configs/charades_ego/svitt.yml

model:
    pretrain: ckpt/svitt-ego.pth
    freeze_vis_backbone: true
    freeze_txt_backbone: true
    num_frames: 16
    config: configs/charades_ego/action-recognition.yaml

data:
    dataset: charades_ego
    root: data/charades_ego/video
    metadata_val: data/charades_ego/csv/{}.csv
    label_map: meta/charades_ego/charades_ego.json
    clip_length: 16
    sparse_sample: true