stable-diffusion-finetune/configs/stable-diffusion/txt2img-multinode-clip-encoder-f16-256-pretraining.yaml

model:
  base_learning_rate: 1.0e-04
  target: ldm.models.diffusion.ddpm.LatentDiffusion
  params:
    linear_start: 0.001
    linear_end: 0.015
    num_timesteps_cond: 1
    log_every_t: 200
    timesteps: 1000
    first_stage_key: "jpg"
    cond_stage_key: "txt"
    image_size: 16
    channels: 16
    cond_stage_trainable: false   # Note: different from the one we trained before
    conditioning_key: crossattn
    monitor: val/loss_simple_ema
    scale_factor: 0.22765929   # magic number

    scheduler_config: # 10000 warmup steps
      target: ldm.lr_scheduler.LambdaLinearScheduler
      params:
        warm_up_steps: [ 10000 ]
        cycle_lengths: [ 10000000000000 ] # incredibly large number to prevent corner cases
        f_start: [ 1.e-6 ]
        f_max: [ 1. ]
        f_min: [ 1. ]

    unet_config:
      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
      params:
        image_size: 16    # not really needed
        in_channels: 16
        out_channels: 16
        model_channels: 320    # TODO: scale model here
        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

    first_stage_config:
      target: ldm.models.autoencoder.AutoencoderKL
      params:
        embed_dim: 16
        monitor: val/rec_loss
        ckpt_path: "models/first_stage_models/kl-f16/model.ckpt"
        ddconfig:
          double_z: True
          z_channels: 16
          resolution: 256
          in_channels: 3
          out_ch: 3
          ch: 128
          ch_mult: [ 1,1,2,2,4 ]  # num_down = len(ch_mult)-1
          num_res_blocks: 2
          attn_resolutions: [ 16 ]
          dropout: 0.0
        lossconfig:
          target: torch.nn.Identity

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


data:
  target: ldm.data.laion.WebDataModuleFromConfig
  params:
    tar_base: "pipe:aws s3 cp s3://s-datasets/laion5b/laion2B-data/"
    batch_size: 55
    num_workers: 4
    multinode: True
    min_size: 256   # TODO: experiment. Note: for 2B, images are stored at max 384 resolution
    train:
      shards: '{000000..231317}.tar -'
      shuffle: 10000
      image_key: jpg
      image_transforms:
      - target: torchvision.transforms.Resize
        params:
          size: 256
          interpolation: 3
      - target: torchvision.transforms.RandomCrop
        params:
          size: 256

    # NOTE use enough shards to avoid empty validation loops in workers
    validation:
      shards: '{231318..231349}.tar -'
      shuffle: 0
      image_key: jpg
      image_transforms:
      - target: torchvision.transforms.Resize
        params:
          size: 256
          interpolation: 3
      - target: torchvision.transforms.CenterCrop
        params:
          size: 256


lightning:
  callbacks:
    image_logger:
      target: main.ImageLogger
      params:
        batch_frequency: 5000
        max_images: 4
        increase_log_steps: False
        log_first_step: False
        log_images_kwargs:
          use_ema_scope: False
          inpaint: False
          plot_progressive_rows: False
          plot_diffusion_rows: False
          N: 4
          unconditional_guidance_scale: 3.0
          unconditional_guidance_label: [""]

  trainer:
    benchmark: True
    val_check_interval: 5000000  # really sorry
    num_sanity_val_steps: 0
    accumulate_grad_batches: 2