magic3d-coarse-if.yaml

name: "magic3d-coarse-if"
tag: "${rmspace:${system.prompt_processor.prompt},_}"
exp_root_dir: "outputs"
seed: 0

data_type: "random-camera-datamodule"
data:
  width: 64
  height: 64
  camera_distance_range: [1.5, 2.0]
  light_sample_strategy: "magic3d"
  eval_camera_distance: 2.0
  eval_fovy_deg: 70.

system_type: "magic3d-system"
system:
  geometry_type: "implicit-volume"
  geometry:
    radius: 2.
    normal_type: analytic
    pos_encoding_config:
      otype: HashGrid
      n_levels: 16
      n_features_per_level: 2
      log2_hashmap_size: 19
      base_resolution: 16
      per_level_scale: 1.4472692374403782 # max resolution 4096
    density_bias: "blob_magic3d"
    density_activation: softplus
    density_blob_scale: 10.
    density_blob_std: 0.5
    isosurface_resolution: 128
    isosurface_threshold: auto
    isosurface_coarse_to_fine: true

  material_type: "diffuse-with-point-light-material"
  material:
    ambient_only_steps: 2001
    soft_shading: true

  background_type: "neural-environment-map-background"

  renderer_type: "nerf-volume-renderer"
  renderer:
    radius: ${system.geometry.radius}
    num_samples_per_ray: 512

  prompt_processor_type: "deep-floyd-prompt-processor"
  prompt_processor:
    pretrained_model_name_or_path: "DeepFloyd/IF-I-XL-v1.0"
    prompt: ???

  guidance_type: "deep-floyd-guidance"
  guidance:
    pretrained_model_name_or_path: "DeepFloyd/IF-I-XL-v1.0"
    weighting_strategy: uniform
    guidance_scale: 20.
    min_step_percent: 0.02
    max_step_percent: 0.98

  loggers:
    wandb:
      enable: false
      project: 'threestudio'
      name: None

  loss:
    lambda_sds: 1.
    lambda_orient: [0, 10., 1000., 5000]
    lambda_sparsity: 1.
    lambda_opaque: 0.
  optimizer:
    name: Adam
    args:
      lr: 0.01
      betas: [0.9, 0.99]
      eps: 1.e-15
    params:
      geometry:
        lr: 0.01
      background:
        lr: 0.001

trainer:
  max_steps: 10000
  log_every_n_steps: 1
  num_sanity_val_steps: 0
  val_check_interval: 200
  enable_progress_bar: true
  precision: 16-mixed

checkpoint:
  save_last: true
  save_top_k: -1
  every_n_train_steps: ${trainer.max_steps}