Onnx opset missmatch (opset 17 only supported in pytorch>=2)

The README says the code requires pytorch>=1.7 but the `onnx_model_example.ipynb` expected `opset=17`. Is far as I can tell `opset 17` is only supported in `pytorch>=2`.

1. Should the readme be updated? Or
2. Can I just set `opset_version=16` in the [notebook](https://github.com/facebookresearch/segment-anything/blob/main/notebooks/onnx_model_example.ipynb)?

```python
with open(onnx_model_path, "wb") as f:
torch.onnx.export(
onnx_model,
tuple(dummy_inputs.values()),
f,
export_params=True,
verbose=False,
opset_version=17, # <---- :-(
do_constant_folding=True,
input_names=list(dummy_inputs.keys()),
output_names=output_names,
dynamic_axes=dynamic_axes,
)
```

Happy to make the changes and make a PR with any desired solution.

Start by comparing the PyTorch requirement in README with the export cell in notebooks/onnx_model_example.ipynb, especially its opset_version=17 setting. Verify which opset works with the stated supported versions, then update the README or notebook so the documented requirement and example agree and the notebook export completes.

Written by the indexing model from the issue text.