Cannot convert torchvision models that have multi-dimensional dictionary outputs (even with a wrapper).

[diegogranziol]

🐞Describing the bug

• Make sure to only create an issue here for bugs in the coremltools Python package. If this is a bug with the Core ML Framework or Xcode, please submit your bug here: https://developer.apple.com/bug-reporting/
• Provide a clear and consise description of the bug. Cannot convert torchvision models that have a complicated output, even if a wrapper is used to make the output into a tensor)

  Stack Trace

• If applicable, please paste the complete stack trace. /usr/local/lib/python3.8/dist-packages/coremltools/converters/mil/mil/types/type_tensor.py in val(self, v) 86 def val(self, v): 87 if not isinstance(v, np.ndarray): ---> 88 raise ValueError( 89 "tensor should have value of type ndarray, got {} instead".format( 90 type(v)

ValueError: tensor should have value of type ndarray, got <class 'numpy.float32'> instead

To Reproduce

• Please add a minimal code example that can reproduce the error when running it.

  
  import torchvision
  import torch
  import torch.nn as nn
  import coremltools as ct
  model = torchvision.models.detection.maskrcnn_resnet50_fpn()
  image = torch.randn(1,3,400,400)

class Wrapped(nn.Module): def init(self): super(Wrapped, self).init() self.model = model.eval() def forward(self, x): res = self.model(x)

Extract the tensor we want from the output dictionary

    x = res[0]['boxes']
    return x

new_model = Wrapped().eval() traced_model = torch.jit.trace(new_model,image) mlmodel = ct.convert(traced_model, inputs=[ct.TensorType(shape=[1,3, 400, 400])], source="pytorch", convert_to="mlprogram")

- If the model conversion succeeds, but there is a numerical mismatch in predictions, please include the code used for comparisons.

## System environment (please complete the following information):
 - coremltools version: 6.1
 - OS (e.g. MacOS version or Linux type): MacOS and Linux
 - Any other relevant version information (e.g. PyTorch or TensorFlow version): pytorch==1.12.1 torchvision==0.13.1 

## Additional context
- Add anything else about the problem here that you want to share.

[TobyRoseman]

With pytorch==1.12.1 and torchvision==0.13.1, I get an error when calling into torch.jit.trace before ct.convert is even called. I think this is an issue with torch or with this model. I don't think this is a coremltools error.