final_config.yaml

GENERAL:
  OUTPUT_DIR: "./model_output/"
  LOG_LEVEL: "INFO"  # Options: DEBUG, INFO, WARNING, ERROR
  USE_NESTED_TENSORS: true

MODEL:
  HIDDEN_DIM: 256
  NUM_HEADS: 8
  NUM_BLOCKS: 12
  HEAD_DEPTH: 3
  MAX_SEQ_LEN: 40
  DROPOUT_RATE: 0.1
  ATTENTION_DROPOUT_RATE: 0.1
  DROPPATH_RATE: 0.1
  ROPE_FREQ_SEQ: 100
  ROPE_FREQ_SPAT: 500
  PREDICTION_HEADS: [
    "center",
    # "mask",
    "visibility",
    # "progress"
  ]
  MINI_BATCH_SIZE: -1
  LAYERNORM_TYPE: "RMSNorm"

TRAINING:
  MODE: "trajectory" # or "query"
  TRAJ_MULTIPLIER: 8
  MULTI_GPU: false
  AUGMENT_ROTATION: false
  AUGMENT_DINO: false
  FRAME_MASK_PROB: 0.0
  TEMPORAL_ROPE_JITTER: false
  JITTER_RANGE: 2.0
  EPOCHS: 40
  VIS_FREQ: -1
  BATCH_SIZE: 8
  OPTIM:
    OPTIMIZER: "muon"
    LR_SCHEDULER: "warmupcosine"
    LR: 5e-4
    WEIGHT_DECAY: 0.05
    GRADIENT_CLIP_NORM: 1.0
  LOSS_WEIGHTS:
    IOU: 1.0
    BCE: 1.0
    DICE: 0.0
    VISIBILITY: 1.0
    FOCAL: 1.0
    CENTER_MSE: 10.0
    DISTANCES_MSE: 6.0
  #USE_CWD: true

DATASET:
  DATA_PATH: ["/media/data_0/dataset_progpre/", "/media/data_1/dataset_tg2/", "/media/data_1/coda_pre/", "/media/data_1/botanic_pre/", "/media/data_1/bonn_pre/"]
  # DATA_PATH: ["/media/data_1/dataset_tg2/"]
  DATASET_NAME: ['habitat', 'tartan', 'coda', 'botanic', 'bonn']
  # DATASET_NAME: ['tartan']
  REAL_DATA_PATH: "/media/data_0/dataset_gs/"
  # DATA_PATH: "/scratch/local/ds_zipped/"
  TYPE: "pre"
  HAS_FEATS: true
  IS_ZIPPED: true
  NUM_WORKERS: 4 # 6