TubeR_CSN152_AVA22_train_with_backbone_init.yaml

yaml

DDP_CONFIG:
  WORLD_SIZE: 1
  WORLD_RANK: 0
  GPU_WORLD_SIZE: 8
  GPU_WORLD_RANK: 0
  # DIST_URL: 'tcp://xxx.xxx.xxx.xxx:xxxx'
  DIST_URL: 'tcp://10.164.0.3:23456'
  WOLRD_URLS: ['xxx.xxx.xxx.xxx']
  # AUTO_RANK_MATCH: True
  AUTO_RANK_MATCH: False
  DIST_BACKEND: 'nccl'
  GPU: 0
  DISTRIBUTED: True

CONFIG:
  EVAL_ONLY: False
  TWO_STREAM: False
  USE_LFB: False
  USE_LOCATION: False
  TRAIN:
    EPOCH_NUM: 20
    BATCH_SIZE: 1
    LR: 1e-4
    MIN_LR: 1e-5
    LR_BACKBONE: 1e-5
    MOMENTUM: 0.9
    W_DECAY: 1e-4
    LR_POLICY: 'step'
    USE_WARMUP: False
    WARMUP_START_LR: 1e-5
    WARMUP_EPOCHS: 4
    LR_MILESTONE: [10, 15]
    STEP: 0.1
    OPTIMIZER:
      NAME: SGD
    AUX_LOSS: True

  VAL:
    BATCH_SIZE: 1
    FREQ: 1

  DATA:
    DATASET_NAME: 'ava'
    LABEL_PATH: 'datasets/ava_action_list_v2.1_for_activitynet_2018.pbtxt'
    ANNO_PATH: 'datasets/ava_{}_v22.json'
    # DATA_PATH: 'ava/frames/{}/'
    DATA_PATH: 'datasets/ava/frames/{}'
    NUM_CLASSES: 80
    MULTIGRID: False
    IMG_SIZE: 256
    IMG_RESHAPE_SIZE: 288
    TEMP_LEN: 32
    FRAME_RATE: 2

  MODEL:
    SINGLE_FRAME: True
    BACKBONE_NAME: CSN-152
    TEMPORAL_DS_STRATEGY: decode
    LAST_STRIDE: False
    GENERATE_LFB: False
    NAME: 'ava_detr_9_224'
    ENC_LAYERS: 6
    DEC_LAYERS: 6
    D_MODEL: 256
    NHEAD: 8
    NUM_ENCODER_LAYERS: 12
    DIM_FEEDFORWARD: 2048
    QUERY_NUM: 15
    NORMALIZE_BEFORE: False
    DROPOUT: 0.1
    DS_RATE: 8
    TEMP_LEN: 32
    SAMPLE_RATE: 2
    PRETRAINED: True
    PRETRAIN_BACKBONE_DIR: "irCSN_152_ft_kinetics_from_ig65m_f126851907.mat"
    PRETRAIN_TRANSFORMER_DIR: "detr.pth"
    # PRETRAINED_PATH: "ADTR_CSN_152_decode_ava_22.pth"
    PRETRAINED_PATH: "TubeR_CSN152_AVA22.pth"
    LOAD: False
    LOAD_FC: False

  MATCHER:
    COST_CLASS: 12
    COST_BBOX: 5
    COST_GIOU: 2
    BNY_LOSS: True
    BEFORE: False

  LOSS_COFS:
    MASK_COF: 1
    DICE_COF: 12
    BBOX_COF: 5
    GIOU_COF: 2
    EOS_COF: 0.1
    WEIGHT: 10
    WEIGHT_CHANGE: 1000
    LOSS_CHANGE_COF: 2
    CLIPS_MAX_NORM: 0.1

  LOG:
    BASE_PATH: 'AVA_Tuber_train_pretrained'
    LOG_DIR: 'tb_log'
    SAVE_DIR: 'checkpoints'
    EVAL_DIR: 'AVA_Tuber_train_pretrained/eval'
    SAVE_FREQ: 1
    RES_DIR: 'tmp_train'