I am optimizing a simple ODE problem with DiffEqFlux to get started with Julia. This code worked on TensorFlow. In the following code, running main(S=5, B=256) works, but main(S=7, B=256) fails. I previously used EnsembleProblem with same code and got same results. The error begins: ERROR: Need an…

DiffEqFlux gradient error: breaks with increasing size of problem, size should not affect autodiff

ChrisRackauckas November 17, 2021, 12:32am 8

Looks like a bug in the map adjoint.

Topic		Replies	Views
Using Flux: gradient on DifferentialEquations: solve results in an error New to Julia question , diffeq , differentiation	11	1003	May 8, 2023
Getting gradients for DiffEq with single derivative Optimization (Mathematical)	3	808	June 17, 2020
Flux.gradient throws MethodError even though loss is evaluated just fine New to Julia question , diffeq	3	557	April 16, 2022
Problems using BacksolveAdjoint for an SDE with DiffEqFlux Machine Learning sde	5	699	September 29, 2021
Calculation of gradient not working for an ODEproblem that uses 2D or 3D arrays General Usage diffeq , flux	1	510	July 24, 2020