training on a new dataset

[ihavetoomanyquestions]

i want to try out these GANs for creating a new pokemon (similar to https://github.com/moxiegushi/pokeGAN but with other GANs).

currently i am doing the following method, but i dont know if it would work or can it have unwanted effects?

basically, i get the data from https://github.com/moxiegushi/pokeGAN/tree/master/data and rename it as data/celebA.

then i change the n_examples in config.py to number of samples in my data.

and then i run convert.py to create the tfrecord formatted files.

since it is very expensive for me to run the method on AWS GPUs, i am not able to do a test run.. but is it possible to tell me if the above approach would work?? or is there other things i should change, too?