# How to re-decompose operator in Inductor pass

**URL:** <https://dev-discuss.pytorch.org/t/how-to-re-decompose-operator-in-inductor-pass/1706>\
**Category:** compiler\
**Created:** [December 6, 2023, 9:50am UTC](https://dev-discuss.pytorch.org/t/how-to-re-decompose-operator-in-inductor-pass/1706 "2023-12-06T09:50:01Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![xingfeng02](https://yyz2.discourse-cdn.com/flex036/user_avatar/dev-discuss.pytorch.org/xingfeng02/32/1646_2.png) [@xingfeng02](https://dev-discuss.pytorch.org/u/xingfeng02)\
**Post date:** [December 6, 2023, 9:50am UTC](https://dev-discuss.pytorch.org/t/how-to-re-decompose-operator-in-inductor-pass/1706/1 "2023-12-06T09:50:01Z")

</div>

I want to add a pass re-decompose operators in Inductor, I am wondering how to do it:

- set “decomposition=None” in compile\_fx(), the forward graph and backward graph traced are not decomposed.
- add a pass in post\_grad\_passes(), which can re-decompose the operators using decomposition\_table.
