We are working with pipelines where we have to convert pre-existing torch.jit models to downstream AI serving formats (e.g. ONNX, ExecuTorch). The current torch.export.export API does not support these, so instead we use torch._export.converter.TS2EPConverter.
We need to add support for dynamic shapes; and internally, the converter computes dynamic shapes. So it seems like this could be as simple as exposing dynamic_shapes in the converter’s API.