'add_gan_loss' is not in the command line for training

[sanazsab]

Hi,

why the store action is true in DRM? It does not include in the command line for training. However, FND and BND are dependent on this argument.

parser.add_argument('--add_gan_loss', action='store_true', help='if specified, use gan loss')

Thanks

[fyviezhao]

Actually, the gan loss is not used and thus this line of add_argument is redundant. FND and BND are therefore not used, either.

[sanazsab]

Actually, the gan loss is not used and thus this line of add_argument is redundant. FND and BND are therefore not used, either.

Thanks,

why they are not used? The accuracy hasn't changed with the gan loss?
And why they were added in the code?

[sanazsab]

Actually, the gan loss is not used and thus this line of add_argument is redundant. FND and BND are therefore not used, either.

Because in this case, there is no discriminator and there is just a generator. Thanks for the clarification.