Description
I am trying to get tensorrt installed in a python3.8 (selected 3.8 because docs said that was what was supported) venv, where the server does not have access to pypi.nvidia.com. We have a local repo used for python packages. When trying to install tensorrt the install is failing indicating it cannot reach pypi.nvidia.com and that I should add the extra-index option to my pip command.
Environment
TensorRT Version: 10.0.1
GPU Type: A100-SXM4-40GB
Nvidia Driver Version: 550.54.15-1.el8
CUDA Version: 11.6.1-1
CUDNN Version:
Operating System + Version: Rocky 8.9
Python Version (if applicable): 3.8
TensorFlow Version (if applicable):
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag) - VM running in GCP. Trying to install directly in OS, not in a container in the VM.
Relevant Files
Please attach or include links to any models, data, files, or scripts necessary to reproduce your issue. (Github repo, Google Drive, Dropbox, etc.)
Steps To Reproduce
python3.8 -m venv tensorrt
source tensorrt/bin/activate
pip install -U pip
pip install cuda-python
pip install wheel
pip install tensorrt
The install fails at “Building wheel for tensorrt-cu12”.
I am afraid as well as not having public internet access, I cannot copy/paste out of the environment.
The errors show that it could not fetch a URL https://pypi.nvidia.com/tenasorrt-cu12-libs/
It reports a cert error, but this is because the security device blocking the public internet is not allowing the request out.
The pip.conf I have points to the local repo, and it works for other packages. It seems like the package itself is hardcoded to go out to pypi.nvidia.com ?
) PIP_EXTRA_INDEX_URL environment variable:
Is there a way to install without access to pypi.nvidia.com?
Please include:
* Exact steps/commands to build your repro
* Exact steps/commands to run your repro
* Full traceback of errors encountered