# Gradients from an SDE solver with callbacks

**URL:** <https://discourse.julialang.org/t/gradients-from-an-sde-solver-with-callbacks/46317>\
**Category:** Machine Learning\
**Tags:** question, sde\
**Created:** [September 9, 2020, 10:10am UTC](https://discourse.julialang.org/t/gradients-from-an-sde-solver-with-callbacks/46317 "2020-09-09T10:10:27Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![MikeKlocCZ](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mikekloccz/32/15357_2.png) [@MikeKlocCZ](https://discourse.julialang.org/u/MikeKlocCZ)\
**Post date:** [September 9, 2020, 10:10am UTC](https://discourse.julialang.org/t/gradients-from-an-sde-solver-with-callbacks/46317/1 "2020-09-09T10:10:27Z")

</div>

I am trying to get gradients of the solution of an SDE where a callback function is used in the `solve` command. It seems that the forward adjoint method is not compatible with the usage of callbacks as I get `"ERROR: type Pairs has no field callback"`. Or is there a way how to make it work?

Here is a simple example derived from the Lotka-Volterra tutorials

```julia
using DifferentialEquations, Flux, DiffEqFlux
using DiffEqSensitivity

function dt!(du, u, p, t)
  x, y = u
  α, β, δ, γ = p
  du[1] = dx = α*x - β*x*y
  du[2] = dy = -δ*y + γ*x*y
end

function dW!(du, u, p, t)
  du[1] = 0.1u[1]
  du[2] = 0.1u[2]
end

u0 = [1.0,1.0]
tspan = (0.0, 10.0)
p = [2.2, 1.0, 2.0, 0.4]
prob_sde = SDEProblem(dt!, dW!, u0, tspan,p)

condition(u,t,integrator) = integrator.t >9.0 #some condition
function affect!(integrator)
	 println("Callback") #some callback
end
cb = DiscreteCallback(condition,affect!,save_positions=(false,false))

function predict_sde(p)
  return Array(solve(prob_sde, EM(), saveat = 0.1,sensealg = ForwardDiffSensitivity(), dt=0.001, callback=cb))
 end

loss_sde(p)= sum(abs2, x-1 for x in predict_sde(p))

ps = Flux.params(p)
@time gs = gradient(ps) do
	loss_sde(p)
end #ERROR: type Pairs has no field callback

```

---

<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:** [September 11, 2020, 4:15am UTC](https://discourse.julialang.org/t/gradients-from-an-sde-solver-with-callbacks/46317/2 "2020-09-11T04:15:22Z")

</div>

It was a bug. Fixed in [https://github.com/SciML/DiffEqSensitivity.jl/pull/333](https://github.com/SciML/DiffEqSensitivity.jl/pull/333) and will get released soon.

---

<div class="post-metadata">

**Author:** ![MikeKlocCZ](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mikekloccz/32/15357_2.png) [@MikeKlocCZ](https://discourse.julialang.org/u/MikeKlocCZ)\
**Post date:** [September 11, 2020, 7:49am UTC](https://discourse.julialang.org/t/gradients-from-an-sde-solver-with-callbacks/46317/3 "2020-09-11T07:49:03Z")

</div>

Thank you!