Observing backward pass while running inference.py with a pb file, which is consuming lot of time

[shrutiramesh1988]

Command Tried: python inference.py --model pretrained/horse2zebra.pb --input=./datasets/horse2zebra/testA --output=./output

Downloaded the pb file from : https://github.com/vanhuyz/CycleGAN-TensorFlow/releases/download/v0.1-alpha/horse2zebra.pb

Issue and Observations: Low throughput observed for inference because backward_data is seen while running inference.py. How do we run it as forward_only to get better throughput?

gemm:blas,backward_data