[Debug] pytorch导出onnx模型时不支持roll操作

BUG:

RuntimeError: Exporting the operator roll to ONNX opset version 11 is not supported. Please open a bug to request ONNX export support for the missing operator.

参考方案:

https://github.com/pytorch/pytorch/blob/d39790340db916e128b2b637cd12f4616fddb87d/torch/onnx/symbolic_opset9.py#L3261

解决方案：

官方已解决此问题，将下面代码添加进opset版本对应的文件即可，如

conda/lib/python3.6/site-packages/torch/onnx/symbolic_opset11.py

@parse_args('v', 'is', 'is')
def roll(g, self, shifts, dims):
    assert len(shifts) == len(dims)

    result = self
    for i in range(len(shifts)):
        shapes = []
        shape = sym_help._slice_helper(g,
                                       result,
                                       axes=[dims[i]],
                                       starts=[-shifts[i]],
                                       ends=[maxsize])
        shapes.append(shape)
        shape = sym_help._slice_helper(g,
                                       result,
                                       axes=[dims[i]],
                                       starts=[0],
                                       ends=[-shifts[i]])
        shapes.append(shape)
        result = g.op("Concat", *shapes, axis_i=dims[i])

    return result