Hi! K. Frank
Thank you for your decent advice, it helps a lot.
I should’ve remembered the operation matrix * matrix doesn’t break the chain rule of differentiation
I rewrote my model and loss function to make them differentiable and ready to run state according to your series of advice, and I succeeded to “execute” my training without an error!
This is a huge advancement for me (plz allow me to say thank you again), though, I faced another problem (it might be a easy peasy to fix for knowledgeable person though) which is,
the weights of the model doesn’t upgraded through its learning over epochs.
This must be another topic, so I posted a new topic The parameters of the model with custom loss function doesn’t upgraded thorough its learning over epochs - discuss.pytorch.org
Thank you!
passiveradio