Are there any limitations to running more than one model at a time?

[lxbhahaha]

When I try to create two new models at the same time (for example, two yolox), the inference of the first model is lost, and the result of the second model is normal. Can I only run one model at a time? Or did I do something wrong?

[asus4]

I tried loading two models in the NanoSAM example without any problem.

@lxbhahaha Can you make a minimal reproducible project? I would like to know your use case in code.

[lxbhahaha]

I tried loading two models in the NanoSAM example without any problem.

@lxbhahaha Can you make a minimal reproducible project? I would like to know your use case in code.

I uploaded the problematic part of the project to TestOnnx_250117.

Here are lines 30 to 37 of the code in file Assets/Scripts/DemoTest.cs.

30    // yolo1 = new DemoYoloxHumanDetect();
31    // hrnet1 = new DemoHRNet();
32
33    ImageAcquisiton.Instance.InitAndPlay();
34    demoYolo= new DemoYoloxHumanDetect();
35
36    // yolo2 = new DemoYoloxHumanDetect();
37    // hrnet2 = new DemoHRNet();

It is currently work fine, But if you uncomment lines 36 or 37, demoYolo will no longer get the result. If you just uncomment lines 30 or 31, demoyolo will still get the results.

It seems that only the most recent model can get results, and I can't figure out why this problem occurs.

[asus4]

@lxbhahaha, thanks for providing the repro project. That is weird. We may need to test with the original C# library to isolate whether it is a Unity or the original OnnxRuntime issue.