landscapy-ml

landscapy-ml is the small bridge between landscapy FitnessLandscape objects and PyTorch models.

More information, documentation and usage instructions can be found on landscapy.

Installation

landscapy-ml is installed by default with pip install landscapy or pip install landscapy-ml. This remains the recommended installation path.

To install directly from a checkout:

python -m pip install .

Python

Use the Python interface to convert a FitnessLandscape into PyTorch-ready records and datasets, then adapt model outputs back into landscape layers as described in the Python usage guide.

from importlib.resources import files

from fitness_landscape.core.landscape import read_csv_landscape

from landscapyml import (
    LandscapeDataset,
    export_landscape_records,
    make_fitness_target_getter,
)

data_path = files("landscapyml").joinpath("data/minimal_landscape.csv")
landscape = read_csv_landscape(
    data_path,
    sequence_col="sequence",
    alphabet=list("AC"),
    graph="hamming",
    numeric_layers=["target"],
    attach_embeddings=False,
)
exported = export_landscape_records(
    landscape,
    fitness_layers=["target"],
    include_embeddings=False,
)
dataset = LandscapeDataset(
    exported.records,
    target_getter=make_fitness_target_getter("target"),
)

features, target = dataset[0]
print(len(dataset), features.shape, target.shape)

CLI

Use the command-line interface to inspect registered models and data builders and run the maintained CSV workflows described in the CLI guide.