models/cldm_v21_512_latctrl_coltrans.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: "tgt"
    cond_stage_key: "pmt"
    control_key: "ctrl"
    image_size: 64  # 这个是默认的latent size
    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
    l_coltrans_weight: 1.0

    control_stage_config:
      target: cldm.cldm.ControlNet
      params:
        use_checkpoint: True
        image_size: 64 # unused，代码中是根据输入的 latent code 自适应的
        in_channels: 4
        hint_channels: 4
        model_channels: 320
        attention_resolutions: [ 4, 2, 1 ]
        num_res_blocks: 2
        channel_mult: [ 1, 2, 4, 4 ]
        num_head_channels: 64 # need to fix for flash-attn
        use_spatial_transformer: True
        use_linear_in_transformer: True
        transformer_depth: 1
        context_dim: 1024
        legacy: False
        latent_control: True  # control 输入形式会影响到 control stage model 的结构

    unet_config:
      target: cldm.cldm.ControlledUnetModel
      params:
        use_checkpoint: True
        image_size: 64 # unused，代码中是根据输入的 latent code 自适应的
        in_channels: 4
        out_channels: 4
        model_channels: 320
        attention_resolutions: [ 4, 2, 1 ]
        num_res_blocks: 2
        channel_mult: [ 1, 2, 4, 4 ]
        num_head_channels: 64 # need to fix for flash-attn
        use_spatial_transformer: True
        use_linear_in_transformer: True
        transformer_depth: 1
        context_dim: 1024
        legacy: False

    first_stage_config:
      target: ldm.models.autoencoder.AutoencoderKL
      params:
        embed_dim: 4
        monitor: val/rec_loss
        ddconfig:
          #attn_type: "vanilla-xformers"
          double_z: true
          z_channels: 4
          resolution: 512
          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

    cond_stage_config:
      target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder
      params:
        freeze: True
        layer: "penultimate"