Tensor dimension error on running example LSTM autoencoders

[fox-ds]

When running the examples from the LSTM autoencoder section (e.g. https://github.com/online-ml/river-torch/blob/master/docs/examples/anomaly/example_lstm_autoencoder.ipynb) we are getting the error:

    input = torch.cat((h, x_flipped), dim=self.time_axis)
RuntimeError: Tensors must have same number of dimensions: got 3 and 2

This is probably due to the line

h = h[-1].view(1, 1, -1)

which transforms the 2-dimensional tensor h into a three-dimensional tensor which cannot be concatenated with x_flipped. A potential fix would be to change the line to

h = h[-1].view(1, 1, -1)[0]

Maybe this issue is also present in other autoencoder examples, and we don't know for sure if this is the proper fix, hence the issue.

[kulbachcedric]

Hi @fox-ds, thanks for the hint (and sorry for the late reply). We'll fix it :-).

Best Cedric