未验证 提交 454f4f21 编写于 作者: Y Yiqun Liu 提交者: GitHub

Rewrite is_empty op to avoid unnecessary data transform. (#15509)

* Rewrite is_empty op to avoid unnecessary data transform.
test=develop

* Add the implementation of InferShape and InferVarType for is_empty op.
test=develop

* Rewrite is_empty op to avoid directly inherit OperatorBase.
test=develop
上级 6724be2b
......@@ -34,9 +34,8 @@ class IsEmptyOp : public framework::OperatorWithKernel {
framework::OpKernelType GetExpectedKernelType(
const framework::ExecutionContext &ctx) const override {
framework::OpKernelType kt = framework::OpKernelType(
ctx.Input<framework::LoDTensor>("X")->type(), platform::CPUPlace());
return kt;
auto *x = ctx.Input<framework::LoDTensor>("X");
return framework::OpKernelType(x->type(), x->place());
}
};
......@@ -58,7 +57,6 @@ It will just return product(tensor.ddims()) > 0;
} // namespace paddle
namespace ops = paddle::operators;
REGISTER_OPERATOR(is_empty, ops::IsEmptyOp, ops::IsEmptyOpMaker,
paddle::framework::EmptyGradOpMaker);
REGISTER_OP_CPU_KERNEL(
......
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "paddle/fluid/operators/is_empty_op.h"
#include "paddle/fluid/framework/op_registry.h"
namespace ops = paddle::operators;
REGISTER_OP_CUDA_KERNEL(
is_empty, ops::IsEmptyOpKernel<paddle::platform::CUDADeviceContext, float>,
ops::IsEmptyOpKernel<paddle::platform::CUDADeviceContext, double>,
ops::IsEmptyOpKernel<paddle::platform::CUDADeviceContext, int>,
ops::IsEmptyOpKernel<paddle::platform::CUDADeviceContext, int64_t>);
......@@ -28,6 +28,9 @@ class IsEmptyOpKernel : public framework::OpKernel<T> {
// get output
auto* output_tensor = context.Output<framework::LoDTensor>("Out");
// Note: is_empty is always executed on CPU and the output data should
// always be allocated for CPUPlace. We reigister CUDA kernel for this op to
// avoid the unnecessary data transform.
output_tensor->mutable_data<bool>(platform::CPUPlace())[0] =
framework::product(input_tensor->dims()) == 0;
}
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册