rbgirshick/yacs

Health v2 · maintenance only

Adoption not part of the score

1338 stars · 90 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

YACS (Yet Another Configuration System) is a lightweight Python library for defining and managing system configurations, originally built for scientific and ML experiment settings. It provides a CfgNode object with sensible defaults, YAML-based serialization, merging of experiment-specific overrides, and a freeze mechanism for reproducibility.

Use cases

• manage hyperparameters for machine learning experiments
• serialize experiment configs to yaml for reproducibility
• override default config values per experiment
• define a single defaults file for all project options
• freeze configuration after setup to prevent accidental changes
• manage configs for computer vision model training like detectron

When to choose

• you run scientific or ML experiments and need reproducible, serializable configs
• you want a small, focused config library with defaults plus per-experiment yaml overrides
• your project follows the detectron/py-faster-rcnn style config paradigm

When to avoid

• you need dynamic or schema-validated configs with complex types
• you want runtime-reloadable or environment-variable-driven configuration for services
• you need a general-purpose settings library for web apps (e.g. pydantic-settings, dynaconf)

Facets

library · maturity maintenance

configuration-management serialization developer-tools machine-learning python cross-platform yaml-config experiment-config reproducibility hyperparameters cfg-node research automation

Member repositories

For agents

markdown · JSON · MCP: product_card(name="rbgirshick/yacs")