Multi-label classification - how to start about tabnet HOT 3 OPEN

clf = TabNetMultiTaskClassifier(
    n_steps=1,
    cat_idxs=cat_idxs,
    cat_dims=cat_dims,
    cat_emb_dim=12,
    optimizer_fn=torch.optim.Adam,
    optimizer_params=dict(lr=2e-2),
    scheduler_params={"step_size":50, "gamma":0.9},
    scheduler_fn=torch.optim.lr_scheduler.StepLR,
    mask_type='entmax',
    lambda_sparse=0, 
)

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
Cell In[61], [line 3]
      [1] max_epochs = 2
----> [3] clf.fit(
      [4]     X_train=X_train.values, y_train=y_train.values,
      [5]     max_epochs=max_epochs ,
      [6]     batch_size=1024,
      [7]     virtual_batch_size=12
      [8]     num_workers=1,
      [9]     drop_last=False,
     [10])

File pytorch_tabnet\abstract_model.py:258, in TabModel.fit(self, X_train, y_train, eval_set, eval_name, eval_metric, loss_fn, weights, max_epochs, patience, batch_size, virtual_batch_size, num_workers, drop_last, callbacks, pin_memory, from_unsupervised, warm_start, augmentations, compute_importance)
    (pytorch_tabnet/abstract_model.py:253) for epoch_idx in range(self.max_epochs):
    (pytorch_tabnet/abstract_model.py:254) 
    (pytorch_tabnet/abstract_model.py:255)     # Call method on_epoch_begin for all callbacks
    (pytorch_tabnet/abstract_model.py:256)     self._callback_container.on_epoch_begin(epoch_idx)
--> (pytorch_tabnet/abstract_model.py:258)     self._train_epoch(train_dataloader)
    (pytorch_tabnet/abstract_model.py:260)     # Apply predict epoch to all eval sets
    (pytorch_tabnet/abstract_model.py:261)     for eval_name, valid_dataloader in zip(eval_names, valid_dataloaders):

File pytorch_tabnet\abstract_model.py:489, in TabModel._train_epoch(self, train_loader)
    (pytorch_tabnet/abstract_model.py:486) for batch_idx, (X, y) in enumerate(train_loader):
    (pytorch_tabnet/abstract_model.py:487)     self._callback_container.on_batch_begin(batch_idx)
--> (pytorch_tabnet/abstract_model.py:489)     batch_logs = self._train_batch(X, y)
    (pytorch_tabnet/abstract_model.py:491)     self._callback_container.on_batch_end(batch_idx, batch_logs)
    (pytorch_tabnet/abstract_model.py:493) epoch_logs = {"lr": self._optimizer.param_groups[-1]["lr"]}

File pytorch_tabnet\abstract_model.py:527, in TabModel._train_batch(self, X, y)
    (pytorch_tabnet/abstract_model.py:524) for param in self.network.parameters():
    (pytorch_tabnet/abstract_model.py:525)     param.grad = None
--> (pytorch_tabnet/abstract_model.py:527) output, M_loss = self.network(X)
    (pytorch_tabnet/abstract_model.py:529) loss = self.compute_loss(output, y)
    (pytorch_tabnet/abstract_model.py:530) # Add the overall sparsity loss

File torch\nn\modules\module.py:1532, in Module._wrapped_call_impl(self, *args, **kwargs)
   (torch/nn/modules/module.py:1530)     return self._compiled_call_impl(*args, **kwargs)  # type: ignore[misc]
   (torch/nn/modules/module.py:1531) else:
-> (torch/nn/modules/module.py:1532)     return self._call_impl(*args, **kwargs)

File torch\nn\modules\module.py:1541, in Module._call_impl(self, *args, **kwargs)
   (torch/nn/modules/module.py:1536) # If we don't have any hooks, we want to skip the rest of the logic in
   (torch/nn/modules/module.py:1537) # this function, and just call forward.
   (torch/nn/modules/module.py:1538) if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   (torch/nn/modules/module.py:1539)         or _global_backward_pre_hooks or _global_backward_hooks
   (torch/nn/modules/module.py:1540)         or _global_forward_hooks or _global_forward_pre_hooks):
-> (torch/nn/modules/module.py:1541)     return forward_call(*args, **kwargs)
   (torch/nn/modules/module.py:1543) try:
  (torch/nn/modules/module.py:1544)     result = None

