project_name: t-final
dataset_path: "/home/apesea/xuhq/codes/DeepData/test/ray/input"
executor_type: ray
ray_address: 127.0.0.1:6388
export_type: jsonl
use_cache: false
export_path: "/home/apesea/xuhq/codes/DeepData/test/ray/out_final"
skip_op_error: false
video_key: video_path
open_monitor: false
text_keys: ['text']
np: 1
keep_stats_in_res_ds: True          # ← 让 video_audio_tags 等 meta 字段导出

process:
  - video_extract_frames_mapper:
      frame_num: 5
      frame_sampling_method: uniform
      video_key: video_path
      output_format: path
      frame_dir: "./frames"
      num_proc: 1

  - video_tagging_from_audio_mapper:
      video_key: video_path
      num_proc: 1
      hf_ast: 'MIT/ast-finetuned-audioset-10-10-0.4593'

  - video_duration_filter:
      video_key: video_path
      min_duration: 3
      max_duration: 60
      num_proc: 1

  - video_aesthetics_filter:
      video_key: video_path
      min_score: 0.4
      num_proc: 1

  - video_captioning_from_frames_mapper:
      video_key: video_path
      hf_img2seq: Salesforce/blip-image-captioning-base   # ~990MB，稳
      frame_sampling_method: uniform
      frame_num: 3
      keep_original_sample: false
      caption_num: 1
      keep_candidate_mode: random_any
      num_proc: 1

  - video_frames_text_similarity_filter:
      video_key: video_path
      min_score: 0.25
      num_proc: 1

  - ray_video_deduplicator:            # ← 不是 ray_video_deduplicator
      video_key: video_path
      backend: ray_actor