vqgan/model.yaml

model:
  base_learning_rate: 4.5e-06
  target: models.taming.models.vqgan.VQSegmentationModel
  params:
    monitor: miou
    embed_dim: 256
    n_embed: 32
    image_key: segmentation
    n_labels: 20
    ddconfig:
      double_z: false
      z_channels: 256
      resolution: 128
      in_channels: 20
      out_ch: 20
      ch: 128
      ch_mult:
      - 1
      - 1
      - 2
      - 2
      - 4
      num_res_blocks: 2
      attn_resolutions:
      - 16
      dropout: 0.0
    lossconfig:
      target: models.taming.modules.losses.segmentation.BCELossWithQuant
      params:
        codebook_weight: 1.0