mmr/tkDNN
Archived
1
0
Fork 0
Code Issues 46 Pull Requests 12 Releases 4 Activity

Fix bad merge

Browse Source
This commit is contained in:
Adam Cuellar
2022-04-11 15:57:40 -04:00
parent 31075be60b
commit a37c14d71d
4 changed files with 273 additions and 195 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/tkDNN/NetworkRT.h
+65 -105
View File
@@ -11,6 +11,7 @@
#include <pluginsRT/ActivationLeakyRT.h>
#include <pluginsRT/ActivationLogisticRT.h>
#include <pluginsRT/ActivationMishRT.h>
#include <pluginsRT/ActivationSwishRT.h>
#include <pluginsRT/ActivationReLUCeilingRT.h>
#include <pluginsRT/DeformableConvRT.h>
#include <pluginsRT/FlattenConcatRT.h>
@@ -26,130 +27,89 @@
#include <pluginsRT/ConstantPaddingRT.h>
#include <pluginsRT/ReflectionPadding.h>
namespace tk { namespace dnn {
template<typename T> void writeBUF(char*& buffer, const T& val)
{
*reinterpret_cast<T*>(buffer) = val;
buffer += sizeof(T);
}
template<typename T> T readBUF(const char*& buffer)
{
T val = *reinterpret_cast<const T*>(buffer);
buffer += sizeof(T);
return val;
}
using namespace nvinfer1;
#include "pluginsRT/ActivationLeakyRT.h"
#include "pluginsRT/ActivationLogisticRT.h"
#include "pluginsRT/ActivationReLUCeilingRT.h"
#include "pluginsRT/ActivationMishRT.h"
#include "pluginsRT/ActivationSwishRT.h"
#include "pluginsRT/ReorgRT.h"
#include "pluginsRT/RegionRT.h"
#include "pluginsRT/RouteRT.h"
#include "pluginsRT/ShortcutRT.h"
#include "pluginsRT/YoloRT.h"
#include "pluginsRT/UpsampleRT.h"
#include "pluginsRT/ResizeLayerRT.h"
#include "pluginsRT/DeformableConvRT.h"
#include "pluginsRT/FlattenConcatRT.h"
#include "pluginsRT/ReshapeRT.h"
#include "pluginsRT/MaxPoolingFixedSizeRT.h"
class PluginFactory : IPluginFactory
{
public:
YoloRT *yolos[16];
int n_yolos;
virtual IPlugin* createPlugin(const char* layerName, const void* serialData, size_t serialLength);
};
namespace tk { namespace dnn {
class NetworkRT {
class NetworkRT {
public:
nvinfer1::DataType dtRT;
nvinfer1::IBuilder *builderRT;
nvinfer1::IRuntime *runtimeRT;
nvinfer1::INetworkDefinition *networkRT;
#if NV_TENSORRT_MAJOR >= 6
nvinfer1::IBuilderConfig *configRT;
public:
nvinfer1::DataType dtRT;
nvinfer1::IBuilder *builderRT;
nvinfer1::IRuntime *runtimeRT;
nvinfer1::INetworkDefinition *networkRT;
#if NV_TENSORRT_MAJOR >= 6
nvinfer1::IBuilderConfig *configRT;
#endif
nvinfer1::ICudaEngine *engineRT;
nvinfer1::IExecutionContext *contextRT;
const static int MAX_BUFFERS_RT = 10;
void* buffersRT[MAX_BUFFERS_RT];
dataDim_t buffersDIM[MAX_BUFFERS_RT];
int buf_input_idx, buf_output_idx;
bool builderActive = false;
dataDim_t input_dim, output_dim;
dnnType *output;
cudaStream_t stream;
nvinfer1::ICudaEngine *engineRT;
nvinfer1::IExecutionContext *contextRT;
std::vector<nvinfer1::YoloRT*> yolo_plugins; // yolo layers in network
const static int MAX_BUFFERS_RT = 10;
void* buffersRT[MAX_BUFFERS_RT];
dataDim_t buffersDIM[MAX_BUFFERS_RT];
int buf_input_idx, buf_output_idx;
bool builderActive = false;
dataDim_t input_dim, output_dim;
dnnType *output;
cudaStream_t stream;
NetworkRT(Network *net, const char *name);
virtual ~NetworkRT();
std::vector<nvinfer1::YoloRT*> yolo_plugins; // yolo layers in network
int getMaxBatchSize() {
if(engineRT != nullptr)
return engineRT->getMaxBatchSize();
else
return 0;
}
NetworkRT(Network *net, const char *name);
virtual ~NetworkRT();
int getBuffersN() {
if(engineRT != nullptr)
return engineRT->getNbBindings();
else
return 0;
}
int getMaxBatchSize() {
if(engineRT != nullptr)
return engineRT->getMaxBatchSize();
else
return 0;
}
/**
Do inference
*/
dnnType* infer(dataDim_t &dim, dnnType* data);
void enqueue(int batchSize = 1);
int getBuffersN() {
if(engineRT != nullptr)
return engineRT->getNbBindings();
else
return 0;
}
