代码拉取完成，页面将自动刷新
This repository contains the Open Source Software (OSS) components of NVIDIA TensorRT. Included are the sources for TensorRT plugins and parsers (Caffe and ONNX), as well as sample applications demonstrating usage and capabilities of the TensorRT platform.
To build the TensorRT OSS components, ensure you meet the following package requirements:
System Packages
GNU Make >= v4.1
CMake >= v3.13
PIP >= v19.0
Essential libraries and utilities
Cross compilation for Jetson platforms requires JetPack's host component installation
Optional Packages
Containerized builds
Code formatting tools
TensorRT Release
NOTE: Along with the TensorRT OSS components, the following source packages will also be downloaded, and they are not required to be installed on the system.
git clone -b master https://github.com/nvidia/TensorRT TensorRT -b release/7.0
cd TensorRT
git submodule update --init --recursive
export TRT_SOURCE=`pwd`
To build the TensorRT OSS, obtain the corresponding TensorRT 7.0 binary release from NVidia Developer Zone. For a list of key features, known and fixed issues, refer to the TensorRT 7.0 Release Notes.
Example: Ubuntu 18.04 with cuda-10.2
Download and extract the latest TensorRT 7.0 GA package for Ubuntu 18.04 and CUDA 10.2
cd ~/Downloads
# Download TensorRT-7.0.0.11.Ubuntu-18.04.x86_64-gnu.cuda-10.2.cudnn7.6.tar.gz
tar -xvzf TensorRT-7.0.0.11.Ubuntu-18.04.x86_64-gnu.cuda-10.2.cudnn7.6.tar.gz
export TRT_RELEASE=`pwd`/TensorRT-7.0.0.11
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$TRT_RELEASE/lib
Example: CentOS/RedHat 7 with cuda-10.0
Download and extract the TensorRT 7.0 GA for CentOS/RedHat 7 and CUDA 10.0 tar package
cd ~/Downloads
# Download TensorRT-7.0.0.11.CentOS-7.6.x86_64-gnu.cuda-10.0.cudnn7.6.tar.gz
tar -xvzf TensorRT-7.0.0.11.CentOS-7.6.x86_64-gnu.cuda-10.0.cudnn7.6.tar.gz
export TRT_RELEASE=`pwd`/TensorRT-7.0.0.11
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$TRT_RELEASE/lib
Using the SDK manager, download the host componets of the PDK version or Jetpack specified in the name of the Dockerfile. To do this:
Linux Jetpack 4.2.1), then click ContinueDownload & Install Options make note of or change the download folder and Select Download now. Install later. then agree to the license terms and click ContinueYou should now have all expected files to build the container. Move these into the docker/jetpack_files folder.
Install the System Packages list of components in the Prerequisites section.
Alternatively, use the build containers as described below:
The docker container can be built using the included Dockerfile. The build container is configured with the environment and packages required for building TensorRT OSS.
Example: Ubuntu 18.04 with cuda-10.2
docker build -f docker/ubuntu.Dockerfile --build-arg UBUNTU_VERSION=18.04 --build-arg CUDA_VERSION=10.2 --tag=tensorrt-ubuntu .
Example: CentOS/RedHat 7 with cuda-10.0
docker build -f docker/centos.Dockerfile --build-arg CENTOS_VERSION=7 --build-arg CUDA_VERSION=10.0 --tag=tensorrt-centos .
Example: Cross compile for JetPack 4.2.1 with cuda-10.0
docker build -f docker/ubuntu-cross-aarch64.Dockerfile --build-arg UBUNTU_VERSION=18.04 --build-arg CUDA_VERSION=10.0 --tag tensorrt-ubuntu-aarch64 .
`
docker run -v $TRT_RELEASE:/tensorrt -v $TRT_SOURCE:/workspace/TensorRT -it tensorrt-ubuntu:latest
NOTE: To run TensorRT/CUDA programs within the build container, install nvidia-docker. Replace the
docker runcommand withnvidia-docker runordocker run --runtime=nvidia.
