Neftune computation is probably wrong with packed training

### System Info v4.46.2 ### Who can help? _No response_ ### Information - [ ] The official example scripts - [ ] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...) - [ ] My own task or dataset (give details below) ### Reproduction Neftune computation is probably wrong with packed training because the scaling factor is `alpha/sqrt(lenght*d)`. The length is the packed length there: https://github.com/huggingface/transformers/blob/a06a0d12636756352494b99b5b264ac9955bc735/src/transformers/trainer_utils.py#L126-L149 ### Expected behavior Should take into account the size of each sentence during computation.