run_title: swu_celeba_bs30
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
  segm_pl:
    weight: 30
    weights_path: ${env:TORCH_HOME}
optimizers:
  generator:
    kind: adam
    lr: 0.001
  discriminator:
    kind: adam
    lr: 0.0001
visualizer:
  key_order:
  - image
  - predicted_image
  - discr_output_fake
  - discr_output_real
  - inpainted
  rescale_keys:
  - discr_output_fake
  - discr_output_real
  kind: directory
  outdir: /group-volume/orc_srr/multimodal/e.logacheva/CelebA-HQ-inpainting/experiments/e.logacheva_2021-03-16_16-00-45_train_ablv2_work_swu_celeba_bs30/samples
location:
  data_root_dir: ${env:HOME}/CelebA-HQ
  out_root_dir: /group-volume/orc_srr/multimodal/${env:USER}/CelebA-HQ-inpainting/experiments
  tb_dir: /group-volume/orc_srr/multimodal/${env:USER}/CelebA-HQ-inpainting/tb_logs
  pretrained_models: /group-volume/orc_srr/multimodal/pretrained/
data:
  batch_size: 5
  val_batch_size: 2
  num_workers: 3
  train:
    indir: ${location.data_root_dir}/train_256
    out_size: 256
    mask_gen_kwargs:
      irregular_proba: 1
      irregular_kwargs:
        max_angle: 4
        max_len: 200
        max_width: 100
        max_times: 5
        min_times: 1
      box_proba: 1
      box_kwargs:
        margin: 10
        bbox_min_size: 30
        bbox_max_size: 150
        max_times: 4
        min_times: 1
      segm_proba: 0
    transform_variant: no_augs
    dataloader_kwargs:
      batch_size: ${data.batch_size}
      shuffle: true
      num_workers: ${data.num_workers}
  val:
    indir: ${location.data_root_dir}/val_256
    img_suffix: .png
    dataloader_kwargs:
      batch_size: ${data.val_batch_size}
      shuffle: false
      num_workers: ${data.num_workers}
  visual_test: null
generator:
  kind: pix2pixhd_global
  input_nc: 4
  output_nc: 3
  ngf: 64
  n_downsampling: 3
  n_blocks: 9
  conv_kind: default
  add_out_act: sigmoid
discriminator:
  kind: pix2pixhd_nlayer
  input_nc: 3
  ndf: 64
  n_layers: 4
evaluator:
  kind: default
  inpainted_key: inpainted
  integral_kind: ssim_fid100_f1
trainer:
  kwargs:
    gpus: -1
    distributed_backend: ddp
    max_epochs: 40
    gradient_clip_val: 1
    log_gpu_memory: None
    limit_train_batches: 25000
    val_check_interval: 2600
    log_save_interval: 250
    precision: 32
    terminate_on_nan: false
    check_val_every_n_epoch: 1
    num_sanity_val_steps: 8
    replace_sampler_ddp: false
  checkpoint_kwargs:
    verbose: true
    save_top_k: 5
    save_last: true
    period: 1
    monitor: val_ssim_fid100_f1_total_mean
    mode: max