# Why are Neural SDEs trained with TrackerAdjoint()?

**URL:** <https://discourse.julialang.org/t/why-are-neural-sdes-trained-with-trackeradjoint/113514>\
**Category:** Machine Learning\
**Tags:** question, sde, zygote\
**Created:** [April 25, 2024, 6:47pm UTC](https://discourse.julialang.org/t/why-are-neural-sdes-trained-with-trackeradjoint/113514 "2024-04-25T18:47:33Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![agelgazzar](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/agelgazzar/32/207319_2.png) [@agelgazzar](https://discourse.julialang.org/u/agelgazzar)\
**Post date:** [April 25, 2024, 6:47pm UTC](https://discourse.julialang.org/t/why-are-neural-sdes-trained-with-trackeradjoint/113514/1 "2024-04-25T18:47:33Z")

</div>

I thought that Tracker was mostly replaced by Zygote and Enzyme for Reverse-mode Autodiff in the sciml ecosystem.  
Why is it still being used for computing the sensitivities in DiffeqFlux?

---

<div class="post-metadata">

**Author:** ![ChrisRackauckas](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/chrisrackauckas/32/77_2.png) [@ChrisRackauckas](https://discourse.julialang.org/u/ChrisRackauckas)\
**Post date:** [April 29, 2024, 5:39am UTC](https://discourse.julialang.org/t/why-are-neural-sdes-trained-with-trackeradjoint/113514/2 "2024-04-29T05:39:31Z")

</div>

Just limitations with higher order AD in Zygote, though those issues were actually fixed a few years ago so it could probably update its default. Also some of @avikpal 's Lux changes for higher order autodiff might come into play here too. So we probably just need to reassess it.