What is the priority order for parameter value resolution?

Values are resolved from highest to lowest priority: CLI arguments, direct assignment, context managers, environment variables, and finally default code values.

How do I run hyperparameter sweeps with this skill?

Use the 'piter' utility to create parameter combinations using Zip, Cartesian product, or Trials operators, then iterate through the resulting configs in your training loop.

Does it support complex Python types?

Yes, it supports standard types (int, float, str, bool), Enums, Literals, Lists, Tuples, and Union types for creating advanced subcommand-like patterns.

How does params-proto handle boolean flags in the CLI?

Boolean parameters defaulting to False generate a standard flag (e.g., --verbose), while parameters defaulting to True generate a negative flag (e.g., --no-cuda) to toggle the value.

Can I use environment variables for sensitive data like API keys?

Yes, the EnvVar utility allows you to map parameters directly to environment variables, supporting required values, multiple fallbacks, and default values.

Params Proto Declarative ML Management

Name: Params Proto Declarative ML Management
Author: geyang

bygeyang

•

データサイエンスとML

Streamlines machine learning experiment configuration using declarative decorators and automatic CLI generation.

params-proto is a specialized library designed for ML/AI researchers and developers to manage hyperparameters and configurations with minimal boilerplate. By leveraging Python decorators like @proto.cli and @proto.prefix, it transforms standard functions and classes into robust command-line interfaces with automatic type-checking and documentation. It excels in complex environments requiring multi-namespace configurations, environment variable overrides, and sophisticated hyperparameter sweeps, ensuring that experiment settings are always reproducible and easy to modify from the shell.

主な機能

01Automatic CLI generation with type-hinted parameters and help text

02Support for Union types to implement clean subcommand-like CLI patterns

03Seamless environment variable integration with priority-based resolution

04Namespaced configuration management using singleton and multi-instance decorators

0518 GitHub stars

06Advanced hyperparameter sweep generation via the piter utility

ユースケース

01Managing hierarchical experiment configurations across multiple model modules

02Automating large-scale hyperparameter searches and grid sweeps

03Scaling single-file ML scripts into production-ready CLI applications

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add geyang/params-proto params-proto

For use in Claude.ai and ChatGPT

Download Skill