Computing Two Intermediate Layers Incurs 2x Latency with Keras TF 2.3

Question

I am trying to compute the outputs of two intermediate layers with Keras in TensorFlow 2.3 and am incurring roughly 2x the latency. I was under the impression that only one forw