Open AdrianEddy opened 3 months ago
@AdrianEddy Thank you for great works! I try to inference with your code. but ,
Traceback (most recent call last):
File "main_stereo.py", line 612, in
@ylab604 Please check now
I see i did change the mask function when i check your code(yesterday). But, important thing is that onnx graph(netron)is not normal campare with pinto0309
Anyway thank you for your kindness. And i will also update the result of excution
What do you mean it's not normal? What's weird about it?
What do you mean it's not normal? What's weird about it?
This means that if converted to onnx or jit, the inference output will be different from the original torch model.
This PR adds types to function signatures to be able to use
torch.jit
ortorch.onnx.export()
. I also had to convert some functions to modulesThe code should be equivalent to the previous one, I verified that with inference (I didn't test training though)
It's easiest to review this without whitespace diff
Related to #29