I tried to update the parameters in ODE by Flux training. In Struct V203, only 2 parameters are selected to be tracked for backpropagation: Flux.trainable(device::V2O3) = (device.Rth, device.Cth) During training, these 2 parameters are indeed tracked, but their values are not updated. No error pop …

No gradient is calculated during Flux training with ODE

ChrisRackauckas July 4, 2020, 7:37am 2

1 Like

Topic		Replies	Views
Request to be deleted General Usage diffeq , flux	4	725	July 4, 2020
The tracked parameters are not updated during Flux.training General Usage diffeq , flux	2	784	July 6, 2020
DiffEqFlux: how do declare a system of ODE which is backpropagatable? Machine Learning diffeq	2	661	February 28, 2019
Optimizing parameters of ODE in DiffEqFlux: Not implemented: convert tracked Tracker.TrackedReal{Float64} to tracked Float64 Machine Learning diffeq , flux , optimization	4	1450	March 31, 2019
Taking gradients when propagating through an ODE solver Machine Learning question	6	622	June 3, 2020