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Layer *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Conv2d *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Activation *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Dense *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Pooling *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Softmax *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Route *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Flatten *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Reshape *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Resize *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Reorg *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Region *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Shortcut *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Yolo *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Upsample *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, DeformConv2d *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input,Padding *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor* input,MulAdd *l);
/**
Do inference
*/
dnnType* infer(dataDim_t &dim, dnnType* data);
void enqueue(int batchSize = 1);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Layer *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Conv2d *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Activation *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Dense *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Pooling *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Softmax *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Route *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Flatten *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Reshape *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Resize *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Reorg *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Region *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Shortcut *l);
nvinfer1::IPluginV2Layer* convert_layer(nvinfer1::ITensor *input, Yolo *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Upsample *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, DeformConv2d *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input,Padding *l);
nvinfer1::ILayer* convert_layer(nvinfer1::ITensor* input,MulAdd *l);
#if NV_TENSORRT_MAJOR > 5 && NV_TENSORRT_MAJOR < 8
bool serialize(const char *filename);
bool serialize(const char *filename);
#else
bool serialize(const char *filename,nvinfer1::IHostMemory *ptr);
bool serialize(const char *filename,nvinfer1::IHostMemory *ptr);
#endif
bool deserialize(const char *filename);
void destroy();
bool deserialize(const char *filename);
void destroy();
};
};
}}
#endif //NETWORKRT_H
}}
#endif //NETWORKRT_H
include/tkDNN/pluginsRT/ActivationSwishRT.h
+61 -40
View File
@@ -1,61 +1,82 @@
#include<cassert>
#include "../kernels.h"
#include <NvInfer.h>
#include <vector>
class ActivationSwishRT : public IPlugin {
namespace nvinfer1 {
class ActivationSwishRT : public IPluginV2 {
public:
ActivationSwishRT() {
public:
ActivationSwishRT() ;
~ActivationSwishRT() ;
ActivationSwishRT(const void *data, size_t length) ;
}
int getNbOutputs() const NOEXCEPT override ;
~ActivationSwishRT(){
Dims getOutputDimensions(int index, const Dims *inputs, int nbInputDims) NOEXCEPT override ;
}
void configureWithFormat(const Dims *inputDims, int nbInputs, const Dims *outputDims, int nbOutputs, DataType type,
PluginFormat format, int maxBatchSize) NOEXCEPT override ;
int getNbOutputs() const override {
return 1;
}
int initialize() NOEXCEPT override ;
Dims getOutputDimensions(int index, const Dims* inputs, int nbInputDims) override {
return inputs[0];
}
void terminate() NOEXCEPT override ;
void configure(const Dims* inputDims, int nbInputs, const Dims* outputDims, int nbOutputs, int maxBatchSize) override {
size = 1;
for(int i=0; i<outputDims[0].nbDims; i++)
size *= outputDims[0].d[i];
}
size_t getWorkspaceSize(int maxBatchSize) const NOEXCEPT override ;
#if NV_TENSORRT_MAJOR > 7
int enqueue(int batchSize, const void *const *inputs, void *const *outputs, void *workspace,cudaStream_t stream) NOEXCEPT override ;
#elif NV_TENSORRT_MAJOR == 7
int32_t enqueue (int32_t batchSize, const void *const *inputs, void **outputs, void *workspace, cudaStream_t stream) override;
#endif
int initialize() override {
size_t getSerializationSize() const NOEXCEPT override ;
return 0;
}
void serialize(void *buffer) const NOEXCEPT override ;
virtual void terminate() override {
}
const char *getPluginType() const NOEXCEPT override ;
