trainer:
  target: trainer.TrainerDifIR
model:
  target: models.unet.UNetModelSwin
  ckpt_path: null
  params:
    image_size: 64
    in_channels: 3
    model_channels: 160
    out_channels: 3
    attention_resolutions:
    - 64
    - 32
    - 16
    - 8
    dropout: 0
    channel_mult:
    - 1
    - 2
    - 2
    - 4
    num_res_blocks:
    - 2
    - 2
    - 2
    - 2
    conv_resample: true
    dims: 2
    use_fp16: false
    num_head_channels: 32
    use_scale_shift_norm: true
    resblock_updown: false
    swin_depth: 2
    swin_embed_dim: 192
    window_size: 8
    mlp_ratio: 4
    cond_lq: true
    lq_size: 64
diffusion:
  target: models.script_util.create_gaussian_diffusion
  params:
    sf: 4
    schedule_name: exponential
    schedule_kwargs:
      power: 0.3
    etas_end: 0.99
    steps: 15
    min_noise_level: 0.04
    kappa: 2.0
    weighted_mse: false
    predict_type: xstart
    timestep_respacing: null
    scale_factor: 1.0
    normalize_input: true
    latent_flag: true
autoencoder:
  target: ldm.models.autoencoder.VQModelTorch
  ckpt_path: weights/autoencoder_vq_f4.pth
  use_fp16: true
  params:
    embed_dim: 3
    n_embed: 8192
    ddconfig:
      double_z: false
      z_channels: 3
      resolution: 256
      in_channels: 3
      out_ch: 3
      ch: 128
      ch_mult:
      - 1
      - 2
      - 4
      num_res_blocks: 2
      attn_resolutions: []
      dropout: 0.0
      padding_mode: zeros
degradation:
  sf: 4
  resize_prob:
  - 0.2
  - 0.7
  - 0.1
  resize_range:
  - 0.15
  - 1.5
  gaussian_noise_prob: 0.5
  noise_range:
  - 1
  - 30
  poisson_scale_range:
  - 0.05
  - 3.0
  gray_noise_prob: 0.4
  jpeg_range:
  - 30
  - 95
  second_order_prob: 0.5
  second_blur_prob: 0.8
  resize_prob2:
  - 0.3
  - 0.4
  - 0.3
  resize_range2:
  - 0.3
  - 1.2
  gaussian_noise_prob2: 0.5
  noise_range2:
  - 1
  - 25
  poisson_scale_range2:
  - 0.05
  - 2.5
  gray_noise_prob2: 0.4
  jpeg_range2:
  - 30
  - 95
  gt_size: 256
  resize_back: false
  use_sharp: false
data:
  train:
    type: realesrgan
    params:
      dir_paths: []
      txt_file_path:
      - /content/ResShift/high_res/train.txt
      im_exts:
      - JPEG
      io_backend:
        type: disk
      blur_kernel_size: 21
      kernel_list:
      - iso
      - aniso
      - generalized_iso
      - generalized_aniso
      - plateau_iso
      - plateau_aniso
      kernel_prob:
      - 0.45
      - 0.25
      - 0.12
      - 0.03
      - 0.12
      - 0.03
      sinc_prob: 0.1
      blur_sigma:
      - 0.2
      - 3.0
      betag_range:
      - 0.5
      - 4.0
      betap_range:
      - 1
      - 2.0
      blur_kernel_size2: 15
      kernel_list2:
      - iso
      - aniso
      - generalized_iso
      - generalized_aniso
      - plateau_iso
      - plateau_aniso
      kernel_prob2:
      - 0.45
      - 0.25
      - 0.12
      - 0.03
      - 0.12
      - 0.03
      sinc_prob2: 0.1
      blur_sigma2:
      - 0.2
      - 1.5
      betag_range2:
      - 0.5
      - 4.0
      betap_range2:
      - 1
      - 2.0
      final_sinc_prob: 0.8
      gt_size: 256
      crop_pad_size: 300
      use_hflip: true
      use_rot: false
      rescale_gt: true
  val:
    type: base
    params:
      dir_path: testdata/Val_SR/lq
      im_exts: png
      transform_type: default
      transform_kwargs:
        mean: 0.5
        std: 0.5
      extra_dir_path: testdata/Val_SR/gt
      extra_transform_type: default
      extra_transform_kwargs:
        mean: 0.5
        std: 0.5
      recursive: false
train:
  lr: 5.0e-05
  lr_min: 2.0e-05
  lr_schedule: null
  warmup_iterations: 100
  batch:
  - 8
  - 1
  microbatch: 1
  num_workers: 4
  prefetch_factor: 2
  weight_decay: 0
  ema_rate: 0.999
  iterations: 1000
  save_freq: 10000
  log_freq:
  - 200
  - 2000
  - 1
  local_logging: true
  tf_logging: false
  use_ema_val: true
  val_freq: ${train.save_freq}
  val_y_channel: true
  val_resolution: ${model.params.lq_size}
  val_padding_mode: reflect
  use_amp: true
  seed: 123456
  global_seeding: false
  compile:
    flag: false
    mode: reduce-overhead
save_dir: logging/
resume: ''
cfg_path: configs/realsr_swinunet_realesrgan256.yaml

Number of parameters: 118.59M
Restoring autoencoder from weights/autoencoder_vq_f4.pth
Number of images in train data set: 1254
Number of images in val data set: 32
Train: 000200/001000, Loss/MSE: t(1):1.6e-01/1.6e-01, t(8):4.5e-01/4.5e-01, t(15):5.9e-01/5.9e-01, lr:5.00e-05
Train: 000400/001000, Loss/MSE: t(1):2.8e-02/2.8e-02, t(8):3.9e-01/3.9e-01, t(15):5.0e-01/5.0e-01, lr:5.00e-05
Train: 000600/001000, Loss/MSE: t(1):2.1e-02/2.1e-02, t(8):3.4e-01/3.4e-01, t(15):4.6e-01/4.6e-01, lr:5.00e-05
Train: 000800/001000, Loss/MSE: t(1):1.4e-02/1.4e-02, t(8):3.5e-01/3.5e-01, t(15):5.1e-01/5.1e-01, lr:5.00e-05
Train: 001000/001000, Loss/MSE: t(1):1.4e-02/1.4e-02, t(8):2.9e-01/2.9e-01, t(15):4.6e-01/4.6e-01, lr:5.00e-05