Description
I am using tritonserver:23.10-pyt-python-py3 as a base image for my triton server. I want to install the tensorrt==8.6.1 python package inside this container. After reading the documentation and some blogs I found that I should do that by installing python3-libnvinfer. Unfotunately apt is not able to find a version compatible with tesnorrt=8.6.1.6, it’s finding higher versions: from 8.6.2 and above (10.x.x).
Environment
TensorRT Version: 8.6.1.6
GPU Type: jetson orin gpu with jetpack 6.0
Nvidia Driver Version: 540.2.0
CUDA Version: 12.2
CUDNN Version: 8.9.5.29
Operating System + Version: ubuntu
Python Version: 3.10.12
Container: tritonserver:23.10-pyt-python-py3
Relevant Files
Dockerfile
FROM nvcr.io/nvidia/tritonserver:23.10-pyt-python-py3
RUN apt-get update && apt-get install -y --no-install-recommends
python3-pip python3-libnvinfer=8.6.1.6-1+cuda12.0 &&
rm -rf /var/lib/apt/lists/*
COPY requirements.txt ./requirements.txt
RUN pip install --no-cache -r requirements.txt
CMD [ “tritonserver”, “–model-repository=./model_registery”, “–log-verbose=1”, “–load-model=unet_resnet50”, “–allow-grpc=true”,“–model-control-mode=explicit” ]
Logs
apt list -a python3-libnvinfer
Listing… Done
python3-libnvinfer/unknown 10.3.0.26-1+cuda12.5 arm64
python3-libnvinfer/unknown 10.2.0.19-1+cuda12.5 arm64
python3-libnvinfer/unknown 10.1.0.27-1+cuda12.4 arm64
python3-libnvinfer/unknown 10.0.1.6-1+cuda12.4 arm64
python3-libnvinfer/unknown 10.0.0.6-1+cuda12.4 arm64
python3-libnvinfer/unknown 8.6.2.2-1+cuda12.0 arm64
Steps To Reproduce
apt update
apt-get install python3-libnvinfer=8.6.1.6-1+cuda12.0