CycleGAN one side training and testing

[heidarinejad]

I have a dataset A and B and as an ultimate goal want to transform A group images into B group form using CycleGAN. for training only have trainA and trainB folders in the dataset folder using the below code: $ python3 train.py --dataroot ./datasets --name maps_cyclegan --model cycle_gan

and for the test, I provide both testA and testB folders inside the dataset folder (still I do not know should I provide testB or not) and then run: $ python3 test.py --dataroot ./datasets --name maps_cyclegan --model test --no_dropout

but as a one-side train and test, the results are almost the same as testA images and not getting close to testB images. what is wrong with my steps? do I run the training script correctly?

[MystericQ]

after tarining, the out put imgs in /checkpoints,how did the fake_B looks like? try python3 test.py --dataroot ./datasets/testA --