model:
  model_type: mamba
  embed_dim: 512
  depth: 24
  predictor_depth: 2
  drop_path_rate: 0.1
  bimamba_type: v2
  if_devide_out: true
  mixer_type: mamba
  momentum: 0.996
  final_momentum: 0.9999
  norm_target: true
data:
  train_list: data/lists/pretrain_train.txt
  val_list: data/lists/pretrain_val.txt
  batch_size: 4
  num_workers: 8
  T_prime: 30
  tau_seconds: 6.0
  default_tr: null
training:
  epochs: 30
  lr: 5.0e-4
  weight_decay: 0.05
  warmup_epochs: 3
  mask_ratio: 0.65
  grad_clip: 1.0
  grad_accumulation_steps: 4
  seed: 42
  use_amp: true
logging:
  log_interval: 20
  checkpoint_dir: ./checkpoints/pretrain/example
  log_dir: ./logs/pretrain/example