"cudnn operators change to cudnn kernel" (#6660)
* "unified operators" * "add CUDNN register" * "add use cudnn attribute" * "add attribute" * "test conv tranpose op" * "remove duplicated attr" * "fix op test" * "add attribute to set cudnn" * "add more log" * "need layout op register support" * "add more log" * "change GetExpectedKernelType " * "fix Get attr in conv_op" * "fix CI" * "fix tests" * "removed kernel priority fallback" * "fix CI" * "fix stack pointer bug" * "refine buggy interface" * "add const cast to save life" * "fix get_output_with_grad" * "fix op test with dataformat" * ""fix pooling * "fix pooling test" * "fix CI" * "fix with_gpu error" * "add transform needed functional check" * "fix unpack list error" * "comment out parallel.do temporary" * "fix CI" * "fix compile doc error" * "make threshold larger"add_depthwiseConv_op_gpu
dzhwinter
7 years ago
committed by
GitHub
parent
61881397d5
commit
5ad1aef051
@ -1,74 +0,0 @@
|
||||
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
|
||||
|
||||
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/operators/conv_op.h"
|
||||
|
||||
namespace paddle {
|
||||
namespace operators {
|
||||
|
||||
class CudnnConv2DOpMaker : public Conv2DOpMaker {
|
||||
public:
|
||||
CudnnConv2DOpMaker(OpProto* proto, OpAttrChecker* op_checker)
|
||||
: Conv2DOpMaker(proto, op_checker) {
|
||||
AddAttr<int>("workspace_size_MB",
|
||||
"workspace size for cudnn, in MB, "
|
||||
"workspace is a section of GPU memory which will be "
|
||||
"allocated/freed each time the operator runs, larger "
|
||||
"workspace size can increase performance but also requires "
|
||||
"better hardware. This size should be chosen carefully.")
|
||||
.SetDefault(4096);
|
||||
}
|
||||
};
|
||||
|
||||
class CudnnConv3DOpMaker : public Conv3DOpMaker {
|
||||
public:
|
||||
CudnnConv3DOpMaker(OpProto* proto, OpAttrChecker* op_checker)
|
||||
: Conv3DOpMaker(proto, op_checker) {
|
||||
AddAttr<int>("workspace_size_MB",
|
||||
"workspace size for cudnn, in MB, "
|
||||
"workspace is a section of GPU memory which will be "
|
||||
"allocated/freed each time the operator runs, larger "
|
||||
"workspace size can increase performance but also requires "
|
||||
"better hardware. This size should be chosen carefully.")
|
||||
.SetDefault(4096);
|
||||
}
|
||||
};
|
||||
|
||||
} // namespace operators
|
||||
} // namespace paddle
|
||||
|
||||
namespace ops = paddle::operators;
|
||||
REGISTER_OP(conv2d_cudnn, ops::ConvOp, ops::CudnnConv2DOpMaker,
|
||||
conv2d_cudnn_grad, ops::ConvOpGrad);
|
||||
|
||||
REGISTER_OP(conv3d_cudnn, ops::ConvOp, ops::CudnnConv3DOpMaker,
|
||||
conv3d_cudnn_grad, ops::ConvOpGrad);
|
||||
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
conv2d_cudnn,
|
||||
ops::GemmConvKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::GemmConvKernel<paddle::platform::CPUDeviceContext, double>);
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
conv2d_cudnn_grad,
|
||||
ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, double>);
|
||||
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
conv3d_cudnn,
|
||||
ops::GemmConvKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::GemmConvKernel<paddle::platform::CPUDeviceContext, double>);
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
conv3d_cudnn_grad,
|
||||
ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::GemmConvGradKernel<paddle::platform::CPUDeviceContext, double>);
|
||||
@ -1,78 +0,0 @@
|
||||
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
|
||||
|
||||
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/operators/conv_transpose_op.h"
|
||||
|
||||
namespace paddle {
|
||||
namespace operators {
|
||||
|
||||
class CudnnConv2DTransposeOpMaker : public Conv2DTransposeOpMaker {
|
||||
public:
|
||||
CudnnConv2DTransposeOpMaker(OpProto* proto, OpAttrChecker* op_checker)
|
||||
: Conv2DTransposeOpMaker(proto, op_checker) {
|
||||
AddAttr<int>("workspace_size_MB",
|
||||
"workspace size for cudnn, in MB, "
|
||||
"workspace is a section of GPU memory which will be "
|
||||
"allocated/freed each time the operator runs, larger "
|
||||
"workspace size can increase performance but also requires "
|
||||
"better hardward. This size should be carefully setted.")
