Toy 8 Modes error

[fabdan]

Executing toy_8-modes.ipynb Jupyter sample code raises error:

"TypeError: tuple indices must be integers or slices, not tuple"
on code:
" l = lambd_predict * loss_fit(output[:, ndim_z:], y[:, ndim_z:])"

"output" variable seems to be a tuple o tensors instead of a tensor.

[Guasong]

Hi fabdan,

I think they updated the package so in the new version, the model gives two output instead of only one. Therefore, the variable "output" will be assigned with two outputs, which makes it a tuple. To fix the problem, you need to use

"output, _ = model(x)" to replace "output = model(x)"

Try to do this replacement whenever call of model(x) appears in this notebook (there are 5 in total). Also, change the line with "pred_c" to be

    pred_c, _ = model(torch.cat((x_samps, torch.zeros(N_samp, ndim_tot - ndim_x)),
                             dim=1).to(device))
    pred_c = pred_c.data[:, -8:].argmax(dim=1)

Then it should work.

[1DreamCollector]

@Guasong Thank you ,you are right.