Question about fine-tuning for text classification

Thank you for the exceptional work! I have a query concerning the fine-tuning process for text classification.
I've been following the code provided in the 'examples/finetune_modernbert_on_glue.ipynb' notebook, but I've encountered some perplexing results. Could you please provide some guidance?

Here are more details:
1. I noticed that the var 'wd' wasn't being utilized, so I included it for weight decay, but this has led to confusing results too. You can see the screenshot of the error here:
2. I received a TypeError stating that ModernBertForSequenceClassification.__init__() received an unexpected keyword argument 'compile'. Here's the code snippet where the error occurred:
```python
hf_model = AutoModelForSequenceClassification.from_pretrained(
checkpoint, num_labels=n_labels, id2label=id2label, label2id=label2id, compile=False
)
```
Could you please provide guidance on how to address these issues? Thank you in advance for your help!

Start with examples/finetune_modernbert_on_glue.ipynb and reproduce the reported weight-decay and compile errors, including the AutoModelForSequenceClassification.from_pretrained call. Compare the notebook's handling of wd and model arguments, then confirm the documented fine-tuning path runs without those errors.

Written by the indexing model from the issue text.