Deepstream 6.1 Segmentation Fault on WSL2

• Hardware Platform (Jetson / GPU)
RTX 3060 on WSL2 Ubuntu 20.04
• DeepStream Version
Deepstream 6.1
• TensorRT Version
TensorRT 8.2.4.2
• NVIDIA GPU Driver Version
512.15
• Issue Type( questions, new requirements, bugs)
Bug
•CUDA Version
11.4
Other Softwares
Cudnn 8.4.1, Gstreamer 1.16.2

I’ve a kind of a bug thing when I use deepstream-app & deepstream-transfer-learning app on my WSL ubuntu setup. When I try to execute TAO examples from base deepstream everything seems fine but on my custom TAO trained model there is not.

I have a .etlt model which trained from TAO and .h264 video file. I need to process them with deepstream but when I execute the command with my config and app files segmentation fault appears with any error messages with GST_DEBUG=3 parameter.

This is the screenshot of my terminal output when I exeucte the command GST_DEBUG=3 deepstream-transfer-learning-app -c deepstream_app_head.txt

PS : I use Batch Size as 1 and checked my GPU consumption and there is not any leak on GPU which might be caused for a segmentation fault.

Sorry, We cannot support WSL or WSL2 env yet. You can test on other linux env or jetson platform. Could you tell us about your work env and the reason for using the WSL2 env? Thanks

1 Like