onnx-simplifier icon indicating copy to clipboard operation
onnx-simplifier copied to clipboard

Published 20 hours ago • verified publisher icon daquexian

Weird Question

Open andrearosasco opened this issue 2 years ago • 1 comments

Is it possible to have the opposite behaviour of onnx-simplifier? I have a torch.nn.functional.grid_sample operation that I can export to onnx using opset=16 as a GridSample node. The problem is that when I feed the onnx to tensorrt, quite unsurprisingly, the node ain't recongnized.

I'd like to split GridSample down to simpler onnx operators but I don't know if it's even possible. Does anyone have an idea on how to do that?

andrearosasco avatar Oct 10 '22 07:10 andrearosasco

If possible, I would try to split the grid_sample into multiple objects within the python code if I were in your shoes.

mchaniotakis avatar Dec 05 '22 21:12 mchaniotakis