PyTorch copy_ operation

[vidursatija]

🌱 Describe your Feature Request

• A clear and concise description of what the problem is.
  Lack of copy_ operation conversion from PyTorch to CoreML
• CoreML / iOS version you are using?
  4.0b3
• Are you interested in contributing?
  Yes

Use cases

• Please describe the use cases
  Copying tensors for general machine learning use cases
• Please provide examples

a = b.copy_()

OR

a = b

Describe alternatives you've considered

A clear and concise description of any alternative solutions or features you've considered.
Reshaping the source array because the reshape operation is implemented.

Additional context

Add any other context or screenshots about the feature request here.

I've used the following piece of code to convert copy operations:

@register_torch_op
def copy_(context, node):
    inputs = _get_inputs(context, node, expected=3)
    x = inputs[0]
    output = mb.reshape(x=x, shape=x.shape, name=node.name)
    context.add(output)

[vidursatija]

Does that solution work? I can send a pull request.

[1duo]

Does that solution work? I can send a pull request.

You're more than welcome to put up a fix! mb.reshape seems unnecessary, we have mb.identity().

[vidursatija]

Yes. Trying it and sending a PR

[1duo]

PR Merged. Thanks @vidursatija!