Skip to content

open-mlpipe

Configuration

loisekk/open-mlpipe

Configuration¶

Default Configuration¶

open-mlpipe works out of the box with smart defaults. No configuration needed.

from open_mlpipe import run

# Auto-detect everything
ctx = run("dataset.csv")

Custom Configuration¶

Python API¶

from open_mlpipe import PipelineConfig, PipelineRunner

config = PipelineConfig(
    project="my-project",
    task="auto",
    data={"path": "data.csv", "target": "price"},
    model_selection={
        "candidates": ["lightgbm", "xgboost"],
        "scoring": ["r2"],
    },
    tuning={"enabled": True, "n_trials": 50},
)

runner = PipelineRunner(config)
ctx = runner.run()

YAML Config¶

# configs/my_pipeline.yaml
project: my-project
task: auto
data:
  path: data.csv
  target: price
model_selection:
  candidates: [lightgbm, xgboost]
tuning:
  enabled: true
  n_trials: 50

open-mlpipe run --config configs/my_pipeline.yaml

Configuration Options¶

Option	Default	Description
`project`	`"mlpipe-run"`	Project name
`task`	`"auto"`	Task type (auto/regression/classification)
`data.test_size`	`0.2`	Test set size
`tuning.enabled`	`True`	Enable hyperparameter tuning
`tuning.n_trials`	`"auto"`	Number of Optuna trials
`feature_selection.enabled`	`True`	Enable feature selection
`evaluation.explainability`	`True`	Enable SHAP explainability
`deployment.enabled`	`False`	Generate deployment artifacts

Next Steps¶