Skip to content

Latest commit

 

History

History
95 lines (75 loc) · 3.67 KB

README.md

File metadata and controls

95 lines (75 loc) · 3.67 KB
Raw

Build Process TensorRT Custom plugin

TensorRT OSS build is required. This is required because TensorRT TRT_EfficientNMSX plugin that are required by models and not in the general TensorRT release.

New plugin

Downloading TensorRT Build

  1. Download TensorRT OSS

    cd TensorRTPlugin
git clone -b release/8.6 https://github.com/levipereira/TensorRT TensorRT
cd TensorRT
git submodule update --init --recursive

Setting Up The Build Environment

For Linux platforms, we recommend that you generate a docker container for building TensorRT OSS as described below.

  1. Generate the TensorRT-OSS build container.

    The TensorRT-OSS build container can be generated using the supplied Dockerfiles and build scripts. The build containers are configured for building TensorRT OSS out-of-the-box.

    Ubuntu 20.04 on x86-64 with cuda-12.1 (DeepStream 7.0)

    ./docker/build.sh --file docker/ubuntu-20.04.Dockerfile --tag tensorrt-ubuntu20.04-cuda12.1

    Example: Ubuntu 20.04 cross-compile for Jetson (aarch64) with cuda-11.4.2 (JetPack SDK)

    ./docker/build.sh --file docker/ubuntu-cross-aarch64.Dockerfile --tag tensorrt-jetpack-cuda11.4

    Example: Ubuntu 20.04 on aarch64 with cuda-11.8

    ./docker/build.sh --file docker/ubuntu-20.04-aarch64.Dockerfile --tag tensorrt-aarch64-ubuntu20.04-cuda11.8 --cuda 11.8.0

  2. Launch the TensorRT-OSS build container.

    Example: Ubuntu 20.04 build container

    ./docker/launch.sh --tag tensorrt-ubuntu20.04-cuda12.1 --gpus all

    NOTE:
    sudo password for Ubuntu build containers is nvidia.

Building TensorRT-OSS

  • Generate Makefiles and build.

    Example: Linux (x86-64) build with default cuda-12.1 - DeepStream 7.0

     cd $TRT_OSSPATH
 mkdir -p build && cd build
 cmake .. -DTRT_LIB_DIR=$TRT_LIBPATH -DTRT_OUT_DIR=`pwd`/out
 make -j$(nproc)

    Example: Linux (aarch64) build with default cuda-12.1

     cd $TRT_OSSPATH
 mkdir -p build && cd build
 cmake .. -DTRT_LIB_DIR=$TRT_LIBPATH -DTRT_OUT_DIR=`pwd`/out -DCMAKE_TOOLCHAIN_FILE=$TRT_OSSPATH/cmake/toolchains/cmake_aarch64-native.toolchain
 make -j$(nproc)

NOTE:

  • Required CMake build arguments are:
    • TRT_LIB_DIR: Path to the TensorRT installation directory containing libraries.
    • TRT_OUT_DIR: Output directory where generated build artifacts will be copied.
  • Optional CMake build arguments:
    • 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:
      • NVidia A100: -DGPU_ARCHS="80"
      • Tesla T4, GeForce RTX 2080: -DGPU_ARCHS="75"
      • Titan V, Tesla V100: -DGPU_ARCHS="70"
      • Multiple SMs: -DGPU_ARCHS="80 75"

After building ends successfully, libnvinfer_plugin.so* will be generated under ${pwd}/out/.
exit from docker container to host and copy libnvinfer_plugin

## exit from container
exit
## copy lib to deepstream-yolov9/TensorRTPlugin
cp build/out/libnvinfer_plugin.so.8.6.1 ../../TensorRTPlugin/

## on this example we copied from 
# ~/deepstream-yolov9/TensorRTPlugin/TensorRT/build/out/libnvinfer_plugin.so.8.6.1  ~/deepstream-yolov9/TensorRTPlugin/

## optional to clean
## you can remove TensorRT dir
rm -rf deepstream-yolov9/TensorRTPlugin/TensorRT