config/cldm_v15_appearance_pose_local_mm.yaml

model:
  target: model_lib.ControlNet.cldm.cldm.ControlLDMReferenceOnly_Temporal_Pose_Local
  params:
    linear_start: 0.00085
    linear_end: 0.0120
    num_timesteps_cond: 1
    log_every_t: 200
    timesteps: 1000
    first_stage_key: "jpg"
    cond_stage_key: "txt"
    control_key: "hint"
    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

    appearance_control_stage_config:
      target: model_lib.ControlNet.cldm.cldm.ControlNetReferenceOnly
      params:
        image_size: 32 # unused
        in_channels: 4
        hint_channels: 3
        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

    pose_control_stage_config:
      target: model_lib.ControlNet.cldm.cldm.ControlNet
      params:
        image_size: 32 # unused
        in_channels: 4
        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

    local_pose_control_stage_config:
      target: model_lib.ControlNet.cldm.cldm.ControlNet
      params:
        image_size: 32 # unused
        in_channels: 4
        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

    unet_config:
      target: model_lib.ControlNet.cldm.cldm.ControlledUnetModelAttn_Temporal_Pose_Local
      params:
        image_size: 32 # unused
        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_heads: 8
        use_spatial_transformer: True
        transformer_depth: 1
        context_dim: 768
        use_checkpoint: True
        legacy: False

        unet_additional_kwargs:
          use_motion_module              : true
          motion_module_resolutions      : [ 1,2,4,8 ]
          unet_use_cross_frame_attention : false
          unet_use_temporal_attention    : false

          motion_module_type: Vanilla
          motion_module_kwargs:
            num_attention_heads                : 8
            num_transformer_block              : 1
            attention_block_types              : [ "Temporal_Self", "Temporal_Self" ]
            temporal_position_encoding         : true
            temporal_position_encoding_max_len : 24
            temporal_attention_dim_div         : 1
            zero_initialize                    : true

    first_stage_config:
      target: model_lib.ControlNet.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

    cond_stage_config:
      target: model_lib.ControlNet.ldm.modules.encoders.modules.FrozenCLIPEmbedder