Closed JunaidMB closed 2 years ago
Hello @JunaidMB
My run of your code with the github version of the package do not raise any issue :
[Epoch 001] Loss: 1.145955
[Epoch 001] Loss: 1.350600
[Epoch 001] Loss: 1.083805
[Epoch 001] Loss: 1.049350
[Epoch 001] Loss: 1.270870
[Epoch 001] Loss: 1.425476
[Epoch 001] Loss: 0.762844
[Epoch 001] Loss: 1.036789
...
[Epoch 001] Loss: 0.926028
[Epoch 001] Loss: 0.965075
[Epoch 001] Loss: 1.117391
> res
# Tuning results
# 5-fold cross-validation
# A tibble: 5 × 5
splits id .order .metrics .notes
<list> <chr> <int> <list> <list>
1 <split [419/105]> Fold1 3 <tibble [8 × 8]> <tibble [0 × 1]>
2 <split [419/105]> Fold3 1 <tibble [8 × 8]> <tibble [0 × 1]>
3 <split [419/105]> Fold4 2 <tibble [8 × 8]> <tibble [0 × 1]>
4 <split [419/105]> Fold2 4 <tibble [8 × 8]> <tibble [0 × 1]>
5 <split [420/104]> Fold5 5 <tibble [8 × 8]> <tibble [0 × 1]>
Could you try latest github version of the package ?
Hi @cregouby
Using the latest github version of the package has made it work! Thank you very much, I have a full workflow using the tidymodels here (for reference):
https://github.com/JunaidMB/playing_with_tabnet/blob/master/tabnet_nhs_stranded.R
I would consider this issue closed!
I have got the same error message when fitting with torch
.
And after repeat loading and detaching loaded packages, I have found that the error is related with future
packages (furrr
, doFuture
and etc.).
In my case, the error will occur when set plan(multisession)
, and if you set plan(sequential)
, the example will work properly.
You might set parallel somewhere else before running the example.
hope this could help.
Hello,
I'm following the Tabnet tutorial found here with a different and significantly smaller dataset and I'm having some issues related to the hyperparameter tuning. I'm getting the following error:
When I go into the
.notes
column I see the following error message:I'm not sure why this error is occurring, if it was related to the size or column types of the dataset I would expect there to be different error messages, however this seems like something related to how tabnet is interacting with tidymodels?
The full code to reproduce the error is:
Any help would be much appreciated. Thanks!