Adjusting parameters makes training results better

I use DCGAN training some data, Modified some parameters, I got some bad results，E.g：AdamOptimizer parameters epsilon=1 or replace the optimizer with tf.train.RMSPropOptimizer , I got some bad results. I want to get better results. How should I modify it?
Thank you.

The issue mentions DCGAN training, AdamOptimizer with epsilon=1, and tf.train.RMSPropOptimizer, but names no files or tests. Start by locating the training entry point and reproducing the reported parameter changes against the baseline; the work is complete only when a specific, reproducible configuration and its improved results are established.

Written by the indexing model from the issue text.