diffusion:
  timesteps: 1000
  schedule_name: cosine
  enforce_zero_terminal_snr: true
  schedule_params:
    beta_start: 0.0001
    beta_end: 0.02
    cosine_s: 0.008
  timestep_respacing: null
  mean_type: VELOCITY
  var_type: LEARNED_RANGE
  loss_type: MSE

optimizer:
  lr: 0.00001
  type: bkh_pytorch_utils.Lion

validation:
  classifier_cond_scale: 4
  protocol: DDPM
  log_original: true
  log_concat: true
  cls_log_indices: [0, 1, 2, 3]

model:
  input_size: 256
  dims: 2
  attention_resolutions: [8, 16, 32]
  channel_mult: [1, 1, 2, 2, 4, 4]
  dropout: 0.0
  in_channels: 2
  out_channels: 2
  model_channels: 128
  num_head_channels: -1
  num_heads: 4
  num_heads_upsample: -1
  num_res_blocks: [2, 2, 2, 2, 2, 2]
  resblock_updown: false
  use_checkpoint: false
  use_new_attention_order: false
  use_scale_shift_norm: true
  scale_skip_connection: false
  
  # conditions
  num_classes: 772
  # num_classes: 4
  concat_channels: 1
  guidance_drop_prob: 0.1
  missing_class_value: null