about the g_loss and the d_loss

At the training stage,the g_loss is not convergent,always increases or decreases.And the d_loss always near the value of zero at the beginning. How I adjust it?What‘s more，what are the requirements for parameter of the batch_size?

The issue names no files, tests, or entry points. Start by locating the training configuration and documentation for g_loss, d_loss, and batch_size, then compare the reported behavior with the repository's documented expectations; done would require a confirmed diagnosis and a clearly scoped change or explanation.

Written by the indexing model from the issue text.