# ChainRulesCore: Custom adjoint ignored in certain cases

**URL:** <https://discourse.julialang.org/t/chainrulescore-custom-adjoint-ignored-in-certain-cases/67814>\
**Category:** Machine Learning\
**Tags:** adjoint, autodiff, chainrulescore\
**Created:** [September 7, 2021, 1:01pm UTC](https://discourse.julialang.org/t/chainrulescore-custom-adjoint-ignored-in-certain-cases/67814 "2021-09-07T13:01:35Z")\
**Posts on this page:** 1\
**Showing post:** 3

<div class="post-metadata">

**Author:** ![Rasmus\_Hoier](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/rasmus_hoier/32/24036_2.png) [@Rasmus\_Hoier](https://discourse.julialang.org/u/Rasmus_Hoier)\
**Post date:** [September 7, 2021, 2:21pm UTC](https://discourse.julialang.org/t/chainrulescore-custom-adjoint-ignored-in-certain-cases/67814/3 "2021-09-07T14:21:10Z")

</div>

Thanks!  
Pretty neat that the call to gradient checks if forward mode will be beneficial! 🙂 I guess we had implicitly made the assumption that the gradient function would always use reverse mode, which is not really justified.

BTW: I was a bit confused about why use ForwardDiff instead of ChainRulesCore.frule. Found an answer to this [here](https://discourse.julialang.org/t/chainrulescore-and-forwarddiff/61705).

---

_[View the full topic](https://discourse.julialang.org/t/chainrulescore-custom-adjoint-ignored-in-certain-cases/67814)._