mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2026-05-10 01:21:55 +08:00
43cabceb6d
* [Backend] TRT backend & PP-Infer backend support pinned memory (#403) * TRT backend use pinned memory * refine fd tensor pinned memory logic * TRT enable pinned memory configurable * paddle inference support pinned memory * pinned memory pybindings Co-authored-by: Jason <jiangjiajun@baidu.com> * [Bug Fix] release task scripts (#411) * Update py_run.bat * Update cpp_run.bat * Update compare_with_gt.py Increase score_diff and boxes_diff_ratio threshold * Update cpp_run.bat * Update release task scripts according to diffrent platforms * Delete CMAKE_CXX_COMPILER in cpp_run.bat * [Doc] add contributor for js application (#413) add contributor * [Other] Refactor js submodule (#415) * Refactor js submodule * Remove change-log * Update ocr module * Update ocr-detection module * Update ocr-detection module * Remove change-log * [Doc] Add PicoDet & PaddleClas Android demo docs (#412) * [Backend] Add override flag to lite backend * [Docs] Add Android C++ SDK build docs * [Doc] fix android_build_docs typos * Update CMakeLists.txt * Update android.md * [Doc] Add PicoDet Android demo docs * [Doc] Update PicoDet Andorid demo docs * [Doc] Update PaddleClasModel Android demo docs * [Doc] Update fastdeploy android jni docs * [Doc] Update fastdeploy android jni usage docs Co-authored-by: Jason <jiangjiajun@baidu.com> * Update README.md * Update README_CN.md * Update README_CN.md * Update README_EN.md * [Doc] Add tutorial of supporting new models (#418) * first commit for yolov7 * pybind for yolov7 * CPP README.md * CPP README.md * modified yolov7.cc * README.md * python file modify * delete license in fastdeploy/ * repush the conflict part * README.md modified * README.md modified * file path modified * file path modified * file path modified * file path modified * file path modified * README modified * README modified * move some helpers to private * add examples for yolov7 * api.md modified * api.md modified * api.md modified * YOLOv7 * yolov7 release link * yolov7 release link * yolov7 release link * copyright * change some helpers to private * change variables to const and fix documents. * gitignore * Transfer some funtions to private member of class * Transfer some funtions to private member of class * Merge from develop (#9) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> * first commit for yolor * for merge * Develop (#11) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> * Yolor (#16) * Develop (#11) (#12) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> * Develop (#13) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> * documents * documents * documents * documents * documents * documents * documents * documents * documents * documents * documents * documents * Develop (#14) * Fix compile problem in different python version (#26) * fix some usage problem in linux * Fix compile problem Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> * Add PaddleDetetion/PPYOLOE model support (#22) * add ppdet/ppyoloe * Add demo code and documents * add convert processor to vision (#27) * update .gitignore * Added checking for cmake include dir * fixed missing trt_backend option bug when init from trt * remove un-need data layout and add pre-check for dtype * changed RGB2BRG to BGR2RGB in ppcls model * add model_zoo yolov6 c++/python demo * fixed CMakeLists.txt typos * update yolov6 cpp/README.md * add yolox c++/pybind and model_zoo demo * move some helpers to private * fixed CMakeLists.txt typos * add normalize with alpha and beta * add version notes for yolov5/yolov6/yolox * add copyright to yolov5.cc * revert normalize * fixed some bugs in yolox * fixed examples/CMakeLists.txt to avoid conflicts * add convert processor to vision * format examples/CMakeLists summary * Fix bug while the inference result is empty with YOLOv5 (#29) * Add multi-label function for yolov5 * Update README.md Update doc * Update fastdeploy_runtime.cc fix variable option.trt_max_shape wrong name * Update runtime_option.md Update resnet model dynamic shape setting name from images to x * Fix bug when inference result boxes are empty * Delete detection.py Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Jason <928090362@qq.com> * add is_dynamic for YOLO series (#22) * modify ppmatting backend and docs * modify ppmatting docs * fix the PPMatting size problem * fix LimitShort's log * retrigger ci * modify PPMatting docs * modify the way for dealing with LimitShort * change develop_a_new_model.md dir Co-authored-by: Jason <jiangjiajun@baidu.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Jason <928090362@qq.com> * [Doc] add readme for js packages (#421) * add contributor * add package readme * refine ocr readme * refine ocr readme Co-authored-by: Wang Xinyu <wangxinyu_es@163.com> Co-authored-by: huangjianhui <852142024@qq.com> Co-authored-by: Double_V <liuvv0203@163.com> Co-authored-by: chenqianhe <54462604+chenqianhe@users.noreply.github.com> Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> Co-authored-by: leiqing <54695910+leiqing1@users.noreply.github.com> Co-authored-by: ziqi-jin <67993288+ziqi-jin@users.noreply.github.com> Co-authored-by: root <root@bjyz-sys-gpu-kongming3.bjyz.baidu.com>
282 lines
7.4 KiB
C++
282 lines
7.4 KiB
C++
// Copyright (c) 2022 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.
