config.yml

logging:
  # Python logging config (passed to dictConfig)
  version: 1
  formatters:
    default:
      format: '[%(asctime)s] %(levelname)s in %(name)s:%(lineno)s: %(message)s'
  handlers:
    console:
      level: INFO
      formatter: default
      class: logging.StreamHandler
      stream: ext://sys.stdout
    log_file:
      level: INFO
      formatter: default
      class: logging.handlers.RotatingFileHandler
      filename: 'logs/server.log'
      mode: 'a'
      maxBytes: 67108864  # 64 MB
      backupCount: 10
  loggers:
    # root logger
    '':
      level: INFO
      handlers:
        - console
        - log_file
    sciencebeam_parser:
      level: DEBUG
    __main__:
      level: DEBUG
    delft:
      level: INFO
    sciencebeam_trainer_delft:
      level: INFO

# The download directory for ScienceBeam Parser resources
# Note: the CV model may download resources to `~/.torch/iopath_cache`,
#    unless FVCORE_CACHE is set. (See `iopath.common.file_io.get_cache_dir`)
download_dir: '/data/.sciencebeam-parser/.cache'

pdfalto:
  path: https://github.com/kermitt2/pdfalto/files/6104204/pdfalto-4b4e983413278a07bb4cc4b2836de03adc8ca6dc-dockcross-linux-64.gz
wapiti:
  install_source: 'https://github.com/kermitt2/Wapiti/archive/a9c25d2bcccd60f1a54a7019689bd5229e866f00.tar.gz'
xslt:
  tei_to_jats:
    parameters:
      # parameters for tei-to-jats.xsl
      output_parameters: 'false'
      output_bold: 'false'
      output_italic: 'false'
      output_empty_figure_graphic: 'true'
      acknowledgement_target: 'ack'
      annex_target: 'back'
lookup:
  country:
    paths:
      - https://raw.githubusercontent.com/kermitt2/grobid/0.8.2/grobid-home/lexicon/countries/CountryCodes.xml
  first_name:
    paths:
      - https://raw.githubusercontent.com/kermitt2/grobid/0.8.2/grobid-home/lexicon/names/firstname.5k
      - https://raw.githubusercontent.com/kermitt2/grobid/0.8.2/grobid-home/lexicon/names/names.female
      - https://raw.githubusercontent.com/kermitt2/grobid/0.8.2/grobid-home/lexicon/names/names.male
  last_name:
    paths:
      - https://raw.githubusercontent.com/kermitt2/grobid/0.8.2/grobid-home/lexicon/names/lastname.5k
      - https://raw.githubusercontent.com/kermitt2/grobid/0.8.2/grobid-home/lexicon/names/names.family
processors:
  fulltext:
    merge_raw_authors: false
    use_cv_model: false
    cv_render_dpi: 100
    use_ocr_model: false
    replace_text_by_cv_graphic: false
    max_graphic_distance: 0.3
models:
  segmentation:
    path: 'https://github.com/kermitt2/grobid/raw/refs/tags/0.6.2/grobid-home/models/segmentation'
    engine: 'wapiti'
    use_first_token_of_block: false
  header:
    path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/v0.0.1/2020-10-04-delft-grobid-header-biorxiv-no-word-embedding.tar.gz'
  name_header:
    # path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/grobid-0.6.1/2021-06-21-grobid-0.6.1-name-header-no-word-embedding-no-layout-features-e800.tar.gz'
    path: 'https://github.com/kermitt2/grobid/raw/0.6.0/grobid-home/models/name/header'
    engine: 'wapiti'
  name_citation:
    # path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/grobid-0.6.1/2021-06-28-grobid-0.6.1-name-citation-no-word-embedding-no-layout-features-e500.tar.gz'
    path: 'https://github.com/kermitt2/grobid/raw/0.6.2/grobid-home/models/name/citation'
    engine: 'wapiti'
  affiliation_address:
    path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/v0.0.1/2020-10-04-delft-grobid-affiliation-address-biorxiv-no-word-embedding.tar.gz'
  fulltext:
    path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/biorxiv-grobid/2021-05-11-delft-grobid-fulltext-biorxiv-10k-auto-v0.0.21-train-1986-e159.tar.gz'
  figure:
    path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/biorxiv-grobid/2021-05-11-delft-grobid-figure-biorxiv-10k-auto-v0.0.18-train-1865-e219.tar.gz'
  table:
    path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/biorxiv-grobid/2021-05-11-delft-grobid-table-biorxiv-10k-auto-v0.0.18-train-1865-e569.tar.gz'
  reference_segmenter:
    path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/v0.0.1/2020-10-04-delft-grobid-reference-segmenter-biorxiv-no-word-embedding.tar.gz'
  citation:
    path: 'https://github.com/elifesciences/sciencebeam-models/releases/download/v0.0.1/2020-10-04-delft-grobid-citation-biorxiv-no-word-embedding.tar.gz'

cv_models:
  default:
    path: 'lp://efficientdet/PubLayNet'
    engine: 'layout_parser'
    score_threshold: 0.1

ocr_models:
  default:
    engine: 'tesserocr'
    lang: 'eng'
    # see https://github.com/sirfz/tesserocr/blob/v2.5.2/tesserocr.pyx#L75-L90
    # (specify literal or int value)
    oem: 'DEFAULT'
    # see https://github.com/sirfz/tesserocr/blob/v2.5.2/tesserocr.pyx#L102-L121
    # (specify literal or int value)
    psm: 'SPARSE_TEXT'

doc_to_pdf:
  enabled: true
  listener:
    port: 2003
    process_timeout: 600
    max_uptime: 10
  convert:
    remove_line_no: true
    remove_header_footer: true
    remove_redline: true

# preload resources and models on startup
preload_on_startup: true