Generate Makefiles and build.
cd $TRT_SOURCE
mkdir -p build && cd build
cmake .. -DTRT_LIB_DIR=$TRT_RELEASE/lib -DTRT_BIN_DIR=`pwd`/out
make -j$(nproc)
NOTE:
- The default CUDA version used by CMake is 10.2. To override this, for example to 10.0, append
-DCUDA_VERSION=10.0to the cmake command.- Samples may fail to link on CentOS7. To work around this create the following symbolic link:
ln -s $TRT_BIN_DIR/libnvinfer_plugin.so $TRT_BIN_DIR/libnvinfer_plugin.so.7
The required CMake arguments are:
TRT_LIB_DIR: Path to the TensorRT installation directory containing libraries.
TRT_BIN_DIR: Output directory where generated build artifacts will be copied.
The following CMake build parameters are optional:
CMAKE_BUILD_TYPE: Specify if binaries generated are for release or debug (contain debug symbols). Values consists of [Release] | Debug
CUDA_VERISON: The version of CUDA to target, for example [10.2].
CUDNN_VERSION: The version of cuDNN to target, for example [7.6].
PROTOBUF_VERSION: The version of Protobuf to use, for example [3.8.x]. Note: Changing this will not configure CMake to use a system version of Protobuf, it will configure CMake to download and try building that version.
CMAKE_TOOLCHAIN_FILE: The path to a toolchain file for cross compilation.
BUILD_PARSERS: Specify if the parsers should be built, for example [ON] | OFF. If turned OFF, CMake will try to find precompiled versions of the parser libraries to use in compiling samples. First in ${TRT_LIB_DIR}, then on the system. If the build type is Debug, then it will prefer debug builds of the libraries before release versions if available.
BUILD_PLUGINS: Specify if the plugins should be built, for example [ON] | OFF. If turned OFF, CMake will try to find a precompiled version of the plugin library to use in compiling samples. First in ${TRT_LIB_DIR}, then on the system. If the build type is Debug, then it will prefer debug builds of the libraries before release versions if available.
BUILD_SAMPLES: Specify if the samples should be built, for example [ON] | OFF.
Other build options with limited applicability:
NVINTERNAL: Used by TensorRT team for internal builds. Values consists of [OFF] | ON.
PROTOBUF_INTERNAL_VERSION: The version of protobuf to use, for example [10.0]. Only applicable if NVINTERNAL is also enabled.
NVPARTNER: For use by NVIDIA partners with exclusive source access. Values consists of [OFF] | ON.
CUB_VERSION: The version of CUB to use, for example [1.8.0].
GPU_ARCHS: GPU (SM) architectures to target. By default we generate CUDA code for all major SMs. Specific SM versions can be specified here as a quoted space-separated list to reduce compilation time and binary size. Table of compute capabilities of NVIDIA GPUs can be found here. Examples:
-DGPU_ARCHS="70"-DGPU_ARCHS="70"-DGPU_ARCHS="75"-DGPU_ARCHS="75"-DGPU_ARCHS="70 75"Copy the build artifacts into the TensorRT installation directory, updating the installation.
TensorRT installation directory is determined as $TRT_LIB_DIR/..
Installation might require superuser privileges depending on the path and permissions of files being replaced.
Installation is not supported in cross compilation scenario. Please copy the result files from build/out folder into the target device.
sudo make install
Verify the TensorRT samples have been installed correctly.
cd $TRT_LIB_DIR/../bin/
./sample_googlenet
If the sample was installed correctly, the following information will be printed out in the terminal.
[08/23/2019-22:08:57] [I] Building and running a GPU inference engine for GoogleNet
[08/23/2019-22:08:59] [I] [TRT] Some tactics do not have sufficient workspace memory to run. Increasing workspace size may increase performance, please check verbose output.
[08/23/2019-22:09:05] [I] [TRT] Detected 1 inputs and 1 output network tensors.
[08/23/2019-22:09:05] [I] Ran /tensorrt/bin/sample_googlenet with:
[08/23/2019-22:09:05] [I] Input(s): data
[08/23/2019-22:09:05] [I] Output(s): prob
&&&& PASSED TensorRT.sample_googlenet # /tensorrt/bin/sample_googlenet
此处可能存在不合适展示的内容，页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容，可点击提交进行申诉，我们将尽快为您处理。
马建仓 AI 助手
![dependabot[bot]](data:image/png;base64,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)
