--- id: yacs version: "0.1.8" license: unclear license_treatment: permissive maintenance: abandoned --- # yacs — Yet Another Configuration System License: permissive · Maintenance: abandoned · Downloads: 3.4M/mo ## What it is and what it does YACS is a configuration system designed for scientific and machine learning projects that need to manage and reproduce experiments reliably. It lets you define default configurations in Python code, override them with YAML files for specific experiments, and further customize them from the command line—all while keeping a single authoritative source for what can be configured. The package centers on the CfgNode object, which acts as a hierarchical configuration container. You create a defaults file (typically config.py) that documents all available options with sensible starting values, then create lightweight YAML files for each experiment that only specify what changes. Your code imports the config, merges in overrides, and can freeze it to prevent accidental modification. This workflow emerged from real systems used in computer vision research and is particularly useful when you need to run many experiment variants and be able to reproduce them later. Use it for: - Define and manage hyperparameters for machine learning model training across multiple experiments with reproducible YAML configs. - Override configuration defaults from command-line arguments without duplicating config logic in argument parsing. - Document all configurable system options in one central Python file as a reference for all team members. - Serialize experimental settings to disk so you can later recall exactly what configuration produced a given result. - Share experiment configs as simple YAML files between team members without requiring code changes or environment variables. ## Worth the install? AI-flagged interpretation of the facts above — verify before relying. YACS is a lightweight configuration management system that lets you define, serialize, and override experimental configurations using YAML files and Python code, designed for reproducible machine learning and scientific workflows. Yes, if you are working on a machine learning or scientific project that needs reproducible configuration management and you do not require ongoing maintenance or updates. The package is stable and does what it claims, but it is abandoned—no security patches, no bug fixes, and no support for future Python versions. Use it for new projects only if you are comfortable maintaining a fork or accepting technical debt. ## Install pip install yacs uv add yacs poetry add yacs ## Installing yacs Before you install: Installation is straightforward with low friction—pure Python wheels for both Python 2 and 3. However, the package is abandoned as of 2020-08-10 with no commits since 2022-04-13, so expect no maintenance, bug fixes, or security updates going forward. License in practice: Licensed under Apache 2.0 (permissive), so you can use, modify, and distribute it freely in commercial and open-source projects without restriction. Quickstart: pip install yacs from yacs.config import CfgNode as CN _C = CN() _C.TRAIN = CN() _C.TRAIN.LR = 0.1 cfg = _C.clone() cfg.merge_from_file('experiment.yaml') print(cfg.TRAIN.LR) Verify before relying: - Whether the package works reliably with modern Python versions beyond what was tested at release - Whether PyYAML dependency has known vulnerabilities or compatibility issues with current Python releases ## Package facts - License: not declared (permissive) - Python support: unspecified - Install friction: low - Maintenance: abandoned - Downloads: 3.4M/month (top 5,000 on PyPI) - Known vulnerabilities: none known ## Tags configuration management system, yaml config framework, experiment hyperparameter config, reproducible ml configuration, config file override system, python config defaults, command line config override, config-management, reproducibility, ml-workflows [View on SkillFed](https://skillfed.io/packages/yacs) · [View on PyPI](https://pypi.org/project/yacs/)