unilm/edgelm/examples/MMPT/projects/task/vtt.yaml

includes: projects/task/ft.yaml
dataset:
  meta_processor: MSRVTTMetaProcessor
  train_path: data/msrvtt/MSRVTT_train.csv
  jsfusion_path: data/msrvtt/MSRVTT_JSFUSION_test.csv
  full_test_path: data/msrvtt/MSRVTT_FULL_test.csv
  dup: 20
  val_path: data/msrvtt/MSRVTT_JSFUSION_test.csv
  vfeat_dir: data/feat/feat_vtt_s3d
  text_processor: MSRVTTTextProcessor
  json_path: data/msrvtt/MSRVTT_data.json
  aligner: DSAligner
  num_iso_layer: 12
model:
  model_cls: MMFusionJoint
  mm_encoder_cls: MMBertForJoint
loss:
  loss_cls: T2VContraLoss
fairseq:
  dataset:
    batch_size: 256
  optimization:
    max_epoch: 10
  checkpoint:
    save_dir: runs/task/vtt