Onnx output and pytorch output are inconsistent

[SeanLiu081]

Thanks for open-sourcing this great work. I have encountered a problem now. When I convert the model to onnx and tensorrt, the landmark result of trt/onnx is different from the result of pytorch model. I find the problem is self.conv_window in onnx，you can see the result before self.conv_window and after self.conv_window. I print the crops[0, :5, :5] before self.conv_window[0] of torch and onnx，they are same,


After self.conv_window[0]，the visual_fts[0,:5,:5] are different，

[SeanLiu081]

the onnx/trt result :

the pytorch result:

[edmosRovi]

Hello, I'm interested in how to convert the modelo to ONNX formato. I have try it but my ONNX modelo nota works, fails in the execution. Please could you tell me how do you get the ONNX model??
I opened an issue here but nobody answer me yet:
#4

[john09282922]

Hi, can u share onnx file?

[edmosRovi]

Sure, I leave a link with two onnx I did when I tested it, one using opset16 and the other opset 17. Any help you can give me to make this network work in onnx will be welcome. Thank you very much.

https://we.tl/t-AEaFfgzM7P

[john09282922]

Thanks for open-sourcing this great work. I have encountered a problem now. When I convert the model to onnx and tensorrt, the landmark result of trt/onnx is different from the result of pytorch model. I find the problem is self.conv_window in onnx，you can see the result before self.conv_window and after self.conv_window. I print the crops[0, :5, :5] before self.conv_window[0] of torch and onnx，they are same, After self.conv_window[0]，the visual_fts[0,:5,:5] are different，

Hi, sean, Did u solve this problem? would you mind sharing onnx file??

[lokeycookie]

@SeanLiu081
is it possible for you to share what steps you took to generate the onnx files?