dvc.lock

schema: '2.0'
stages:
  extract_data:
    cmd: tar -xzf data/data.tar.gz && md5sum data/data.tar.gz > data/data.tar.gz.hash
    deps:
    - path: data/data.tar.gz
      md5: 73d66051b6e8cb435f1804e10ee4aa2f
      size: 3020370695
    outs:
    - path: data/data.tar.gz.hash
      md5: e1965d41201a3bd4c0ea8881f8df8f1c
      size: 51
  combine_dfs:
    cmd: python src/abodybuilder3/stages/data/combine_data_dfs.py
    deps:
    - path: data/data.tar.gz.hash
      md5: e1965d41201a3bd4c0ea8881f8df8f1c
      size: 51
    outs:
    - path: data/structures/success_df.csv
      md5: 9ac3522949828c5f96d1d585320c0116
      size: 610263
    - path: data/structures/summary_df.csv
      md5: 1e15be365726adeeb2c40890893548ab
      size: 6033272
  embeddings:
    cmd: python src/abodybuilder3/stages/data/language_model_embeddings.py
    deps:
    - path: data/structures/summary_df.csv
      md5: 1e15be365726adeeb2c40890893548ab
      size: 6033272
    params:
      params.yaml:
        language:
          model:
          chunk_size: 1000
    outs:
    - path: data/structures/structures_plm.tar.gz
      md5: e593237bce149f736040580ec432e3b0
      size: 64
  extract_embeddings:
    cmd: tar -xzf data/structures/structures_plm.tar.gz && md5sum data/structures/structures_plm.tar.gz
      > data/structures/structures_plm.tar.gz.hash
    deps:
    - path: data/structures/structures_plm.tar.gz
      md5: e593237bce149f736040580ec432e3b0
      size: 64
    outs:
    - path: data/structures/structures_plm.tar.gz.hash
      md5: 4742735d41a9fe7318c456c169adc478
      size: 72
  cluster:
    cmd: python src/abodybuilder3/stages/data/cluster_data.py
    deps:
    - path: data/structures/summary_df.csv
      md5: 1e15be365726adeeb2c40890893548ab
      size: 6033272
    params:
      params.yaml:
        cluster:
          min_seq_id: 0.95
          coverage: 0.8
    outs:
    - path: data/clusters.csv
      md5: fcdbf74ff2d8995da77075d79671ba20
      size: 260832
  filter:
    cmd: python src/abodybuilder3/stages/data/filter_data.py
    deps:
    - path: data/structures/summary_df.csv
      md5: 1e15be365726adeeb2c40890893548ab
      size: 6033272
    params:
      params.yaml:
        filter:
          resolution_cutoff: 3.5
          abangle_cutoff: 3.5
          cdrh3_length_cutoff: 30
          total_cdr_length_cutoff: 500
          rare_species_cutoff: 15
    outs:
    - path: data/filters.csv
      md5: c3767cf2580815f26546ff2189d161e7
      size: 904862
  split:
    cmd: python src/abodybuilder3/stages/data/split_data.py
    deps:
    - path: data/clusters.csv
      md5: fcdbf74ff2d8995da77075d79671ba20
      size: 260832
    - path: data/filters.csv
      md5: c3767cf2580815f26546ff2189d161e7
      size: 904862
    - path: data/structures/summary_df.csv
      md5: 1e15be365726adeeb2c40890893548ab
      size: 6033272
    params:
      params.yaml:
        split:
          seed: 0
          validation_size: 150
          test_size: 100
          valid_test_max_cluster_size: 10
          test_valid_resolution_cutoff: 2.3
          test_valid_cdrh3_cutoff: 22
          use_clustering: false
          use_cdr_pseudosequence: true
    outs:
    - path: data/split.csv
      md5: 5c181f849e8651132e19098457db2262
      size: 259543
    - path: data/valid_test_set_similarity.csv
      md5: 8454f1a7ee95be455318409cf48d9b43
      size: 123455348