|
|
|
|
#pragma once
|
|
|
|
#include <cuda_runtime_api.h>
|
|
|
|
#include <algorithm>
|
|
#include <iostream>
|
|
#include <map>
|
|
#include <memory>
|
|
#include <numeric>
|
|
#include <string>
|
|
#include <vector>
|
|
|
|
#include "NvInfer.h"
|
|
#include "fastdeploy/core/allocate.h"
|
|
#include "fastdeploy/core/fd_tensor.h"
|
|
#include "fastdeploy/utils/utils.h"
|
|
|
|
namespace fastdeploy {
|
|
|
|
struct FDInferDeleter {
|
|
template <typename T>
|
|
void operator()(T* obj) const {
|
|
if (obj) {
|
|
delete obj;
|
|
// obj->destroy();
|
|
}
|
|
}
|
|
};
|
|
|
|
template <typename T>
|
|
using FDUniquePtr = std::unique_ptr<T, FDInferDeleter>;
|
|
|
|
int64_t Volume(const nvinfer1::Dims& d);
|
|
|
|
nvinfer1::Dims ToDims(const std::vector<int>& vec);
|
|
nvinfer1::Dims ToDims(const std::vector<int64_t>& vec);
|
|
|
|
size_t TrtDataTypeSize(const nvinfer1::DataType& dtype);
|
|
|
|
FDDataType GetFDDataType(const nvinfer1::DataType& dtype);
|
|
|
|
nvinfer1::DataType ReaderDtypeToTrtDtype(int reader_dtype);
|
|
|
|
std::vector<int> ToVec(const nvinfer1::Dims& dim);
|
|
|
|
template <typename T>
|
|
std::ostream& operator<<(std::ostream& out, const std::vector<T>& vec) {
|
|
out << "[";
|
|
for (size_t i = 0; i < vec.size(); ++i) {
|
|
if (i != vec.size() - 1) {
|
|
out << vec[i] << ", ";
|
|
} else {
|
|
out << vec[i] << "]";
|
|
}
|
|
}
|
|
return out;
|
|
}
|
|
|
|
template <typename AllocFunc, typename FreeFunc>
|
|
class FDGenericBuffer {
|
|
public:
|
|
//!
|
|
//! \brief Construct an empty buffer.
|
|
//!
|
|
explicit FDGenericBuffer(nvinfer1::DataType type = nvinfer1::DataType::kFLOAT)
|
|
: mSize(0),
|
|
mCapacity(0),
|
|
mType(type),
|
|
mBuffer(nullptr),
|
|
mExternal_buffer(nullptr) {}
|
|
|
|
//!
|
|
//! \brief Construct a buffer with the specified allocation size in bytes.
|
|
//!
|
|
FDGenericBuffer(size_t size, nvinfer1::DataType type)
|
|
: mSize(size), mCapacity(size), mType(type) {
|
|
if (!allocFn(&mBuffer, this->nbBytes())) {
|
|
throw std::bad_alloc();
|
|
}
|
|
}
|
|
|
|
//!
|
|
//! \brief This use to skip memory copy step.
|
|
//!
|
|
FDGenericBuffer(size_t size, nvinfer1::DataType type, void* buffer)
|
|
: mSize(size), mCapacity(size), mType(type) {
|
|
mExternal_buffer = buffer;
|
|
}
|
|
|
|
FDGenericBuffer(FDGenericBuffer&& buf)
|
|
: mSize(buf.mSize),
|
|
mCapacity(buf.mCapacity),
|
|
mType(buf.mType),
|
|
mBuffer(buf.mBuffer) {
|
|
buf.mSize = 0;
|
|
buf.mCapacity = 0;
|
|
buf.mType = nvinfer1::DataType::kFLOAT;
|
|
buf.mBuffer = nullptr;
|
|
}
|
|
|
|
FDGenericBuffer& operator=(FDGenericBuffer&& buf) {
|
|
if (this != &buf) {
|
|
freeFn(mBuffer);
|
|
mSize = buf.mSize;
|
|
mCapacity = buf.mCapacity;
|
|
mType = buf.mType;
|
|
mBuffer = buf.mBuffer;
|
|
// Reset buf.
|
|
buf.mSize = 0;
|
|
buf.mCapacity = 0;
|
|
buf.mBuffer = nullptr;
|
|
}
|
|
return *this;
|
|
}
|
|
|
|
//!
|
|
//! \brief Returns pointer to underlying array.
|
|
//!
|
|
void* data() {
|
|
if (mExternal_buffer != nullptr) return mExternal_buffer;
|
|
return mBuffer;
|
|
}
|
|
|
|
//!
|
|
//! \brief Returns pointer to underlying array.
|
|
//!
|
|
const void* data() const {
|
|
if (mExternal_buffer != nullptr) return mExternal_buffer;
|
|
return mBuffer;
|
|
}
|
|
|
|
//!
|
|
//! \brief Returns the size (in number of elements) of the buffer.
|
|
//!
|
|
size_t size() const { return mSize; }
|
|
|
|
//!
|
|
//! \brief Returns the size (in bytes) of the buffer.
|
|
//!
|
|
size_t nbBytes() const { return this->size() * TrtDataTypeSize(mType); }
|
|
|
|
//!
