Hi all, I’m running into a problem when trying to compute gradients of an ODE solution on GPU (with Metal.jl) using SciMLSensitivity ( SciMLSensitivity: Automatic Differentiation and Adjoints for (Differential) Equation Solvers · SciMLSensitivity.jl ). Here’s a minimal working example: using Random…

oh, aparently you need to set abstol and reltol as Float32 also. We should make those be of utype also… Edit: it looks like we already do this. I’m very confused as to what’s going wrong…

SciMLSensitivity fails on GPU?

Specific Domains Machine Learning

Oscar_Smith September 22, 2025, 5:40pm 2

What is the stack trace associated with the InvalidIRError? There might be a place in OrdinaryDiffEq where we use a Float64 by accident.

1 Like

Topic		Replies	Views
Trouble with NeuralODE and Metal Numerics question , sciml	3	432	December 28, 2023
Is this impossible for Zygote Adjoint to handle General Usage zygote	4	613	April 7, 2021
sciML_train func running but not optimizing New to Julia diffeq , flux , optimization , sciml	5	560	April 14, 2021
Simple SciML case - core dumped after switching from julia 1.7 to 1.8 General Usage question	7	470	September 21, 2022
DiffEqGPU.jl with CUDA: Error computing gradients through SDE solver Machine Learning	12	326	January 10, 2024

SciMLSensitivity fails on GPU?

Related topics