Toby Roseman

Here is a minimal example to reproduce the issue: ```python import torch import coremltools as ct class Net(torch.nn.Module): def forward(self, x): return x + x x = torch.rand(1, 3, 256,...

The [PyTorch documentation for this op](https://pytorch.org/vision/main/generated/torchvision.ops.deform_conv2d.html) doesn't contains a lot of details. The [PyTorch forward implementation for `deform_conv2d`](https://github.com/pytorch/vision/blob/b1123cfd543d35d282de9bb28067e48ebec18afe/torchvision/csrc/ops/cpu/deform_conv2d_kernel.cpp#L867-L1086) seems quite complex. Can someone share mathematical formulas for what this operation...

Thanks @Volutionn for the concise information. So the `deform_conv2d` PyTorch op uses just the "Deformable Convolution" formula, is that correct?

We don't support Python 3.12 yet. That is the root cause here. Related to #1846

@snej - Right at the top of the document you reference, it says: ``` The current version of coremltools includes wheels for Python 3.7, 3.8, 3.9, 3.10, and 3.11. ```

> This was addressed in #2170 @YifanShenSZ There are several other things that need to be addressed before we can support Python 3.12. At a minimum, we still need to...

Thanks for the self contained example code. I can reproduce your issue. Our support for untraced PyTorch models is only "experimental". You should have received an error message saying that....

`upsample_bicubic2d` [is now supported](https://github.com/apple/coremltools/blob/eda9f8ed220c5cfb58873bc00c735579ae33e89a/coremltools/converters/mil/frontend/torch/ops.py#L2886).

@darrenxyli - you're correct. Sorry for the confusion. Reopening this issue.

@yohanessatria - I am unable to run your code. Please add the relevant import statements.