Illegal instruction on python import of tensorrt 10.7 on GPU RTX 3090 in WSL2

[YonatanOfek]

Description

I Installed TensorRT 10.7.0 GA on an ubuntu 22.04 distro in WSL2, and then also pip installed tensorrt. trtexec command works, the maskrcnn onnx conversion example works. But now whenver I import tensorrt in python I get Illegal action.

Environment

TensorRT Version: 10.7.0

NVIDIA GPU: RTX 3090

NVIDIA Driver Version: 560.94

CUDA Version: 12.6.20

CUDNN Version: 9.5.1

Illegal action might be related to cpu so here's my model: AMD Threadripper 2950x

Operating System:

Python Version (if applicable): 3.9.21

Tensorflow Version (if applicable): --

PyTorch Version (if applicable): Probably irrelevant but I have 2.5.1 in my conda environment.

Baremetal or Container (if so, version): Baremetal.

Relevant Files

Model link:
No model, happens on import to python.

Steps To Reproduce

Commands or scripts:

1. Create and enter a WSL2 ubuntu 22.04 distro on windows 10 home edition.

2. Follow section 3.2 of these instructions: https://docs.nvidia.com/deeplearning/tensorrt/install-guide/index.html
   Download and install this version of 10.7: TensorRT 10.7 GA for Ubuntu 20.04 and CUDA 12.0 to 12.6 DEB local repo Package

3. pip install tensorrt==10.7.0

Have you tried the latest release?: Yes

Can this model run on other frameworks? For example run ONNX model with ONNXRuntime (polygraphy run <model.onnx> --onnxrt):

[YonatanOfek]

My mistake: I installed TensorRT with python 3.12.4 and then tried pip installing tensorrt=10.7.0 in a conda environment with python 3.9.21...
pip install tensorrt + import tensorrt works in the 3.12.4 python env...