dataloader: 
  batch_size : 16
  shuffle: true
  num_workers : 64
  drop_last : true
  pin_memory : true
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: "jpg"
    cond_stage_key: "txt"
    control_key: "hint"
    image_size: 64
    channels: 100
    cond_stage_trainable: true
    # conditioning_key: crossattn
    monitor: val/loss_simple_ema
    scale_factor: 0.18215
    use_ema: False
    only_mid_control: False

    # control_stage_config:
    #   target: cldm.cldm.ControlNet
    #   params:
    #     image_size: 32 # unused
    #     in_channels: 100
    #     hint_channels: 768
    #     model_channels: 128
    #     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
    refer_config:
      target: cldm.cldm.ReferenceNet
      params:
        image_size: 32 # unused
        hint_in_channels: 1024
        hint_out_channels: 128
        in_channels: 100
        out_channels: 100
        model_channels: 1024
        attention_resolutions: [ 4, 2, 1 ]
        num_res_blocks: 1
        channel_mult: [ 1, 1 ]
        num_heads: 8
        use_spatial_transformer: True
        transformer_depth: 1
        context_dim: 512
        use_checkpoint: True
        dims: 1
        legacy: False


    unet_config:
      target: tortoise_model.DiffusionTts
      params:
        model_channels: 512
        num_layers: 8
        in_channels: 100
        in_latent_channels: 1024
        out_channels: 100
        dropout: 0
        use_fp16: False
        num_heads: 16
        layer_drop: .1
        unconditioned_percentage: .1
      # target: cldm.cldm.ControlledUnetModel
      # params:
      #   image_size: 32 # unused
      #   hint_in_channels: 1024
      #   hint_out_channels: 128
      #   in_channels: 100
      #   out_channels: 100
      #   model_channels: 1024
      #   attention_resolutions: [ 4, 2, 1 ]
      #   num_res_blocks: 1
      #   resblock_updown: True
      #   channel_mult: [ 1, 1]
      #   num_heads: 8
      #   use_spatial_transformer: True
      #   transformer_depth: 1
      #   context_dim: 512
      #   use_checkpoint: True
      #   dims: 1
      #   legacy: False

    cond_stage_config:
      target: cldm.cond_emb.CLIP
      params: 
        embed_dim: 512
        vision_cfg: 
          layers: 6
          width: 512
          head_width: 64
          mlp_ratio: 4.0
          patch_dropout: 0.4
          attentional_pool: False
          patch_size: 64
          image_size: 1000
          in_channels: 100
          pool_type: 'tok'
          pos_embed_type: 'learnable'
          final_ln_after_pool: false

train:
  train_batch_size : 32
  gradient_accumulate_every : 1
  train_lr : 0.0001
  train_num_steps : 1000000
  ema_update_every : 10
  ema_decay : 0.995
  adam_betas : [0.9, 0.99]
  save_and_sample_every : 1000
  timesteps : 1000
  sampling_timesteps : 1000
  results_folder : "results"
  logs_folder  : "ttts/AA_diffusion/logs"
  num_workers : 32
  eps : 0.000000001
  keep_ckpts : 3
  all_in_mem : false
dataset:
  path : "/home/hyc/tortoise_plus_zh/ttts/datasets/databaker_data.jsonl"
  gpt_path : "/home/hyc/tortoise_plus_zh/ttts/gpt/logs/2023-12-24-14-22-14/model-70.pt"