configs/VITON.yaml

model:
  target: cldm.cldm.ControlLDM
  params:
    linear_start: 0.00085
    linear_end: 0.0120
    num_timesteps_cond: 1
    log_every_t: 200
    timesteps: 1000
    first_stage_key: "image"
    first_stage_key_cond: ["agn", "agn_mask", "image_densepose"]
    cond_stage_key: "cloth"
    control_key: "cloth"
    image_size: 64
    channels: 4
    cond_stage_trainable: False
    conditioning_key: crossattn
    monitor: val/loss_simple_ema
    scale_factor: 0.18215
    use_ema: False
    only_mid_control: False
    use_VAEdownsample: True
    use_lastzc: True 
    use_imageCLIP: True
    use_pbe_weight: True
    u_cond_percent: 0.2
    use_attn_mask: False
    mask1_key: "agn_mask"
    mask2_key: "cloth_mask"

    control_stage_config:
      target: cldm.warping_cldm_network.NoZeroConvControlNet
      params:
        image_size: 32
        in_channels: 13
        hint_channels: 3
        model_channels: 320
        attention_resolutions: [ 4, 2, 1 ]
        num_res_blocks: 2
        channel_mult: [ 1, 2, 4, 4 ]
        num_heads: 8
        use_spatial_transformer: True
        transformer_depth: 1
        context_dim: 768
        use_checkpoint: True
        legacy: False
        cond_first_ch: 4  

    unet_config:
      target: cldm.warping_cldm_network.StableVITON
      params:
        image_size: 32
        in_channels: 13
        out_channels: 4
        model_channels: 320
        attention_resolutions: [ 4, 2, 1 ]
        num_res_blocks: 2
        channel_mult: [ 1, 2, 4, 4 ]
        num_heads: 8
        use_spatial_transformer: True
        transformer_depth: 1
        context_dim: 768
        use_checkpoint: True
        legacy: False
        dim_head_denorm: 1

    first_stage_config:
      target: ldm.models.autoencoder.AutoencoderKL
      params:
        embed_dim: 4
        monitor: val/rec_loss
        ddconfig:
          double_z: true
          z_channels: 4
          resolution: 256
          in_channels: 3
          out_ch: 3
          ch: 128
          ch_mult:
          - 1
          - 2
          - 4
          - 4
          num_res_blocks: 2
          attn_resolutions: []
          dropout: 0.0
        lossconfig:
          target: torch.nn.Identity
    validation_config:
      ddim_steps: 50
      eta: 0.0
      scale: 1.0

    cond_stage_config:
      target: ldm.modules.image_encoders.modules.FrozenCLIPImageEmbedder
dataset_name: VITONHDDataset
resume_path: ./pretrained_models/VITONHD_PBE_pose.ckpt
default_prompt: ""
log_images_kwargs:
  unconditional_guidance_scale: 5.0