Train Test Split Python Packages

kennard-stone

This is an algorithm for evenly partitioning.

safesplit

Leakage-safe train/test splits — group-aware, time-aware, sequence-aware (homology), de-duplicated. The companion that prevents what LeakHound detects.

graph-part

GraphPart, a data partitioning method for ML on biological sequences

bboxconverter

This library allows reading and converting bounding box annotations in many popular formats

insurance-cv

Temporal cross-validation for insurance pricing - respects policy time structure, CatBoost, Polars

protclust

protclust is a Python library for protein sequence analysis that integrates MMseqs2 for fast clustering and provides tools for creating robust machine learning datasets. It offers cluster-aware data splitting to prevent sequence similarity bias in model evaluation, along with comprehensive protein embedding capabilities for feature generation.

jano

Temporal partitioning and backtesting for time-correlated datasets

python-splitter

📁 Repo for python_splitter Python package. This package can split Images into Train, Test, Validation folders automatically by shuffling media/images for machine learning.

spltr

`Spltr` is a simple PyTorch-based data loader and splitter. It may be used to load arrays and matrices or Pandas DataFrames and CSV files containing numerical data with subsequent split it into train, test (validation) subsets in the form of PyTorch DataLoader objects.

maestros

Multi-label stratified splits, while preserving group independence. Includes a stratification chart and report.

splitcheck

Detect row overlap and leakage between dataset splits.

mmseqspy

protclust is a Python library for protein sequence analysis that integrates MMseqs2 for fast clustering and provides tools for creating robust machine learning datasets. It offers cluster-aware data splitting to prevent sequence similarity bias in model evaluation, along with comprehensive protein embedding capabilities for feature generation.

bboxtools-2

This library allows reading and converting bounding box annotations in many popular formats