|
||||
.SetDefault(4096);
|
||||
}
|
||||
};
|
||||
|
||||
class CudnnConv3DTransposeOpMaker : public Conv3DTransposeOpMaker {
|
||||
public:
|
||||
CudnnConv3DTransposeOpMaker(OpProto* proto, OpAttrChecker* op_checker)
|
||||
: Conv3DTransposeOpMaker(proto, op_checker) {
|
||||
AddAttr<int>("workspace_size_MB",
|
||||
"workspace size for cudnn, in MB, "
|
||||
"workspace is a section of GPU memory which will be "
|
||||
"allocated/freed each time the operator runs, larger "
|
||||
"workspace size can increase performance but also requires "
|
||||
"better hardward. This size should be carefully setted.")
|
||||
.SetDefault(4096);
|
||||
}
|
||||
};
|
||||
|
||||
} // namespace operators
|
||||
} // namespace paddle
|
||||
|
||||
namespace ops = paddle::operators;
|
||||
REGISTER_OP(conv2d_transpose_cudnn, ops::ConvTransposeOp,
|
||||
ops::CudnnConv2DTransposeOpMaker, conv2d_transpose_cudnn_grad,
|
||||
ops::ConvTransposeOpGrad);
|
||||
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
conv2d_transpose_cudnn,
|
||||
ops::GemmConvTransposeKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::GemmConvTransposeKernel<paddle::platform::CPUDeviceContext, double>);
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
conv2d_transpose_cudnn_grad,
|
||||
ops::GemmConvTransposeGradKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::GemmConvTransposeGradKernel<paddle::platform::CPUDeviceContext,
|
||||
double>);
|
||||
|
||||
REGISTER_OP(conv3d_transpose_cudnn, ops::ConvTransposeOp,
|
||||
ops::CudnnConv3DTransposeOpMaker, conv3d_transpose_cudnn_grad,
|
||||
ops::ConvTransposeOpGrad);
|
||||
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
conv3d_transpose_cudnn,
|
||||
ops::GemmConvTransposeKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::GemmConvTransposeKernel<paddle::platform::CPUDeviceContext, double>);
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
conv3d_transpose_cudnn_grad,
|
||||
ops::GemmConvTransposeGradKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::GemmConvTransposeGradKernel<paddle::platform::CPUDeviceContext,
|
||||
double>);
|
||||
@ -1,39 +0,0 @@
|
||||
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
|
||||
|
||||
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/operators/pool_cudnn_op.h"
|
||||
|
||||
namespace ops = paddle::operators;
|
||||
|
||||
REGISTER_OP(pool2d_cudnn, ops::PoolOp, ops::Pool2dOpMaker, pool2d_cudnn_grad,
|
||||
ops::PoolOpGrad);
|
||||
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
pool2d_cudnn, ops::PoolKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::PoolKernel<paddle::platform::CPUDeviceContext, double>);
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
pool2d_cudnn_grad,
|
||||
ops::PoolGradKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::PoolGradKernel<paddle::platform::CPUDeviceContext, double>)
|
||||
|
||||
REGISTER_OP(pool3d_cudnn, ops::PoolOp, ops::Pool3dOpMaker, pool3d_cudnn_grad,
|
||||
ops::PoolOpGrad);
|
||||
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
pool3d_cudnn, ops::PoolKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::PoolKernel<paddle::platform::CPUDeviceContext, double>);
|
||||
REGISTER_OP_CPU_KERNEL(
|
||||
pool3d_cudnn_grad,
|
||||
ops::PoolGradKernel<paddle::platform::CPUDeviceContext, float>,
|
||||
ops::PoolGradKernel<paddle::platform::CPUDeviceContext, double>)
|
||||
@ -1,19 +0,0 @@
|
||||
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
|
||||
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. */
|
||||
|
||||
#pragma once
|
||||
|
||||
#include "paddle/framework/op_registry.h"
|
||||
#include "paddle/operators/pool_op.h"
|
||||
|
||||
namespace paddle {
|
||||
namespace operators {} // namespace operators
|
||||
} // namespace paddle
|
||||
Some files were not shown because too many files have changed in this diff Show More
Loading…
Reference in new issue