Getting Started (skyulf-core)

This page is the fastest path to running skyulf-core locally.

Install

From the repository root:

pip install -e ./skyulf-core

Minimal end-to-end example

SkyulfPipeline expects a configuration with:

• preprocessing: a list of steps
• modeling: a single model config

import pandas as pd

from skyulf.pipeline import SkyulfPipeline

df = pd.DataFrame(
    {
        "age": [10, 20, None, 40],
        "city": ["A", "B", "A", "C"],
        "target": [0, 1, 0, 1],
    }
)

config = {
    "preprocessing": [
        {
            "name": "impute_age",
            "transformer": "SimpleImputer",
            "params": {"columns": ["age"], "strategy": "mean"},
        },
        {
            "name": "encode_city",
            "transformer": "OneHotEncoder",
            "params": {"columns": ["city"], "drop_original": True},
        },
    ],
    "modeling": {
        "type": "logistic_regression",
        "params": {"max_iter": 1000},
    },
}

pipeline = SkyulfPipeline(config)
metrics = pipeline.fit(df, target_column="target")
preds = pipeline.predict(df.drop(columns=["target"]))

print(metrics)
print(preds.head())

Next steps

• Read the User Guide section “Pipeline Quickstart” for train/test splits.
• Use the Reference section for supported preprocessing and modeling nodes.