Installation¶
Base Installation¶
ml4t-models keeps the base dependency set small. The default install gives you:
- typed batch and result contracts
- closed-form and NumPy-based model families
- pipeline composition utilities
- frame adapters that do not require heavy optional dependencies
Optional Extras¶
Neural Models¶
Install torch-backed models:
This extra is required for:
CAEModelSAEModelStochasticDiscountFactorModelLSTMPortfolioModelDeepPortfolioModel
Cross-Library Integration¶
Install tabular and spec helpers:
This extra is useful when you want:
ResultsFrame.to_polars()- parquet writing via
write_backtest_frames ml4t-specs-aware schema resolution
Documentation¶
Build the docs locally:
Everything¶
Python Version¶
ml4t-models currently targets:
- Python
>=3.12,<3.14
Development Setup¶
Using uv:
Run the quality gates:
Build the docs:
Related Libraries¶
ml4t-models is designed to integrate at boundaries with the rest of the ML4T stack:
ml4t-datafor dataset loading and canonical schema metadataml4t-engineerfor feature generation and labelsml4t-diagnosticfor IC, validation, and report generationml4t-backtestfor execution and backtest state transitions