InformationBottleneck

class torchelie.nn.InformationBottleneck(in_channels, num_latents, strength=1, kl_reduction='mean')
training: bool