Task gridsΒΆ

You can use tasks[*].grid to create multiple tasks from a single task declaration, for example, to train various models with different parameters:

graph LR load[Load] --> process[Process] --> exp1[Train n_estimators=5, criterion=gini] process --> exp2[Train n_estimators=10, criterion=gini] process --> exp3[Train n_estimators=20, criterion=gini] process --> exp4[Train n_estimators=5, criterion=entropy] process --> exp5[Train n_estimators=10, criterion=entropy] process --> exp6[Train n_estimators=20, criterion=entropy]
# execute independent tasks in parallel
executor: parallel

  - source: random-forest.py
    # generates random-forest-5-gini, random-forest-10-gini, ..., random-forest-20-entropy
    name: random-forest-[[n_estimators]]-[[criterion]]
    product: random-forest-[[n_estimators]]-[[criterion]].html
        # creates 6 tasks (3 * 2)
        n_estimators: [5, 10, 20]
        criterion: [gini, entropy]

Download example:

pip install ploomber
ploomber examples -n cookbook/grid -o grid
cd grid
pip install -r requirements.txt
ploomber build

Click here to see the complete example.

For full details, see the grid API documentation.

An in-depth tutorial showing how to use grid and MLflow for experiment tracking is available here.