RuntimeError: Exporting the operator repeat_interleave to ONNX opset version 11 is not supported.

在将pth转onnx过程中，报错：RuntimeError: Exporting the operator repeat_interleave to ONNX opset version 11 is not supported.

由于repeat_interleave函数是将张量中的元素按照某一维度复制n次，比如torch.repeat_interleave(x, k, dim=2), 因此可以用repeat和view函数代替进行实现。

（如有问题，请指正!）

例如：

import torch

x = torch.randn(2, 1, 4, 1)
print(x)
print("--------------------------------------------------")

y1 = torch.repeat_interleave(x, 2, dim=2)
print(y1)
print("--------------------------------------------------")

y2 = x.repeat(1, 1, 1, 2).view(x.shape[0], x.shape[1], -1, x.shape[3])
print(y2)

得到的结果是：

tensor([[[[-1.9781],
          [-1.3968],
          [-0.6847],
          [ 1.1089]]],


        [[[-0.5929],
          [ 1.7383],
          [-0.6736],
          [-1.2737]]]])

------------------------------------------------------------------------

tensor([[[[-1.9781],
          [-1.9781],
          [-1.3968],
          [-1.3968],
          [-0.6847],
          [-0.6847],
          [ 1.1089],
          [ 1.1089]]],


        [[[-0.5929],
          [-0.5929],
          [ 1.7383],
          [ 1.7383],
          [-0.6736],
          [-0.6736],
          [-1.2737],
          [-1.2737]]]])

------------------------------------------------------------------------

tensor([[[[-1.9781],
          [-1.9781],
          [-1.3968],
          [-1.3968],
          [-0.6847],
          [-0.6847],
          [ 1.1089],
          [ 1.1089]]],


        [[[-0.5929],
          [-0.5929],
          [ 1.7383],
          [ 1.7383],
          [-0.6736],
          [-0.6736],
          [-1.2737],
          [-1.2737]]]])