hog/train.yml

data:
  base_dir: /work/narus/HOGraspNet
  base_path: /work/narus/HOGraspNet
  data_root: /work/narus
  dataset: hog
  dataset_path: /work/narus/HOGraspNet
  gt_joints_path: /work/narus/FreiHAND/evaluation_xyz.json
  mixed:
  - name: hog
    path: /work/narus/HOGraspNet
    weight: 1.0
  - name: ho3d
    path: /work/narus/HO3D/train
    weight: 1.0
  - name: freihand
    path: /work/narus/FreiHAND
    weight: 1.0
  - name: dexycb_sampled
    path: /work/narus/DEXYCB_sampled
    weight: 1.0
  num_workers: 4
  setup: s1
device: cuda
diffusion:
  beta_end: 0.01
  beta_schedule: linear
  beta_start: 0.0001
  diffuse_space: original
  num_diffusion_timesteps: 1000
model:
  coords_dim: 45
  default:
    coords_dim: 45
    custom_pe: false
    downsample: 1
    dropout: 0.25
    hamer_cond_prob: 0.2
    hid_dim: 2056
    n_head: 4
    num_layer: 7
    skips:
    - 1
    - 3
    - 5
    - 6
    unconditinal: true
    zero_mask_token: false
  ema: false
  ema_rate: 0.999
  emb_dim: 512
  hid_dim: 2056
  text_feature: distilbert
  uniaff:
    add_img_pe: true
    coords_dim: 45
    custom_pe: false
    dropout: 0.25
    hid_dim: 2056
    mask_by_sample_img: true
    mask_by_sample_txt: true
    n_head: 4
    num_layer: 1
    zero_mask_token: false
  var_type: fixedsmall
optim:
  amsgrad: false
  decay: 20
  eps: 1.0e-08
  grad_clip: 1.0
  lr: 0.001
  lr_gamma: 0.9
  optimizer: Adam
save_samples:
  anti_pen: true
  caption_index: gt_taxonomy_summary_strict
  diffusion_space: original
  image_cond: true
  skip: 10
  skip_type: uniform
  test_num_diffusion_timesteps: 1000
  test_times: 1
  test_timesteps: 100
  vis: true
  vis_batch: 2500
  vis_epoch: 4
testing:
  anti_pen: true
  skip: 10
  test_num_diffusion_timesteps: 1000
  test_times: 1
  test_timesteps: 5
  vis: true
  vis_batch: 2500
  vis_epoch: 4
training:
  batch_size: 512
  custom_pe: false
  decay: 60
  desc_simple: true
  hamer_cond: false
  image_cond: true
  image_feat_type: hamer
  lr: 5.0e-05
  lr_gamma: 0.9
  mano_grad: true
  n_epochs: 1000
  n_iters: 5000000
  normalize: false
  reg_mean: true
  reg_var: true
  snapshot_freq: 5000
  text_cond_drop_prob: 0.1
  text_mask_prob: 0.1
  unconditinal: false
  use_all_captions: true
  validation_freq: 2000
  vis: true
  vis_batch: 2500
  vit_feat_drop_prob: 0.3