configurations/experiment/process_data.yaml

defaults:
  - base_experiment
  - _self_

tasks: [visualize_dataset]  # add the method names you want to run, e.g. [cache_prompt_embed]
new_data_root: null # newly created csv and files will be saved here. null will defaults to the output_dir of this run.

visualize_dataset:
  n_samples: 32
  disable_filtering: false
  use_processed: true  # if true, will use processed videos from __getitem__ instead of raw files

cache_prompt_embed:
  batch_size: 32

create_gemini_caption:
  n_workers: 12

run_hand_pose_estimation:
  n_workers: 12 # not used

run_human_detection:
  total_workers: 2 # not used
  job_id: 0
  # save_dir: "outputs/"

benchmark_dataloader:
  batch_size: 4
  num_workers: 8