mvit.yaml

LOG_MODEL_INFO: False
TRAIN:
  DATASET: ava
  BATCH_SIZE: 16
  BATCH_SIZE: 8

  EVAL_PERIOD: 2
  CHECKPOINT_PERIOD: 1
  AUTO_RESUME: True
  CHECKPOINT_EPOCH_RESET: True
  CHECKPOINT_IN_INIT: True
  CHECKPOINT_FILE_PATH: ""
  CHECKPOINT_TYPE: pytorch
  CHECKPOINT_EPOCH_RESET: True

DATA:
  USE_OFFSET_SAMPLING: True
  DECODING_BACKEND: torchvision
  NUM_FRAMES: 40
  SAMPLING_RATE: 3
  TRAIN_JITTER_SCALES: [356, 446]
  TRAIN_CROP_SIZE: 312
  TEST_CROP_SIZE: 312 # use if TEST.NUM_SPATIAL_CROPS: 1
  INPUT_CHANNEL_NUM: [3]
  PATH_TO_DATA_DIR: ""

  TRAIN_JITTER_SCALES_RELATIVE: [0.08, 1.0]
  TRAIN_JITTER_ASPECT_RELATIVE: [0.75, 1.3333]
  MEAN: [0.485, 0.456, 0.406]
  STD: [0.229, 0.224, 0.225]
MVIT:
  ZERO_DECAY_POS_CLS: False
  SEP_POS_EMBED: True
  DEPTH: 24
  NUM_HEADS: 1
  EMBED_DIM: 96
  PATCH_KERNEL: (3, 7, 7)
  PATCH_STRIDE: (2, 4, 4)
  PATCH_PADDING: (1, 3, 3)
  MLP_RATIO: 4.0
  QKV_BIAS: True
  DROPPATH_RATE: 0.0
  NORM: "layernorm"
  EMBED_DIM: 144
  NUM_HEADS: 2
  DEPTH: 48 # [2, 6, 36, 2]
  DIM_MUL: [[2, 2.0], [8, 2.0], [44, 2.0]]
  HEAD_MUL: [[2, 2.0], [8, 2.0], [44, 2.0]]
  POOL_Q_STRIDE: [[2, 1, 2, 2], [8, 1, 2, 2], [44, 1, 2, 2]]
  DROPPATH_RATE: 0.0

  POOL_KV_STRIDE_ADAPTIVE: [1, 8, 8]
  POOL_KVQ_KERNEL: [3, 3, 3]
  USE_ABS_POS: False # default: True
  REL_POS_SPATIAL: True # default: false
  REL_POS_TEMPORAL: True # default: false
  MODE: "conv_unshared"

  POOL_Q_STRIDE: [[0, 1, 1, 1], [1, 1, 1, 1], [2, 1, 2, 2], [3, 1, 1, 1], [4, 1, 1, 1], [5, 1, 1, 1], [6, 1, 1, 1], [7, 1, 1, 1], [8, 1, 2, 2], [9, 1, 1, 1], [10, 1, 1, 1],
  [11, 1, 1, 1], [12, 1, 1, 1], [13, 1, 1, 1], [14, 1, 1, 1], [15, 1, 1, 1], [16, 1, 1, 1], [17, 1, 1, 1], [18, 1, 1, 1], [19, 1, 1, 1], [20, 1, 1, 1],
  [21, 1, 1, 1], [22, 1, 1, 1], [23, 1, 1, 1], [24, 1, 1, 1], [25, 1, 1, 1], [26, 1, 1, 1], [27, 1, 1, 1], [28, 1, 1, 1], [29, 1, 1, 1], [30, 1, 1, 1],
  [31, 1, 1, 1], [32, 1, 1, 1], [33, 1, 1, 1], [34, 1, 1, 1], [35, 1, 1, 1], [36, 1, 1, 1], [37, 1, 1, 1], [38, 1, 1, 1], [39, 1, 1, 1], [40, 1, 1, 1],
  [41, 1, 1, 1], [42, 1, 1, 1], [43, 1, 1, 1], [44, 1, 2, 2], [45, 1, 1, 1], [46, 1, 1, 1], [47, 1, 1, 1] ]
  MODE: "conv"
  RESIDUAL_POOLING: True
  SEPARATE_QKV: True
  CLS_EMBED_ON: False # defauult: True

BN:
  USE_PRECISE_STATS: False
  NUM_BATCHES_PRECISE: 200

DETECTION:
  ENABLE: True
  ALIGNED: True
  SPATIAL_SCALE_FACTOR: 32
AVA:
  BGR: False
  DETECTION_SCORE_THRESH: 0.9
  TRAIN_PREDICT_BOX_LISTS: [
    "ava_train_v2.2.csv",
    "person_box_67091280_iou90/ava_detection_train_boxes_and_labels_include_negative_v2.2.csv",
  ]
  TEST_PREDICT_BOX_LISTS: ["person_box_67091280_iou90/ava_detection_val_boxes_and_labels.csv"]
  ANNOTATION_DIR: /datasets01/AVA/080720/frame_list/
  FRAME_LIST_DIR: /datasets01/AVA/080720/frame_list/
  FRAME_DIR: /datasets01/AVA/080720/frames/
  FULL_TEST_ON_VAL: True


SOLVER:
  CLIP_GRAD_L2NORM: 2.0
  ZERO_WD_1D_PARAM: True
  BASE_LR_SCALE_NUM_SHARDS: True
  BASE_LR: 0.075
  COSINE_AFTER_WARMUP: True
  COSINE_END_LR: 1e-7
  WARMUP_START_LR: 1e-8
  WARMUP_EPOCHS: 5.0
  LR_POLICY: cosine
  MAX_EPOCH: 20
  MOMENTUM: 0.9
  WEIGHT_DECAY: 1e-8
  OPTIMIZING_METHOD: sgd
MODEL:
  NUM_CLASSES: 80
  HEAD_ACT: sigmoid
  # NUM_CLASSES: 600
  ARCH: mvit
  MODEL_NAME: MViT
  LOSS_FUNC: bce # soft_cross_entropy # default cross_entropy
  DROPOUT_RATE: 0.0
  ACT_CHECKPOINT: True # for test flops

TEST:
  ENABLE: True
  DATASET: ava
  BATCH_SIZE: 1
  NUM_SPATIAL_CROPS: 1
  CHECKPOINT_FILE_PATH:  /home/jathu/mvit.pyth

DATA_LOADER:
  NUM_WORKERS: 4
  PIN_MEMORY: True
NUM_GPUS: 1
NUM_SHARDS: 1
SHARD_ID: 0
RNG_SEED: 0
OUTPUT_DIR: .

DEMO:
  ENABLE: True
  LABEL_FILE_PATH: /private/home/jathushan/3D/slowfast/ava_names.json
  WEBCAM: -1
  INPUT_VIDEO: /private/home/jathushan/datasets/ttv/webm2/82FE8F069F1354550003607470080_1fcf1757309.4.7.mp4
  OUTPUT_FILE: output.mp4


# #dbg
# DATA_LOADER:
#   NUM_WORKERS: 0
# NUM_GPUS: 1