Weird error in training

[ianni67]

Hi, while running the example pynb, I got the following error:

======== Epoch 1 / 3 ========
Training...

╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ in <module>:89                                                                                   │
│                                                                                                  │
│    86 │   │   # calculate the average loss at the end. `loss` is a Tensor containing a           │
│    87 │   │   # single value; the `.item()` function just returns the Python value               │
│    88 │   │   # from the tensor.                                                                 │
│ ❱  89 │   │   total_train_loss += loss.item()                                                    │
│    90 │   │                                                                                      │
│    91 │   │   # Perform a backward pass to calculate the gradients.                              │
│    92 │   │   loss.backward()                                                                    │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
AttributeError: 'str' object has no attribute 'item'

Looks like something is wrong, the model is not producing a correct output.
Any hint to solve this?

[parrrthhh]

I'm getting the same issue