dataset:
  name: 'mp3d'
  num_views_low_res: 12
  num_views_high_res: 1
  image_root_dir: training/mp3d_skybox
  fov: 90
  rot_low_res: 30
  rot_high_res: 10
  resolution: 256
  crop_size_high_res: 256
train:
  log_dir: high_res_upsample
  lr: 0.0001

test:
  fuse_type: diffcollage # multidiff

model:
  guidance_scale: 9.
  model_type: base
  low_res_noise_level: 20
  upsample_model:
    num_coarse_cp_blocks: 5
    lora_layers: True
    homo_cp_attn: True
    diff_timestep: 75
  base_model:
    model_id: DeepFloyd/IF-I-XL-v1.0
    lora_layers: False
    single_image_ft: False
    diff_timestep: 50