# How to make Zygote avoid differentiating with respect to some fields in struct

**URL:** <https://discourse.julialang.org/t/how-to-make-zygote-avoid-differentiating-with-respect-to-some-fields-in-struct/65520>\
**Category:** Machine Learning\
**Tags:** question, machine-learning, zygote, struct, autodiff\
**Created:** [July 29, 2021, 8:05pm UTC](https://discourse.julialang.org/t/how-to-make-zygote-avoid-differentiating-with-respect-to-some-fields-in-struct/65520 "2021-07-29T20:05:57Z")\
**Posts on this page:** 1\
**Showing post:** 3

<div class="post-metadata">

**Author:** ![CarloLucibello](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/carlolucibello/32/3278_2.png) [@CarloLucibello](https://discourse.julialang.org/u/CarloLucibello)\
**Post date:** [July 30, 2021, 7:05pm UTC](https://discourse.julialang.org/t/how-to-make-zygote-avoid-differentiating-with-respect-to-some-fields-in-struct/65520/3 "2021-07-30T19:05:17Z")

</div>

A workaround is to define an auxiliary function blocking the gradient:

```julia
julia> nograd(x) = x
nograd (generic function with 1 method)

julia> Zygote.@nograd nograd

julia> function loss(model)
         loss = 0.0
         for n in 1:N
           c = nograd(model.C[n]) 
           loss += sum(c*(model(X[n]) .- Y[n]).^2)
         end
         loss
       end
loss (generic function with 1 method)

julia> dmodel = gradient(loss, model)[1]
(W = [-2.6711034296853375 -0.2254551144087391 … 1.4033779589003235 3.1219721905976305; 2.5271537498909424 -0.8264030561248071 … -1.663924873993493 -1.3721983219005442], b = [1.8739815151086066, -0.8248439951554849], C = nothing)

```

---

_[View the full topic](https://discourse.julialang.org/t/how-to-make-zygote-avoid-differentiating-with-respect-to-some-fields-in-struct/65520)._