Mathis/CUDANet
1
0
Fork 0
mirror of https://github.com/lordmathis/CUDANet.git synced 2025-11-06 01:34:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add inception block A

Browse Source
This commit is contained in:
Matúš Námešný
2024-05-19 18:19:07 +02:00
parent d0b974dd9f
commit e8bffe22d5
1 changed files with 68 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

80
examples/inception_v3/inception_v3.cpp
View File

@@ -14,35 +14,91 @@ int main(int argc, const char *const argv[]) {
class BasicConv2d : public CUDANet::Module { class BasicConv2d : public CUDANet::Module {
public: public:
BasicConv2d( BasicConv2d(
const int inputSize, const int inputSize,
const int inputChannels, const int inputChannels,
const int outputChannels, const int outputChannels,
const int kernelSize, const int kernelSize,
const int stride, const int stride,
const int padding, const int padding,
const std::string& prefix const std::string &prefix
) )
: inputSize(inputSize), : inputSize(inputSize),
inputChannels(inputChannels), inputChannels(inputChannels),
outputChannels(outputChannels) { outputChannels(outputChannels) {
// Create the convolution layer // Create the convolution layer
CUDANet::Layers::Conv2d *conv = new CUDANet::Layers::Conv2d( CUDANet::Layers::Conv2d *conv = new CUDANet::Layers::Conv2d(
inputSize, inputChannels, kernelSize, stride, outputChannels, padding, CUDANet::Layers::ActivationType::NONE inputSize, inputChannels, kernelSize, stride, outputChannels,
padding, CUDANet::Layers::ActivationType::NONE
); );
int batchNormSize = conv->getOutputSize(); int batchNormSize = conv->getOutputSize();
CUDANet::Layers::BatchNorm *batchNorm = CUDANet::Layers::BatchNorm *batchNorm = new CUDANet::Layers::BatchNorm(
new CUDANet::Layers::BatchNorm( batchNormSize, outputChannels, 1e-3f,
batchNormSize, outputChannels, 1e-3f, CUDANet::Layers::ActivationType::RELU CUDANet::Layers::ActivationType::RELU
); );
addLayer(prefix + ".conv", conv); addLayer(prefix + ".conv", conv);
addLayer(prefix + ".bn", batchNorm); addLayer(prefix + ".bn", batchNorm);
} }
float* forward(const float* d_input) {
for (auto& layer : layers) {
d_input = layer.second->forward(d_input);
}
return d_input;
}
private: private:
int inputSize; int inputSize;
int inputChannels; int inputChannels;
int outputChannels; int outputChannels;
}; };
class InceptionA : public CUDANet::Module {
public:
InceptionA(
const int inputSize,
const int inputChannels,
const int poolFeatures,
const std::string &prefix
)
: inputSize(inputSize),
inputChannels(inputChannels),
poolFeatures(poolFeatures) {
// Branch 1x1
CUDANet::Module *branch1x1 = new BasicConv2d(
inputSize, inputChannels, 64, 1, 1, 0, prefix + ".branch1x1"
);
// Branch 5x5
CUDANet::Module *branch5x5_1 = new BasicConv2d(
inputSize, inputChannels, 48, 1, 1, 0, prefix + ".branch5x5_1"
);
CUDANet::Module *branch5x5_2 = new BasicConv2d(
inputSize, 48, 64, 5, 1, 2, prefix + ".branch5x5_2"
);
// Branch 3x3
CUDANet::Module *branch3x3_1 = new BasicConv2d(
inputSize, inputChannels, 64, 1, 1, 0, prefix + ".branch3x3_1"
);
CUDANet::Module *branch3x3_2 = new BasicConv2d(
inputSize, 64, 96, 3, 1, 1, prefix + ".branch3x3_2"
);
CUDANet::Module *branch3x3_3 = new BasicConv2d(
inputSize, 96, 96, 3, 1, 1, prefix + ".branch3x3_3"
);
// Branch Pool
CUDANet::Module *branchPool = new BasicConv2d(
inputSize, inputChannels, poolFeatures, 1, 1, 0, prefix + ".branchPool"
);
}
private:
int inputSize;
int inputChannels;
int poolFeatures;
};