zeronvs_config.yaml

model:
  base_learning_rate: 0.0001
  target: ldm.models.diffusion.ddpm.LatentDiffusion
  params:
    linear_start: 0.00085
    linear_end: 0.012
    num_timesteps_cond: 1
    log_every_t: 200
    timesteps: 1000
    first_stage_key: image_target
    cond_stage_key: image_cond
    image_size: 32
    channels: 4
    cond_stage_trainable: false
    conditioning_key: hybrid
    monitor: val/loss_simple_ema
    scale_factor: 0.18215
    conditioning_config:
      params:
        mode: 7dof_quantile_scale
        embedding_dim: 19
        depth_model_name: midas
    scheduler_config:
      target: ldm.lr_scheduler.LambdaLinearScheduler
      params:
        warm_up_steps:
        - 100
        cycle_lengths:
        - 10000000000000
        f_start:
        - 1.0e-06
        f_max:
        - 1.0
        f_min:
        - 1.0
    unet_config:
      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
      params:
        image_size: 32
        in_channels: 8
        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
    eval_config:
      params:
        scale: 3.0
        ddim_steps: 100
        ddim_eta: 1.0
        lpips_model_path: null
    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
    cond_stage_config:
      target: ldm.modules.encoders.modules.FrozenCLIPImageEmbedder
data:
  target: ldm.data.simple.WDSGenericDataModule
  params:
    train_config:
      batch_size: 48
      num_workers: 6
      shuffle_buffer_size: 500
      prefetch_factor: 4
      dataset_config_1:
        dataset_n_shards: 127
        dataset_name: co3d
        views_per_scene: 100
        dataset_n_scenes: 18432
        rate: 0.025
        probability: 0.34
        dataset_url: null
      dataset_config_2:
        dataset_n_shards: 127
        dataset_name: re10k
        views_per_scene: 200
        dataset_n_scenes: 65280
        probability: 0.33
        rate: 0.025
        dataset_url: null
      dataset_config_3:
        dataset_n_shards: 127
        dataset_name: acid
        views_per_scene: 100
        dataset_n_scenes: 12032
        probability: 0.33
        rate: 0.025
        dataset_url: null
    val_config:
      batch_size: 1
      subsample: 1.0
      scene_scale: 1.0
      dataset_n_shards: 1
      dataset_name: co3d
      dataset_n_scenes: 150
      num_workers: 1
      shuffle_buffer_size: 20
      rate: 0.1
      dataset_url: null
--lightning:
  trainer:
    accumulate_grad_batches: 4
  modelcheckpoint:
    params:
      every_n_train_steps: 2500
--data:
  params:
    train_config:
      batch_size: 48
    val_config:
      batch_size: 1
--model:
  params:
    conditioning_config:
      params:
        mode: 7dof_quantile_scale
        embedding_dim: 19
    eval_config:
      params:
        ddim_steps: 100
  base_learning_rate: 0.0001
--args:
  finetune_from: null