includes: projects/task/vtt.yaml
model:
  model_cls: MMFusionSeparate
  mm_encoder_cls: 
  video_encoder_cls: MMBertForEncoder
  text_encoder_cls: BertModel
  num_hidden_video_layers: 6
fairseq:
  dataset:
    batch_size: 224
#   model_cls: MMFusionShare
#   mm_encoder_cls: MMBertForEncoder