dvc.lock

schema: '2.0'
stages:
  scrape:
    cmd: python -m telex.pipeline.scrape.scraping --output-folder "data/raw/raw"
    deps:
    - path: telex/pipeline/scrape
      md5: 7da4762f14f0811c50e13c610379d4c4.dir
      size: 2265
      nfiles: 2
    params:
      params.py:
        MAX_PAGES: 187
        PER_PAGE: 400
        SLEEP_TIME: 0.5
    outs:
    - path: data/raw
      md5: 170069251545593c2f2fd9f421249eeb.dir
      size: 592970611
      nfiles: 74711
  preprocess:
    cmd: python -m telex.pipeline.preprocess.preprocessing --input-folder "data/raw/raw"
      --output-folder "data/preprocessed"
    deps:
    - path: data/raw
      md5: 170069251545593c2f2fd9f421249eeb.dir
      size: 592970611
      nfiles: 74711
    - path: telex/pipeline/preprocess
      md5: 68ddf686b26a16aac649f8e905483b3b.dir
      size: 1883
      nfiles: 2
    outs:
    - path: data/preprocessed
      md5: 4c7d4461073c9b5b8cd526c4fada9878.dir
      size: 288517245
      nfiles: 1
  train:
    cmd: python -m telex.pipeline.train.training --input-file "data/preprocessed/telex.json"
      --output-folder "data/models"
    deps:
    - path: data/preprocessed
      md5: 4c7d4461073c9b5b8cd526c4fada9878.dir
      size: 288517245
      nfiles: 1
    - path: telex/pipeline/train
      md5: cfa4c55926d00d6ad8d680f142c83c08.dir
      size: 1769
      nfiles: 2
    params:
      params.py:
        BigramParams:
          MODEL: bigram
        MODEL: transformer
        TrainingParams:
          SEED: 42
          LR: 0.001
          SPLIT:
          - 0.9
          - 0.05
          - 0.05
          BLOCK_SIZE: 100
          BATCH_SIZE: 64
          EPOCHS: 1
          STEPS: 5000
        TransformerParams:
          MODEL: transformer
          D_MODEL: 120
          HEADS: 6
          D_FF: 512
          DROPOUT: 0.1
          NUM_LAYERS: 2
        VOCAB: "AÁBCDEÉFGHIÍJKLMNOÓÖŐPQRSTUÚÜŰVWXYZaábcdeéfghiíjklmnoóöőpqrstuúüűvwxyz0123456789.,;:?!-–_()[]{}'\"\
          /\\@#$%^&*+=|~` "
    outs:
    - path: data/models
      md5: c70961aae9d65e913b3bc57c689df1a5.dir
      size: 2111358
      nfiles: 1
  evaluate:
    cmd: python -m telex.pipeline.evaluate.evaluation --model-file "data/models/model.pt"
      --input-file "data/preprocessed/telex.json"
    deps:
    - path: data/models
      md5: 49cf68016696a153031893a3e48d6141.dir
      size: 2124928
      nfiles: 2
    - path: data/preprocessed
      md5: 4c7d4461073c9b5b8cd526c4fada9878.dir
      size: 288517245
      nfiles: 1
    - path: telex/pipeline/evaluate
      md5: 35911ce397d1fd191bbd311332f69408.dir
      size: 3247
      nfiles: 2
    params:
      params.py:
        BigramParams:
          MODEL: bigram
        MODEL: transformer
        TrainingParams:
          SEED: 42
          LR: 0.001
          SPLIT:
          - 0.9
          - 0.05
          - 0.05
          BLOCK_SIZE: 100
          BATCH_SIZE: 64
          EPOCHS: 1
          STEPS: 5000
        TransformerParams:
          MODEL: transformer
          D_MODEL: 120
          HEADS: 6
          D_FF: 512
          DROPOUT: 0.1
          NUM_LAYERS: 2
        VOCAB: "AÁBCDEÉFGHIÍJKLMNOÓÖŐPQRSTUÚÜŰVWXYZaábcdeéfghiíjklmnoóöőpqrstuúüűvwxyz0123456789.,;:?!-–_()[]{}'\"\
          /\\@#$%^&*+=|~` "