class torchelie.models.AutoGAN(arch: List[int], n_skip_max: int = 2, in_noise: int = 256, out_ch: int = 3, batchnorm_in_output: bool = False)

Experimental: Generator discovered in AutoGAN: Neural Architecture Search for Generative Adversarial Networks.

  • arch (list) – architecture specification: a list of output channel for each block. Each block doubles the resolution of the generated image. Example: [512, 256, 128, 64, 32].

  • n_skip_max (int) – how many blocks far back will be used for the skip connections maximum.

  • in_noise (int) – dimension of the input noise vector

  • out_ch (int) – number of channels on the image

  • batchnorm_in_output (bool) – whether to have a batchnorm just before projecting to RGB. I have found it better on False, but the official AutoGAN repo has it.


AutoGAN() is experimental, and may change or be deleted soon if not already broken

blocks: torch.nn.modules.container.ModuleList
make_noise: torch.nn.modules.module.Module
n_skip_max: int
to_rgb: torch.nn.modules.container.Sequential