data:
  dataloader:
    batch_size: 16
    eval_batch_size: 32
    num_workers: 2
    persistent_workers: true
    pin_memory: false
    prefetch_factor: 2
    seed: 42
  dataset:
    clips_dir: /data/qing/seamless/clips/n64_fps4_crop224_short224
    hflip_prob: 0.5
    image_size: 224
    mean:
    - 0.485
    - 0.456
    - 0.406
    num_frames: 64
    overfit: false
    std:
    - 0.229
    - 0.224
    - 0.225
    test_txt: /home/qing/mesh3d/gen3d/seamless_interaction/dataset/test.txt
    train_txt: /home/qing/mesh3d/gen3d/seamless_interaction/dataset/train.txt
model:
  hf_model_path: /data/qing/seamless/hf_cache/videomae-base-finetuned-kinetics
  image_size: 224
  num_frames: 64
training:
  common:
    end_lr: 1.0e-06
    epochs: 200
    gradient_checkpointing: false
    lr: 0.0001
    lr_backbone_mult: 0.1
    lr_warmup_epochs: 10
    weight_decay: 0.05