virtual size_t getWorkspaceSize(int maxBatchSize) const override {
return 0;
}
const char *getPluginVersion() const NOEXCEPT override ;
virtual int enqueue(int batchSize, const void*const * inputs, void** outputs, void* workspace, cudaStream_t stream) override {
void destroy() NOEXCEPT override { delete this; }
activationSwishForward((dnnType*)reinterpret_cast<const dnnType*>(inputs[0]),
reinterpret_cast<dnnType*>(outputs[0]), batchSize*size, stream);
return 0;
}
bool supportsFormat(DataType type, PluginFormat format) const NOEXCEPT override ;
const char *getPluginNamespace() const NOEXCEPT override ;
virtual size_t getSerializationSize() override {
return 1*sizeof(int);
}
void setPluginNamespace(const char *plguinNamespace) NOEXCEPT override ;
virtual void serialize(void* buffer) override {
char *buf = reinterpret_cast<char*>(buffer),*a=buf;
tk::dnn::writeBUF(buf, size);
assert(buf == a + getSerializationSize());
}
IPluginV2 *clone() const NOEXCEPT override ;
int size;
};
int size;
private:
std::string mPluginNamespace;
};
class ActivationSwishRTPluginCreator : public IPluginCreator {
public:
ActivationSwishRTPluginCreator() ;
void setPluginNamespace(const char *pluginNamespace) NOEXCEPT override ;
const char *getPluginNamespace() const NOEXCEPT override ;
IPluginV2 *deserializePlugin(const char *name, const void *serialData, size_t serialLength) NOEXCEPT override ;
IPluginV2 *createPlugin(const char *name, const PluginFieldCollection *fc) NOEXCEPT override ;
const char *getPluginName() const NOEXCEPT override ;
const char *getPluginVersion() const NOEXCEPT override ;
const PluginFieldCollection *getFieldNames() NOEXCEPT override ;
private:
static PluginFieldCollection mFC;
static std::vector<PluginField> mPluginAttributes;
std::string mPluginNamespace;
};
REGISTER_TENSORRT_PLUGIN(ActivationSwishRTPluginCreator);
};
src/NetworkRT.cpp
+14 -50
View File
@@ -646,17 +646,12 @@ ILayer* NetworkRT::convert_layer(ITensor *input, Activation *l) {
checkNULL(lRT);
return lRT;
}
else if(l->act_mode == CUDNN_ACTIVATION_ELU || l->act_mode == ACTIVATION_ELU){
IActivationLayer *lRT = networkRT->addActivation(*input,ActivationType::kELU);
else if(l->act_mode == ACTIVATION_SWISH) {
IPlugin *plugin = new ActivationSwishRT();
IPluginLayer *lRT = networkRT->addPlugin(&input, 1, *plugin);
checkNULL(lRT);
return lRT;
else if(l->act_mode == CUDNN_ACTIVATION_ELU || l->act_mode == ACTIVATION_ELU) {
IActivationLayer *lRT = networkRT->addActivation(*input, ActivationType::kELU);
}
else if(l->act_mode == ACTIVATION_LOGISTIC) {
IPlugin *plugin = new ActivationLogisticRT();
IPluginLayer *lRT = networkRT->addPlugin(&input, 1, *plugin);
else if(l->act_mode == ACTIVATION_SWISH) {
IPluginV2 *plugin = new ActivationSwishRT();
ILayer *lRT = networkRT->addPluginV2(&input, 1, *plugin);
checkNULL(lRT);
return lRT;
}
@@ -1034,48 +1029,17 @@ bool NetworkRT::deserialize(const char *filename) {
}
#if NV_TENSORRT_MAJOR > 7
void NetworkRT::destroy() {
delete contextRT;
if(builderActive) {
delete engineRT;
delete builderRT;
IPlugin* PluginFactory::createPlugin(const char* layerName, const void* serialData, size_t serialLength) {
const char * buf = reinterpret_cast<const char*>(serialData),*bufCheck = buf;
std::string name(layerName);
//std::cout<<name<<std::endl;
if(name.find("ActivationLeaky") == 0) {
ActivationLeakyRT *a = new ActivationLeakyRT(readBUF<float>(buf));
a->size = readBUF<int>(buf);
assert(buf == bufCheck + serialLength);
return a;
}
if(name.find("ActivationMish") == 0) {
ActivationMishRT *a = new ActivationMishRT();
a->size = readBUF<int>(buf);
assert(buf == bufCheck + serialLength);
return a;
}
if(name.find("ActivationSwish") == 0) {
ActivationSwishRT *a = new ActivationSwishRT();
a->size = readBUF<int>(buf);
assert(buf == bufCheck + serialLength);
return a;
}
if(name.find("ActivationLogistic") == 0) {
ActivationLogisticRT *a = new ActivationLogisticRT();
a->size = readBUF<int>(buf);
return a;
}
}
void NetworkRT::destroy() {
delete contextRT;
if(builderActive) {
delete engineRT;
delete builderRT;
}
}
#elif NV_TENSORRT_MAJOR <=7
void NetworkRT::destroy() {
void NetworkRT::destroy() {
}
#endif
}}
}}
src/pluginsRT/ActivationSwishRT.cpp
+133
View File
@@ -0,0 +1,133 @@
//