|
|
//! \brief Set user memory buffer for TRT Buffer
|
|
//!
|
|
void SetExternalData(size_t size, nvinfer1::DataType type, void* buffer) {
|
|
mSize = mCapacity = size;
|
|
mType = type;
|
|
mExternal_buffer = const_cast<void*>(buffer);
|
|
}
|
|
|
|
//!
|
|
//! \brief Set user memory buffer for TRT Buffer
|
|
//!
|
|
void SetExternalData(const nvinfer1::Dims& dims, const void* buffer) {
|
|
mSize = mCapacity = Volume(dims);
|
|
mExternal_buffer = const_cast<void*>(buffer);
|
|
}
|
|
|
|
//!
|
|
//! \brief Resizes the buffer. This is a no-op if the new size is smaller than
|
|
//! or equal to the current capacity.
|
|
//!
|
|
void resize(size_t newSize) {
|
|
mExternal_buffer = nullptr;
|
|
mSize = newSize;
|
|
if (mCapacity < newSize) {
|
|
freeFn(mBuffer);
|
|
if (!allocFn(&mBuffer, this->nbBytes())) {
|
|
throw std::bad_alloc{};
|
|
}
|
|
mCapacity = newSize;
|
|
}
|
|
}
|
|
|
|
//!
|
|
//! \brief Overload of resize that accepts Dims
|
|
//!
|
|
void resize(const nvinfer1::Dims& dims) { return this->resize(Volume(dims)); }
|
|
|
|
~FDGenericBuffer() {
|
|
mExternal_buffer = nullptr;
|
|
freeFn(mBuffer);
|
|
}
|
|
|
|
private:
|
|
size_t mSize{0}, mCapacity{0};
|
|
nvinfer1::DataType mType;
|
|
void* mBuffer;
|
|
void* mExternal_buffer;
|
|
AllocFunc allocFn;
|
|
FreeFunc freeFn;
|
|
};
|
|
|
|
using FDDeviceBuffer = FDGenericBuffer<FDDeviceAllocator, FDDeviceFree>;
|
|
using FDDeviceHostBuffer = FDGenericBuffer<FDDeviceHostAllocator,
|
|
FDDeviceHostFree>;
|
|
|
|
class FDTrtLogger : public nvinfer1::ILogger {
|
|
public:
|
|
static FDTrtLogger* logger;
|
|
static FDTrtLogger* Get() {
|
|
if (logger != nullptr) {
|
|
return logger;
|
|
}
|
|
logger = new FDTrtLogger();
|
|
return logger;
|
|
}
|
|
void log(nvinfer1::ILogger::Severity severity,
|
|
const char* msg) noexcept override {
|
|
if (severity == nvinfer1::ILogger::Severity::kINFO) {
|
|
// Disable this log
|
|
// FDINFO << msg << std::endl;
|
|
} else if (severity == nvinfer1::ILogger::Severity::kWARNING) {
|
|
// Disable this log
|
|
// FDWARNING << msg << std::endl;
|
|
} else if (severity == nvinfer1::ILogger::Severity::kERROR) {
|
|
FDERROR << msg << std::endl;
|
|
} else if (severity == nvinfer1::ILogger::Severity::kINTERNAL_ERROR) {
|
|
FDASSERT(false, "%s", msg);
|
|
}
|
|
}
|
|
};
|
|
|
|
struct ShapeRangeInfo {
|
|
explicit ShapeRangeInfo(const std::vector<int64_t>& new_shape) {
|
|
shape.assign(new_shape.begin(), new_shape.end());
|
|
min.resize(new_shape.size());
|
|
max.resize(new_shape.size());
|
|
is_static.resize(new_shape.size());
|
|
for (size_t i = 0; i < new_shape.size(); ++i) {
|
|
if (new_shape[i] > 0) {
|
|
min[i] = new_shape[i];
|
|
max[i] = new_shape[i];
|
|
is_static[i] = 1;
|
|
} else {
|
|
min[i] = -1;
|
|
max[i] = -1;
|
|
is_static[i] = 0;
|
|
}
|
|
}
|
|
}
|
|
|
|
std::string name;
|
|
std::vector<int64_t> shape;
|
|
std::vector<int64_t> min;
|
|
std::vector<int64_t> max;
|
|
std::vector<int64_t> opt;
|
|
std::vector<int8_t> is_static;
|
|
// return
|
|
// -1: new shape is inillegal
|
|
// 0 : new shape is able to inference
|
|
// 1 : new shape is out of range, need to update engine
|
|
int Update(const std::vector<int64_t>& new_shape);
|
|
int Update(const std::vector<int>& new_shape) {
|
|
std::vector<int64_t> new_shape_int64(new_shape.begin(), new_shape.end());
|
|
return Update(new_shape_int64);
|
|
}
|
|
|
|
friend std::ostream& operator<<(std::ostream& out,
|
|
const ShapeRangeInfo& info) {
|
|
out << "Input name: " << info.name << ", shape=" << info.shape
|
|
<< ", min=" << info.min << ", max=" << info.max << std::endl;
|
|
return out;
|
|
}
|
|
};
|
|
|
|
} // namespace fastdeploy
|