dvc.yaml

vars:
  - trials:
    - name: hipattack
      title: HIPATTACK
    - name: poise
      title: POISE1
    - name: manage
      title: MANAGE
    - name: compass
      title: COMPASS
    - name: tips3
      title: TIPS3
    - name: hope3
      title: HOPE-3
    - name: rely
      title: RE-LY
  - output_directory: data

stages:
  diff:
    foreach: ${trials}
    do:
      cmd: >
        poetry run python3 -m traq.preprocessing diff
        --config_filename configs/${item.name}.toml
        --output_directory ${output_directory}
      deps:
        - traq/preprocessing/diff.py
        - configs/${item.name}.toml
        - data/raw/${item.title}
      outs:
        - data/diffs/${item.name}

          #  summarize:
          #    foreach: ${trials}
          #    do:
          #      cmd: >
          #        poetry run python3 -m traq.preprocessing summarize
          #        --config_filename configs/${item.name}.toml
          #        --output_directory ${output_directory}
          #      deps:
          #        - traq/preprocessing/summarize.py
          #        - configs/${item.name}.toml
          #        - data/diffs/${item.name}
          #      outs:
          #        - data/summaries/${item.name}

  derive_labels:
    foreach: ${trials}
    do:
      cmd: >
        poetry run python3 -m traq.preprocessing derive_labels
        --config_filename configs/${item.name}.toml
        --output_directory ${output_directory}
      deps:
        - traq/preprocessing/derive_labels.py
        - configs/${item.name}.toml
        - data/diffs/${item.name}
      outs:
        - data/derived/${item.name}

  explode_pickles:
    foreach: ${trials}
    do:
      cmd: >
        poetry run python3 -m traq.services.explode_pickles
        --input_directory data/derived/${item.name}
        --output_filename data/exploded/${item.name}.pkl
      deps:
        - data/derived/${item.name}
        - traq/utils.py
        - traq/services/explode_pickles.py
      outs:
        - data/exploded/${item.name}.pkl

  compute_meta_features:
    foreach: ${trials}
    do:
      cmd: >
        python3.7 -W ignore -m traq.services.generate_metafeatures
        --input_filename data/exploded/${item.name}.pkl
        --output_filename data/metafeatures/${item.name}.pkl
      deps:
        - data/exploded/${item.name}.pkl
        - traq/services/generate_metafeatures.py
      outs:
        - data/metafeatures/${item.name}.pkl

  compare_od_models:
    foreach: ${trials}
    do:
      cmd: >
        poetry run python3 -W ignore -m traq.compare_unsupervised
        --input_directory data/derived/${item.name}
        --output_filename output/comparisons/${item.name}.csv
      deps:
        - traq/compare_unsupervised.py
        - traq/utils.py
        - traq/evaluation.py
        - traq/data/pickle.py
        - traq/data/base.py
        - traq/grid.py
        - data/derived/${item.name}
      outs:
        - output/comparisons/${item.name}.csv

  fit_metaod_models:
    foreach: ${trials}
    do:
      cmd: >
        PYTHONPATH=. python3.7 traq/services/train_metaod.py
        --metafeatures_directory data/metafeatures
        --performance_directory output/comparisons
        --output_filename models/metaod/${item.name}.pkl
        --exclude_from_training ${item.name}
      deps:
        - traq/services/train_metaod.py
        - data/metafeatures
        - output/comparisons
      outs:
        - models/metaod/${item.name}.pkl

  comparison_viz:
    cmd: >
      mkdir -p output/viz &&
      poetry run python3 -m traq.format_comparisons
      --input_directory output/comparisons
      --output_directory output/viz > output/viz/summary.txt
    deps:
      - traq/format_comparisons.py
      - output/comparisons
    outs:
      - output/viz

  compare_meta_models:
    foreach: ${trials}
    do:
      cmd: >
        poetry run python3 -W ignore -m traq.compare_selection
        --input_directory data/derived/${item.name}
        --output_filename output/meta_comparisons/${item.name}.csv
      deps:
        - traq/metaod_wrapper.py
        - traq/compare_selection.py
        - traq/utils.py
        - traq/evaluation.py
        - traq/data/pickle.py
        - traq/data/base.py
        - traq/grid.py
        - data/derived/${item.name}
        - data/metafeatures/${item.name}.pkl
        - models/metaod/${item.name}.pkl
      outs:
        - output/meta_comparisons/${item.name}.csv

  meta_comparison_viz:
    cmd: >
      mkdir -p output/meta_viz &&
      poetry run python3 -m traq.format_comparisons
      --input_directory output/meta_comparisons
      --output_directory output/meta_viz > output/meta_viz/summary.txt
    deps:
      - traq/format_comparisons.py
      - output/meta_comparisons
    outs:
      - output/meta_viz