Model did not return a loss --- but why?

🤗Transformers
1

I’m trying to fine-tune a DistilBert model on a new dataset and when I attempt to train I get this:

ValueError: The model did not return a loss from the inputs, only the following keys: last_hidden_state. For reference, the inputs it received are input_ids,attention_mask.

Why??

I’ve followed Huggingface tutorials all the way along.
As far as I can tell, my dataset is in the right shape:

DatasetDict({
    train: Dataset({
        features: ['text', 'labels'],
        num_rows: 44330
    })
    test: Dataset({
        features: ['text', 'labels'],
        num_rows: 11083
    })
})

After passing it through tokenizer = DistilBertTokenizer.from_pretrained(‘distilbert-base-cased’), it looks like this:

DatasetDict({
    train: Dataset({
        features: ['text', 'labels', 'input_ids', 'attention_mask'],
        num_rows: 44330
    })
    test: Dataset({
        features: ['text', 'labels', 'input_ids', 'attention_mask'],
        num_rows: 11083
    })
})

I cannot find any solutions, or even guidance as to how to debug this, anywhere.

A little help would be awesome.