Is it possible to start TensorFlow from an intermediate layer?

Hello,

I was wondering

Lets say if I get the output of an intermediary layer, would it be possible to feed the data back into the intermediary layer and resume the processing only from that layer?

Just wondering…..