lama/configs/training/big-lama-celeba.yaml

run_title: ''

training_model:
  kind: default
  visualize_each_iters: 1000
  concat_mask: true
  store_discr_outputs_for_vis: true
losses:
  l1:
    weight_missing: 0
    weight_known: 10
  perceptual:
    weight: 0
  adversarial:
    kind: r1
    weight: 10
    gp_coef: 0.001
    mask_as_fake_target: true
    allow_scale_mask: true
  feature_matching:
    weight: 100
  resnet_pl:
    weight: 30
    weights_path: ${env:TORCH_HOME}

generator:
  kind: ffc_resnet
  input_nc: 4
  output_nc: 3
  ngf: 64
  n_downsampling: 3
  n_blocks: 18
  add_out_act: sigmoid
  init_conv_kwargs:
    ratio_gin: 0
    ratio_gout: 0
    enable_lfu: false
  downsample_conv_kwargs:
    ratio_gin: ${generator.init_conv_kwargs.ratio_gout}
    ratio_gout: ${generator.downsample_conv_kwargs.ratio_gin}
    enable_lfu: false
  resnet_conv_kwargs:
    ratio_gin: 0.75
    ratio_gout: ${generator.resnet_conv_kwargs.ratio_gin}
    enable_lfu: false

defaults:
  - location: celeba
  - data: abl-04-256-mh-dist-celeba
  - discriminator: pix2pixhd_nlayer
  - optimizers: default_optimizers
  - visualizer: directory
  - evaluator: default_inpainted
  - trainer: any_gpu_large_ssim_ddp_final_celeba
  - hydra: overrides