This post was temporarily hidden by the community for possibly being off-topic, inappropriate, or spammy.
How is this different from No gradient is calculated during Flux training with ODE?
Thanks for the reply.
Yes, they are the same issue. There was no reply in that post, so I simplify the code and post here, requesting for more possible attention.
I’d be grateful if you could kindly provide any insight. This problem has bothered me for a while. Thanks
I have further simplified the code: The tracked parameters are not updated during Flux.training
I’ve also requested to delete this post.
It would probably be easier to open one issue to track all of these simplifications.
Let’s get an issue open and I’m going to ping Flux developers for some help.