D1:
 x2t_interleave_local:
  dataset_type: x2t_general
  raw_bytes_input: true
  is_image: true
  task_type: "tv2t"
  text_template: true
  dataset_names:
    i2t_local_256:
      data_dir:
        - datasets/image2text/local_256.parquet
  debug_parquet_repeat: 256  # debug option for quick smoke tests on small local datasets

  video_transform_args:
    resolution: 616
    mode: "bucket"
    divisible_crop_size: 28
    aspect_ratios: ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"]
    mean: [0.48145466, 0.4578275, 0.40821073]
    std: [0.26862954, 0.26130258, 0.27577711]

  is_mandatory: false
  weight: 1.0

D2:
 x2t_interleave_local:
  dataset_type: x2t_general
  raw_bytes_input: true
  is_image: false
  task_type: "tv2t"
  text_template: true
  dataset_names:
    v2t_local_256:
      data_dir:
        - datasets/video2text/local_256.parquet
  debug_parquet_repeat: 256  # debug option for quick smoke tests on small local datasets

  video_transform_args:
    resolution: 616
    mode: "bucket"
    divisible_crop_size: 28
    aspect_ratios: ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"]
    mean: [0.48145466, 0.4578275, 0.40821073]
    std: [0.26862954, 0.26130258, 0.27577711]

  video_frame_sampler_args:
    type: "multi_clips"
    res_dump: ["12fps_480p"]
    params:
      temporal: 4
      sample_fps: 12
      max_duration: 2
      assert_seconds: false
      truncate: false

  is_mandatory: false
  weight: 1.0