Quick Start¶

The fastest way to try MELITE is with the bundled synthetic example dataset.

Install¶

git clone https://github.com/NanoBiostructuresRG/melite.git
cd melite
python -m pip install -e .

Run a Smoke Benchmark¶

melite run --smoke --config examples/example_config.toml

Smoke mode uses a small grid and fast cross-validation settings. It is useful for checking installation and data flow, but it is not benchmark-quality.

Export a Model Artifact¶

melite export --row 0 --csv examples/output/results.csv --outdir examples/output/

This retrains the selected model on all available example data and writes a .pkl artifact.

Predict from Python¶

import numpy as np
from melite import predict

X_new = np.load("examples/sample_PCA70.npz")["X"]
result = predict("examples/output/Model_SVC_sample_pca70.pkl", X_new)

print(result["predictions"])    # shape (n_samples,)
print(result["probabilities"])  # shape (n_samples, n_classes)

Expected Workflow¶

X / y -> melite run -> results.csv -> melite export -> .pkl -> predict()