dvc-demo/dvc.yaml

stages:
  split:
    cmd: python3 main.py split
    deps:
    - data/CrossValidated-Questions-Nov-2020.csv
    - main.py
    params:
    - paths.dataset
    - paths.test_df
    - paths.train_df
    - split.seed
    outs:
    - outputs/test.csv.gz
    - outputs/train.csv.gz
  featurize:
    cmd: python3 main.py featurize
    deps:
    - main.py
    - outputs/test.csv.gz
    - outputs/train.csv.gz
    params:
    - paths.test_df
    - paths.test_df_featurized
    - paths.train_df
    - paths.train_df_featurized
    outs:
    - outputs/test-featurized.csv.gz
    - outputs/train-featurized.csv.gz
  tfidf:
    cmd: python3 main.py tfidf
    deps:
    - main.py
    - outputs/test-featurized.csv.gz
    - outputs/train-featurized.csv.gz
    params:
    - paths.test_df_featurized
    - paths.test_tfidf
    - paths.tfidf
    - paths.train_df_featurized
    - paths.train_tfidf
    - tfidf.max_features
    outs:
    - outputs/test-tfidf.npz
    - outputs/tfidf.pkl
    - outputs/train-tfidf.npz
  train:
    cmd: python3 main.py train
    deps:
    - main.py
    - outputs/train-featurized.csv.gz
    - outputs/train-tfidf.npz
    params:
    - paths.model
    - paths.train_df_featurized
    - paths.train_tfidf
    - train.loss
    - train.seed
    - paths.train_metrics
    outs:
    - outputs/model.pkl
    metrics:
    - metrics-train.yaml:
        cache: false
  test:
    cmd: python3 main.py test
    deps:
    - main.py
    - outputs/test-featurized.csv.gz
    - outputs/test-tfidf.npz
    - outputs/model.pkl
    params:
    - paths.model
    - paths.test_df_featurized
    - paths.test_tfidf
    - paths.test_metrics
    metrics:
    - metrics-test.yaml:
        cache: false