mmagic.models.editors.biggan.biggan_discriminator
¶
Module Contents¶
Classes¶
BigGAN Discriminator. The implementation refers to |
- class mmagic.models.editors.biggan.biggan_discriminator.BigGANDiscriminator(input_scale, num_classes=0, in_channels=3, out_channels=1, base_channels=96, sn_eps=1e-06, sn_style='ajbrock', act_cfg=dict(type='ReLU'), with_spectral_norm=True, blocks_cfg=dict(type='BigGANDiscResBlock'), arch_cfg=None, init_cfg=dict(type='ortho'))[source]¶
Bases:
mmengine.model.BaseModule
BigGAN Discriminator. The implementation refers to https://github.com/ajbrock/BigGAN-PyTorch/blob/master/BigGAN.py # noqa.
In BigGAN, we use a SAGAN-based architecture composing of an self-attention block and number of convolutional residual blocks with spectral normalization.
More details can be found in: Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR2019).
The design of the model structure is highly corresponding to the output resolution. For the original BigGAN’s generator, you can set
output_scale
as you need and use the default value ofarch_cfg
andblocks_cfg
. If you want to customize the model, you can set the arguments in this way:arch_cfg
: Config for the architecture of this generator. You can refer the_default_arch_cfgs
in the_get_default_arch_cfg
function to see the format of thearch_cfg
. Basically, you need to provide information of each block such as the numbers of input and output channels, whether to perform upsampling, etc.blocks_cfg
: Config for the convolution block. You can replace the block type to your registered customized block and adjust block params here. However, you should notice that some params are shared among these blocks likeact_cfg
,with_spectral_norm
,sn_eps
, etc.- Parameters
input_scale (int) – The scale of the input image.
num_classes (int, optional) – The number of conditional classes. Defaults to 0.
in_channels (int, optional) – The channel number of the input image. Defaults to 3.
out_channels (int, optional) – The channel number of the final output. Defaults to 1.
base_channels (int, optional) – The basic channel number of the discriminator. The other layers contains channels based on this number. Defaults to 96.
sn_eps (float, optional) – Epsilon value for spectral normalization. Defaults to 1e-6.
sn_style (str, optional) – The style of spectral normalization. If set to ajbrock, implementation by ajbrock(https://github.com/ajbrock/BigGAN-PyTorch/blob/master/layers.py) will be adopted. If set to torch, implementation by PyTorch will be adopted. Defaults to ajbrock.
act_cfg (dict, optional) – Config for the activation layer. Defaults to dict(type=’ReLU’).
with_spectral_norm (bool, optional) – Whether to use spectral normalization. Defaults to True.
blocks_cfg (dict, optional) – Config for the convolution block. Defaults to dict(type=’BigGANDiscResBlock’).
arch_cfg (dict, optional) – Config for the architecture of this discriminator. Defaults to None.
init_cfg (dict, optional) – Initialization config dict. If type is Pretrained, the pretrain model will be loaded. Otherwise, type will be parsed as the name of initialization method. Support values are ‘ortho’, ‘N02’, ‘xavier’. Defaults to dict(type=’ortho’).