search

pipeless-ai / pipeless

An open-source computer vision framework to build and deploy apps in minutes
https://pipeless.ai
Apache License 2.0
726 stars 35 forks source link

Not running in ONNX Runtime #151

Open vaghawan opened 3 months ago

vaghawan commented 3 months ago

Hi, I'm currently using pipeless v1 currently in Jetson Orien with Jetpack 5.1.2, I've the ONNX runtime installed, and if I run the independent code out form pipeless, my inference runs fine. But when I run it using pipeless using onnx runtime, it throws the following error 

terminate called after throwing an instance of 'onnxruntime::OnnxRuntimeException'
  what():  /home/user/onnxruntime/include/onnxruntime/core/framework/ort_value.h:85 const T& OrtValue::Get() const [with T = onnxruntime::Tensor] IsTensor() was false. Trying to get a Tensor, but got: (null)

This happens for the given example of onnx-yolo as well. And if I were to use the Ultralytics YOLO the .pt version of the model, the pipeless gets stucks and doesn't reach to the point where it process frame.

Any help would be appreciated.

Thanks

miguelaeh commented 3 months ago

Hi @vaghawan ,

I have not seen that error before, which version of the onnx runtime are you using?

vaghawan commented 3 months ago

It's the 1.17.0 version : onnxruntime_gpu-1.17.0-cp38-cp38-linux_aarch64.whl

It's available through nvidia complied whl file, it seems this is the only option in Orien devices. These build are available through this : https://elinux.org/Jetson_Zoo#ONNX_Runtime