// Created by Adam on 4/11/2022
//
#include <tkDNN/pluginsRT/ActivationSwishRT.h>
using namespace nvinfer1;
std::vector<PluginField> ActivationSwishRTPluginCreator::mPluginAttributes;
PluginFieldCollection ActivationSwishRTPluginCreator::mFC{};
ActivationSwishRT::ActivationSwishRT() {
}
ActivationSwishRT::~ActivationSwishRT() {
}
ActivationSwishRT::ActivationSwishRT(const void *data, size_t length) {
const char *buf = reinterpret_cast<const char *>(data), *bufCheck = buf;
size = readBUF<int>(buf);
assert(buf == bufCheck + length);
}
int ActivationSwishRT::getNbOutputs() const NOEXCEPT { return 1; }
Dims ActivationSwishRT::getOutputDimensions(int index, const Dims *inputs, int nbInputDims) NOEXCEPT { return inputs[0]; }
void ActivationSwishRT::configureWithFormat(const Dims *inputDims, int nbInputs, const Dims *outputDims, int nbOutputs, DataType type,
PluginFormat format, int maxBatchSize) NOEXCEPT {
assert(format == PluginFormat::kLINEAR);
size = 1;
for (int i = 0; i < outputDims[0].nbDims; i++)
size *= outputDims[0].d[i];
}
int ActivationSwishRT::initialize() NOEXCEPT { return 0; }
void ActivationSwishRT::terminate() NOEXCEPT {}
size_t ActivationSwishRT::getWorkspaceSize(int maxBatchSize) const NOEXCEPT { return 0; }
#if NV_TENSORRT_MAJOR > 7
int ActivationSwishRT::enqueue(int batchSize, const void *const *inputs, void *const *outputs, void *workspace,
cudaStream_t stream) NOEXCEPT {
activationSwishForward((dnnType *) reinterpret_cast<const dnnType *>(inputs[0]),
reinterpret_cast<dnnType *>(outputs[0]), batchSize * size, stream);
return 0;
}
#elif NV_TENSORRT_MAJOR == 7
int32_t ActivationSwishRT::enqueue(int32_t batchSize, const void *const *inputs, void **outputs, void *workspace,
cudaStream_t stream) {
activationSwishForward((dnnType *) reinterpret_cast<const dnnType *>(inputs[0]),
reinterpret_cast<dnnType *>(outputs[0]), batchSize * size, stream);
return 0;
}
#endif
size_t ActivationSwishRT::getSerializationSize() const NOEXCEPT {
return 1 * sizeof(int);
}
void ActivationSwishRT::serialize(void *buffer) const NOEXCEPT {
char *buf = reinterpret_cast<char *>(buffer), *a = buf;
writeBUF(buf, size);
assert(buf == a + getSerializationSize());
}
const char* ActivationSwishRT::getPluginType() const NOEXCEPT {
return "ActivationSwishRT_tkDNN";
}
const char *ActivationSwishRT::getPluginVersion() const NOEXCEPT {
return "1";
}
bool ActivationSwishRT::supportsFormat(DataType type, PluginFormat format) const NOEXCEPT {
return (type == DataType::kFLOAT && format == PluginFormat::kLINEAR);
}
const char *ActivationSwishRT::getPluginNamespace() const NOEXCEPT {
return mPluginNamespace.c_str();
}
void ActivationSwishRT::setPluginNamespace(const char *plguinNamespace) NOEXCEPT {
mPluginNamespace = plguinNamespace;
}
IPluginV2 *ActivationSwishRT::clone() const NOEXCEPT {
auto *p = new ActivationSwishRT();
p->setPluginNamespace(mPluginNamespace.c_str());
return p;
}
ActivationSwishRTPluginCreator::ActivationSwishRTPluginCreator() {
mPluginAttributes.clear();
mFC.nbFields = mPluginAttributes.size();
mFC.fields = mPluginAttributes.data();
}
void ActivationSwishRTPluginCreator::setPluginNamespace(const char *pluginNamespace) NOEXCEPT {
mPluginNamespace = pluginNamespace;
}
const char *ActivationSwishRTPluginCreator::getPluginNamespace() const NOEXCEPT {
return mPluginNamespace.c_str();
}
IPluginV2 *ActivationSwishRTPluginCreator::deserializePlugin(const char *name, const void *serialData, size_t serialLength) NOEXCEPT {
auto *pluginObj = new ActivationSwishRT(serialData, serialLength);
pluginObj->setPluginNamespace(mPluginNamespace.c_str());
return pluginObj;
}
IPluginV2 *ActivationSwishRTPluginCreator::createPlugin(const char *name, const PluginFieldCollection *fc) NOEXCEPT {
const PluginField *fields = fc->fields;
auto *pluginObj = new ActivationSwishRT();
pluginObj->setPluginNamespace(mPluginNamespace.c_str());
return pluginObj;
}
const char *ActivationSwishRTPluginCreator::getPluginName() const NOEXCEPT {
return "ActivationSwishRT_tkDNN";
}
const char *ActivationSwishRTPluginCreator::getPluginVersion() const NOEXCEPT{
return "1";
}
const PluginFieldCollection *ActivationSwishRTPluginCreator::getFieldNames() NOEXCEPT {
return &mFC;
}