colin
/
TIM-VX
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add map for UnMaxpool2d (#83) 

Signed-off-by: zhao.xia <zhao.xia@verisilicon.com>
This commit is contained in:
Nightingale 2021-05-28 17:09:26 +08:00 committed by GitHub
parent 18a928ee69
commit 9c60671031
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 222 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
include/tim/vx/ops/unmaxpool2d.h Normal file
View File

@ -0,0 +1,59 @@
/****************************************************************************
*
* Copyright (c) 2021 Vivante Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*
*****************************************************************************/
#ifndef TIM_VX_OPS_UNMAXPOOL2D_H_
#define TIM_VX_OPS_UNMAXPOOL2D_H_
#include <array>
#include "tim/vx/operation.h"
#include "tim/vx/types.h"
namespace tim {
namespace vx {
namespace ops {
/**
* ## UnMaxpool2d
*
* Performs an 2-D Max pooling operation upsample
*
* - stride : stride along each spatial axis.
* - ksize : filter size.
*/
class UnMaxpool2d : public Operation {
public:
UnMaxpool2d(Graph* graph, const std::array<uint32_t, 2>& ksize,
const std::array<uint32_t, 2>& stride, DataLayout layout = DataLayout::WHCN);
protected:
const std::array<uint32_t, 2> ksize_;
const std::array<uint32_t, 2> stride_;
};
} // namespace ops
} // namespace vx
} // namespace tim
#endif /* TIM_VX_OPS_UNMAXPOOL2D_H_ */

2
src/tim/vx/ops/README.md
View File

@ -18,7 +18,7 @@ DeConv2d|DECONVOLUTION|Mapped|[tf.nn.conv2d_transpose](https://tensorflow.google
Reshape|RESHAPE|Mapped|[tf.reshape](https://tensorflow.google.cn/api_docs/python/tf/reshape)
Transpose|PERMUTE|Mapped|[tf.transpose](https://tensorflow.google.cn/api_docs/python/tf/transpose)
Prelu|PRELU|Mapped|[tf.keras.layers.PReLU](https://tensorflow.google.cn/api_docs/python/tf/keras/layers/PReLU)
||UPSAMPLE|Unmapped
UnMaxpool2d|UPSAMPLE|Mapped| Recover pixel from the outputs of MaxpoolWithArgmax.
Relu|RELU|Mapped|[tf.nn.relu](https://tensorflow.google.cn/api_docs/python/tf/nn/relu)
||RELUN|Deprecated|[tf.keras.layers.ReLU(max_value=N)](https://tensorflow.google.cn/api_docs/python/tf/keras/layers/ReLU)
Reorg|REORG|Mapped|[darknet.reorg](https://github.com/pjreddie/darknet/blob/master/src/reorg_layer.c)

46
src/tim/vx/ops/unmaxpool2d.cc Normal file
View File

@ -0,0 +1,46 @@
/****************************************************************************
*
* Copyright (c) 2021 Vivante Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*
*****************************************************************************/
#include "tim/vx/ops/unmaxpool2d.h"
#include "operation_private.h"
#include "type_utils.h"
#include "vsi_nn_pub.h"
namespace tim {
namespace vx {
namespace ops {
UnMaxpool2d::UnMaxpool2d(Graph* graph, const std::array<uint32_t, 2>& ksize,
const std::array<uint32_t, 2>& stride, DataLayout layout)
: Operation(graph, VSI_NN_OP_UPSAMPLE, 2, 1, layout),
ksize_(ksize), stride_(stride) {
this->impl()->node()->nn_param.upsample.scale[0] = stride_[0];
this->impl()->node()->nn_param.upsample.scale[1] = stride_[1];
this->impl()->node()->nn_param.upsample.size[0] = ksize_[0];
this->impl()->node()->nn_param.upsample.size[1] = ksize_[1];
}
} // namespace ops
} // namespace vx
} // namespace tim

