cudnn+tensorRTに対応

[masakifujiwara1]

サポートバージョン対応表 https://docs.nvidia.com/deeplearning/tensorrt/support-matrix/index.html

[masakifujiwara1]

• tensorRT 10.0
• cuda 11.3
• cudnn 8.9.7

[masakifujiwara1]

tensorrt download https://developer.nvidia.com/tensorrt/download/10x

やり方 https://qiita.com/chin_self_driving_car/items/c3bbf88d9a0ee203a7f7

apt installするもの tensorrt

[masakifujiwara1]

cudnn download https://developer.nvidia.com/cudnn-archive

やり方 tensorrtと同様（但し，installはlibcudnn8）

apt installするもの libcudnn8 libcudnn8-dev

[masakifujiwara1]

TensorRT OSS 22.02 install（-DGPU_ARCHS: 1080の場合）

git clone -b 22.02 https://github.com/NVIDIA/TensorRT.git TensorRT
cd TensorRT
git submodule update --init --recursive
mkdir -p build && cd build
export PATH=/usr/local/cuda-11.3/bin:$PATH
cmake .. -DCUDA_VERSION=$CUDA_VERSION -DGPU_ARCHS="61" -DCUDNN_LIB=/usr/lib/x86_64-linux-gnu/libcudnn.so.8.9.7
make nvinfer_plugin -j$(nproc)
make nvinfer_plugin_static -j$(nproc)
cp libnvinfer_plugin.so.8.2.* /usr/lib/$ARCH-linux-gnu/libnvinfer_plugin.so.8.2.3
cp libnvinfer_plugin_static.a /usr/lib/$ARCH-linux-gnu/libnvinfer_plugin_static.a

[masakifujiwara1]

• tensorRT 8.2.x
• cuda 11.3
• cudnn 8.2.1