unimernet_base.yaml

model:
  arch: unimernet
  load_finetuned: False
  load_pretrained: False
  pretrained: "path/to/pretrained/weight"
  finetuned: ""
  tokenizer_name: nougat
  tokenizer_config:
    path: ./models/unimernet
  model_name: unimernet
  model_config:
    max_seq_len: 384


preprocess:
  vis_processor:
    train:
      name: "formula_image_train"
      image_size:
        - 192
        - 672
    eval:
      name: "formula_image_eval"
      image_size:
        - 192
        - 672
  text_processor:
    train:
      name: "blip_caption"
    eval:
      name: "blip_caption"