Apply :code:`function()` to :code:`getattr(module, name)` on each forward pass. Allows to implement things such as weight normalization,

Multiply :code:`getattr(module, name)` by :code:`scale` on forward pass as a hook. Used to implement equalized LR for StyleGAN

A packed block with Masked Conv-BN-ReLU Args: in_ch (int): input channels out_ch (int): output channels ks (int): ke

A packed block with Masked Conv-Norm-ReLU Args: in_ch (int): input channels out_ch (int): output channels ks (int):

A Classification recipe with a default froward training / testing pass using cross entropy and mixup, and extended with AdaBelief and Red

Build the discriminator for StyleGAN2 Args: input_sz (int): image size max_ch (int): maximum number of chann

StyleGAN2 Recipe distributed with DistributedDataParallel Args: G (nn.Module): a Generator. D (nn.Module): a Discriminator.

Add a bilinear upsampling layer before the conv that doubles the spatial size

Sets the learning rate to the initial LR decayed by 10 every 30 epochs

Applies transforms on x Args: x (torch.Tensor): input Returns: transformed x

Manipulate an array of logit predictions so that binary prediction is not a special case anymore. Outputs `preds` as (Batch size, Num classes

Build a attention56 network Args: num_classes (int): number of classes in_ch (int): number of channels in the images

Patch Discriminator from pix2pix

Change brightness by a factor alpha Args: alpha (float): scale factor

Conditions the layer before the forward pass if z will not be present when calling forward Args: z (2D tensor, o

Conditions the layer before the forward pass if z will not be present when calling forward Args: z (2D tensor, o

Conditions the layer before the forward pass if z will not be present when calling forward Args: z (2D tensor, o

Scale contrast by factor alpha Args: alpha (float): scale factor

initialize a NCCL process group with default port / address. For internal use.