
- fill_constant_batch_size_like,
+ ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::GPUPlace, float>,
  ops::FillConstantBatchSizeLikeOpKernel<paddle::platform::GPUPlace, double>);

paddle/operators/fill_constant_batch_size_like_op.h

@@ -25,7 +25,7 @@ class FillConstantBatchSizeLikeOpKernel : public framework::OpKernel<T> {
  void Compute(const framework::ExecutionContext& ctx) const override {
  auto* out = ctx.Output<framework::Tensor>("Out");
  out->mutable_data<T>(ctx.GetPlace());
- auto value = ctx.Attr<T>("value");
+ auto value = ctx.Attr<float>("value");
 
  auto out_eigen = framework::EigenVector<T>::Flatten(*out);
  auto place = ctx.GetEigenDevice<Place>();