def create_model(options): latent_size = 0 discriminator = apply_spectral_norm(Discriminator()) #discriminator = Discriminator() generator = Generator() lr_base = 0.0002 #* 10 #* 0.1 model = Gan( discriminator=discriminator, generator=generator, latent_size=latent_size, optimizer_discriminator_fn=functools.partial(optimizer_fn, lr=lr_base), # TTUR optimizer_generator_fn=functools.partial(optimizer_fn, lr=lr_base), real_image_from_batch_fn=get_image, image_pool=GanDataPool(100)) model.apply(init_weights) return model
def create_model(): latent_size = 64 generator = Generator(latent_size=latent_size) discriminator = Discriminator() optimizer_fn = functools.partial(torch.optim.Adam, lr=0.001, betas=(0.5, 0.999)) model = Gan(discriminator=discriminator, generator=generator, latent_size=latent_size, optimizer_discriminator_fn=optimizer_fn, optimizer_generator_fn=optimizer_fn, real_image_from_batch_fn=get_image) return model
def create_model(): latent_size = 0 #netD = apply_spectral_norm(Discriminator()) discriminator = Discriminator() generator = Generator() lr_base = 0.0002 model = Gan( discriminator=discriminator, generator=generator, latent_size=latent_size, optimizer_discriminator_fn=functools.partial(optimizer_fn, lr=lr_base), # TTUR optimizer_generator_fn=functools.partial(optimizer_fn, lr=lr_base), real_image_from_batch_fn=get_image, ) return model