Sequence classification at the word level

I am working on Morph analysis problem. The input is a sentene and the model should predict morphological labels (features ) for each word. Consider the given format of my dataset.

Word / (features)
I / (0,0,0,1) Love / (1,1,0,1) Pizza / (1,1,0,1).

I want to use existing transformer model and fine tune it for the above task.

The Automodelforsequenceclassification model takes the whole sentence and then it predicts labels. Can I modify it or configure it so that it works at the word level ?

Is there any other alternative ?