# Understanding Token classification head

**URL:** <https://discuss.huggingface.co/t/understanding-token-classification-head/32743>\
**Category:** Beginners\
**Created:** [February 28, 2023, 8:14pm UTC](https://discuss.huggingface.co/t/understanding-token-classification-head/32743 "2023-02-28T20:14:15Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![ChaitanyaC](https://avatars.discourse-cdn.com/v4/letter/c/2acd7d/32.png) [@ChaitanyaC](https://discuss.huggingface.co/u/ChaitanyaC)\
**Post date:** [February 28, 2023, 8:14pm UTC](https://discuss.huggingface.co/t/understanding-token-classification-head/32743/1 "2023-02-28T20:14:15Z")

</div>

I was trying to understand how AutoModels work in hugging face. There is a proper description for BertForTokenClassification’s classification heads but not for the generic AutoModelForTokenClassificaton. I wanna know what is this this token classification head so that I can use it directly or write a custom model with a required classification head

---

<div class="post-metadata">

**Author:** ![ChaitanyaC](https://avatars.discourse-cdn.com/v4/letter/c/2acd7d/32.png) [@ChaitanyaC](https://discuss.huggingface.co/u/ChaitanyaC)\
**Post date:** [April 4, 2023, 12:14pm UTC](https://discuss.huggingface.co/t/understanding-token-classification-head/32743/2 "2023-04-04T12:14:58Z")

</div>

Posted this on [Stack overflow](https://stackoverflow.com/questions/75890430/what-is-the-classification-head-of-a-hugging-face-automodelfortokenclassificatio/75908709#75908709) and it’s just a Dropout layer and a linear layer attached to it. Refer to the link for more details on how to check this through code.