File pytorch_tabnet\tab_network.py:615, in TabNet.forward(self, x)
    (pytorch_tabnet/tab_network.py:614) def forward(self, x):
--> (pytorch_tabnet/tab_network.py:615)     x = self.embedder(x)
    (pytorch_tabnet/tab_network.py:616)     return self.tabnet(x)

File torch\nn\modules\module.py:1532, in Module._wrapped_call_impl(self, *args, **kwargs)
   (torch/nn/modules/module.py:1530)     return self._compiled_call_impl(*args, **kwargs)  # type: ignore[misc]
   (torch/nn/modules/module.py:1531) else:
-> (torch/nn/modules/module.py:1532)     return self._call_impl(*args, **kwargs)

File torch\nn\modules\module.py:1541, in Module._call_impl(self, *args, **kwargs)
   (torch/nn/modules/module.py:1536) # If we don't have any hooks, we want to skip the rest of the logic in
   (torch/nn/modules/module.py:1537) # this function, and just call forward.
   (torch/nn/modules/module.py:1538) if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   (torch/nn/modules/module.py:1539)         or _global_backward_pre_hooks or _global_backward_hooks
   (torch/nn/modules/module.py:1540)         or _global_forward_hooks or _global_forward_pre_hooks):
-> (torch/nn/modules/module.py:1541)     return forward_call(*args, **kwargs)
   (torch/nn/modules/module.py:1543) try:
   (torch/nn/modules/module.py:1544)     result = None

File pytorch_tabnet\tab_network.py:890, in EmbeddingGenerator.forward(self, x)
    (pytorch_tabnet/tab_network.py:887)         cols.append(x[:, feat_init_idx].float().view(-1, 1))
    (pytorch_tabnet/tab_network.py:888)     else:
    (pytorch_tabnet/tab_network.py:889)         cols.append(
--> (pytorch_tabnet/tab_network.py:890)             self.embeddings[cat_feat_counter](x[:, feat_init_idx].long())
    (pytorch_tabnet/tab_network.py:891)         )
    (pytorch_tabnet/tab_network.py:892)         cat_feat_counter += 1
    (pytorch_tabnet/tab_network.py:893) # concat

File torch\nn\modules\module.py:1532, in Module._wrapped_call_impl(self, *args, **kwargs)
   (torch/nn/modules/module.py:1530)     return self._compiled_call_impl(*args, **kwargs)  # type: ignore[misc]
   (torch/nn/modules/module.py:1531) else:
-> (torch/nn/modules/module.py:1532)     return self._call_impl(*args, **kwargs)

File torch\nn\modules\module.py:1541, in Module._call_impl(self, *args, **kwargs)
   (torch/nn/modules/module.py:1536) # If we don't have any hooks, we want to skip the rest of the logic in
   (torch/nn/modules/module.py:1537) # this function, and just call forward.
   (torch/nn/modules/module.py:1538) if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   (torch/nn/modules/module.py:1539)         or _global_backward_pre_hooks or _global_backward_hooks
   (torch/nn/modules/module.py:1540)         or _global_forward_hooks or _global_forward_pre_hooks):
-> (torch/nn/modules/module.py:1541)     return forward_call(*args, **kwargs)
   (torch/nn/modules/module.py:1543) try:
   (torch/nn/modules/module.py:1544)     result = None

File torch\nn\modules\sparse.py:163, in Embedding.forward(self, input)
    (torch/nn/modules/sparse.py:162) def forward(self, input: Tensor) -> Tensor:
--> (torch/nn/modules/sparse.py:163)     return F.embedding(
    (torch/nn/modules/sparse.py:164)         input, self.weight, self.padding_idx, self.max_norm,
    (torch/nn/modules/sparse.py:165)         self.norm_type, self.scale_grad_by_freq, self.sparse)

File torch\nn\functional.py:2264, in embedding(input, weight, padding_idx, max_norm, norm_type, scale_grad_by_freq, sparse)
   (torch/nn/functional.py:2258)     # Note [embedding_renorm set_grad_enabled]
   (torch/nn/functional.py:2259)     # XXX: equivalent to
   (torch/nn/functional.py:2260)     # with torch.no_grad():
   (torch/nn/functional.py:2261)     #   torch.embedding_renorm_
   (torch/nn/functional.py:2262)     # remove once script supports set_grad_enabled
   (torch/nn/functional.py:2263)     _no_grad_embedding_renorm_(weight, input, max_norm, norm_type)
-> torch/nn/functional.py:2264) return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)

IndexError: index out of range in self