torchvision Resize export doesn't support antialias=True, ::_upsample_bicubic2d_aa to ONNX opset version 16 is not supported

[tobaiMS]

trafficstars

🐛 Describe the bug

'''python

resize = torchvision.transforms.Resize(size=(224, 224), interpolation=torchvision.transforms.InterpolationMode.BICUBIC, max_size=None, antialias=True)

dummy_input = torch.randn(1,3,2000,2000)
torch.onnx.export(resize, dummy_input, preprocess_model_path, input_names=['image'], output_names=['processed_image'], dynamic_axes={
                  # dict value: manually named axes
                  "image": {2:'h', 3:'w'},
              })

'''

error "torch.onnx.symbolic_registry.UnsupportedOperatorError: Exporting the operator ::_upsample_bicubic2d_aa to ONNX opset version 16 is not supported."

Versions

ONNX opset version 16 is not supported

cc @neginraoof @BowenBao

[tobaiMS]

seems it's available since opset 18: onnx/Operators.md at main · onnx/onnx (github.com) , can we support it?

[datumbox]

@BowenBao could you offer your thoughts here?

[tobaiMS]

Hi, @datumbox @BowenBao , want to follow up on this thread :)

[BowenBao]

Hi @tobaiMS thanks for reporting. we are blocked on https://github.com/pytorch/pytorch/pull/83201 moving ONNX to opset18 inside PyTorch. Once that is merged, this issue can be attempted by using the opset18 Resize. Tracking this issue in our project board.

[thiagocrepaldi]

#83201 was merged. feel free to retry and reopen the issue if the issue persists