Could not initialize cublas. Please check CUDA installation

Description

I get below error when I try to create an engine from .etlt model using tao_converter.

[ERROR] 1: [wrapper.cpp::CublasWrapper::85] Error Code 1: Cublas (Could not initialize cublas. Please check CUDA installation.)
[ERROR] Unable to create engine

Environment

TensorRT Version: 8.2.3
GPU Type: Geforce RTX 3060
Nvidia Driver Version: 530.30.02
CUDA Version: 11.6.1
CUDNN Version: 8.3.3
Operating System + Version: Ubuntu 20.04
Python Version (if applicable): 3.8
TensorFlow Version (if applicable): NA
PyTorch Version (if applicable): NA
Baremetal or Container (if container which image + tag): Container - nvcr.io/nvidia/tensorrt:22.03-py3

This is my Dockerfile:

FROM nvcr.io/nvidia/tensorrt:22.03-py3

RUN apt-get update && apt-get install libgl1 -y

RUN pip3 install -U pip

ADD requirements.txt /tmp/requirements.txt

RUN pip3 install -r /tmp/requirements.txt

RUN /opt/tensorrt/install_opensource.sh

CMD ["bash"]

The image got built without any error, but I get the above mentioned error when I try to convert the etlt model to engine using tao_converter(8.2 version). Kindly help me fix this issue.

Hi,
Please refer to the installation steps from the below link if in case you are missing on anything

Also, we suggest you to use TRT NGC containers to avoid any system dependency related issues.

Thanks!

Hi @AakankshaS , Thanks for the reply. I’m already using the TensorRT container from NGC. Kindly refer the my comment in the environment section. The container version I’m using - nvcr.io/nvidia/tensorrt:22.03-py3

Hi Team, Any lead on this please?

Hi @arivarasan.e ,
In the container, and your environment details, i see a TRT and CUDA version
mismatch from what the container holds.

[quote=“arivarasan.e, post:1, topic:250843”]
TensorRT Version: 8.2.3
GPU Type: Geforce RTX 3060
Nvidia Driver Version: 530.30.02
CUDA Version: 11.6.1
[/quote].
Can you please check the container components and retry again.

Thanks