Is there anyway I can convert csp pytorch trained model to ONNX?

[huyhoangle86]

Actually I'm struggling with this. I’m trying to export my trained model that in .pt format to ONNX but after I execute the script the program output bunch of logging stuff like in the pic below and the converted onnx format seems not appear in my local disk I don’t know what I got wrong here? I tried to give it full path but no luck.

And I just read a website it says about onnx " However, whenever a model contains control flow, like for loops or if statements, the tracer method will fail, simply because the tracer(onnx) is never aware of the existence of the control flow statements, it faithfully records the flow based on the supplied input." . Is that correct?

[WongKinYiu]

https://github.com/linghu8812/tensorrt_inference/tree/master/Yolov4

[huyhoangle86]

@WongKinYiu oh Come on..... I'm using Scaled Yolov4 that was written in pytorch and .pt format I see that repo only support Darknet and .weights format.

[WongKinYiu]

just convert .pt to .weights by from models import *; convert('your.cfg', 'your.pt', 'your.weights')

[huyhoangle86]

@WongKinYiu Thank you. The convert was successfully One more question. Do you know how do I inference the model that was just converted to Onnx format?

[WongKinYiu]

i do not know how to inference onnx model.

[huyhoangle86]

thank you btw :D

[jaideep11061982]

just convert .pt to .weights by from models import *; convert('your.cfg', 'your.pt', 'your.weights')

@huyhoangle86 @WongKinYiu could u please help understand this discussion.. Onnx is some thing new for me. I am usign Pytorch , do i need any weight conversion ?

[huyhoangle86]

@jaideep11061982 No , you don't, just gg search for Onnx, try to convert if you feel interested in.

[lantudou]

@huyhoangle86 I hope it is still helpful for you even if it seems too late. https://github.com/WongKinYiu/ScaledYOLOv4/issues/371