Help for design a neurnet

[tranquocde]

hi, I have question like that: I want to design a DL model with a feature as follow: In forward(), the input goes through layers, but I want the layers stop whenever the strop-trigger value from output of previous layer is satisfied (if the iteration exceeds maximum value then just stop and get the output , but I want to stop early if I can). Can I design this features by the loop inside forward()