Hello everyone, I’m trying to convert my code so that it would run on GPU, but I still encountered some issues. My code is mainly using DiffEqFlux, with some array reshaping required inside the Neural ODE function itself. When the the code gets to the sciml_train part, I always get this same error…

CuArray supports bits type error when performing automatic differentiation in sciml_train (possibly due to array reshaping)

ChrisRackauckas July 27, 2021, 4:58pm 2

Yes, ReverseDiff doesn’t not support GPUs. Use ZygoteVJP (or maybe EnzymeVJP)

Topic		Replies	Views
Code using Flux slow on GPU GPU flux	9	3119	November 6, 2019
Flux gpu gradient failing General Usage question	5	882	August 18, 2020
Zygote: CuArray only supports bits types General Usage	6	1349	October 25, 2020
DiffEqFlux neural_ode used with Flux.Train! is slower on GPU than CPU GPU flux , ode	0	738	December 31, 2019
CuArray and Optim GPU optim , optimization	9	3424	September 4, 2018