116
src/tim/vx/ops/unmaxpool2d_test.cc Normal file
View File

@ -0,0 +1,116 @@
/****************************************************************************
*
* Copyright (c) 2021 Vivante Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*
*****************************************************************************/
#include "tim/vx/context.h"
#include "tim/vx/graph.h"
#include "tim/vx/ops/unmaxpool2d.h"
#include "gtest/gtest.h"
TEST(UnMaxpool2d, shape_2_2_1_fp32_kernel_2_stride_2) {
auto ctx = tim::vx::Context::Create();
auto graph = ctx->CreateGraph();
tim::vx::ShapeType in_shape({2, 2, 1});
tim::vx::ShapeType out_shape({3, 3, 1});
tim::vx::TensorSpec values_spec(tim::vx::DataType::FLOAT32,
in_shape, tim::vx::TensorAttribute::INPUT);
tim::vx::TensorSpec indices_spec(tim::vx::DataType::UINT8,
in_shape, tim::vx::TensorAttribute::INPUT);
tim::vx::TensorSpec output_spec(tim::vx::DataType::FLOAT32,
out_shape, tim::vx::TensorAttribute::OUTPUT);
auto values_tensor = graph->CreateTensor(values_spec);
auto indices_tensor = graph->CreateTensor(indices_spec);
auto output_tensor = graph->CreateTensor(output_spec);
std::vector<float> values = {
5, 6,
8, 9 };
std::vector<uint8_t> indices = {
3, 2,
1, 0 };
std::vector<float> golden = {
0, 0, 0,
0, 5, 6,
0, 8, 9 };
EXPECT_TRUE(values_tensor->CopyDataToTensor(values.data(), values.size()*4));
EXPECT_TRUE(indices_tensor->CopyDataToTensor(indices.data(), indices.size()*4));
std::array<uint32_t, 2> ksize = {2, 2};
std::array<uint32_t, 2> stride = {2, 2};
auto op = graph->CreateOperation<tim::vx::ops::UnMaxpool2d>(ksize, stride);
(*op).BindInputs({values_tensor, indices_tensor}).BindOutputs({output_tensor});
EXPECT_TRUE(graph->Compile());
EXPECT_TRUE(graph->Run());
std::vector<float> output(golden.size());
EXPECT_TRUE(output_tensor->CopyDataFromTensor(output.data()));
EXPECT_EQ(golden, output);
}
TEST(UnMaxpool2d, shape_2_2_1_uint8_kernel_2_stride_2) {
auto ctx = tim::vx::Context::Create();
auto graph = ctx->CreateGraph();
tim::vx::ShapeType in_shape({2, 2, 1});
tim::vx::ShapeType out_shape({4, 4, 1});
tim::vx::Quantization io_quant(tim::vx::QuantType::ASYMMETRIC, 1, 0);
tim::vx::TensorSpec values_spec(tim::vx::DataType::UINT8,
in_shape, tim::vx::TensorAttribute::INPUT, io_quant);
tim::vx::TensorSpec indices_spec(tim::vx::DataType::UINT8,
in_shape, tim::vx::TensorAttribute::INPUT);
tim::vx::TensorSpec output_spec(tim::vx::DataType::UINT8,
out_shape, tim::vx::TensorAttribute::OUTPUT, io_quant);
auto values_tensor = graph->CreateTensor(values_spec);
auto indices_tensor = graph->CreateTensor(indices_spec);
auto output_tensor = graph->CreateTensor(output_spec);
std::vector<uint8_t> values = {
5, 6,
11, 12};
std::vector<uint8_t> indices = {
3, 2,
3, 2};
std::vector<uint8_t> golden = {
0, 0, 0, 0,
0, 5, 6, 0,
0, 0, 0, 0,
0, 11, 12, 0 };
EXPECT_TRUE(values_tensor->CopyDataToTensor(values.data(), values.size()));
EXPECT_TRUE(indices_tensor->CopyDataToTensor(indices.data(), indices.size()));
std::array<uint32_t, 2> ksize = {2, 2};
std::array<uint32_t, 2> stride = {2, 2};
auto op = graph->CreateOperation<tim::vx::ops::UnMaxpool2d>(ksize, stride);
(*op).BindInputs({values_tensor, indices_tensor}).BindOutputs({output_tensor});
EXPECT_TRUE(graph->Compile());
EXPECT_TRUE(graph->Run());
std::vector<uint8_t> output(golden.size());
EXPECT_TRUE(output_tensor->CopyDataFromTensor(output.data()));
EXPECT_EQ(golden, output);
}