Verified:GitHub

ml-config-manager

Generates Hydra configuration files for PyTorch Lightning ML training and Optuna sweeps.

Updated Feb 6, 2026

Search Tags:#configuration management #hydra #optuna #pytorch lightning #hyperparameter sweeps #ml config #experiment composition

Ready to run

npx skills add https://github.com/nishide-dev/claude-code-ml-research --skill ml-config-manager

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: ml-config-manager
Source: https://github.com/nishide-dev/claude-code-ml-research/tree/main/skills/ml-config-manager
Command: npx skills add https://github.com/nishide-dev/claude-code-ml-research --skill ml-config-manager

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

ML experiments get slowed down when configuration files (model, data, trainer, logger, and sweeps) are hand-written inconsistently, causing Hydra composition issues and repeatable experiment setup to be error-prone.

Core Features & Use Cases

Hydra config generation for full ML workflows: produce structured config files for model, data, trainer, logger, and end-to-end experiment composition.
Hyperparameter sweep templates: create sweep configuration (including Optuna-based Bayesian/TPESampler setups) for systematic search and multirun execution.
Modular best-practice structure: guides consistent naming, DRY defaults composition, and resolution-friendly _target_ patterns for PyTorch Lightning and Hydra.

Quick Start

Tell the AI: “Create a Hydra config for a ResNet model on CIFAR-10 with a W&B logger and an Optuna hyperparameter sweep optimizing val/loss.”

Frequently Asked Questions about ml-config-manager

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I generate Hydra config files for PyTorch Lightning experiments?▼

To generate Hydra config files for PyTorch Lightning, specify your model, data, trainer, and logger requirements to produce structured YAML templates with valid defaults overrides and correct _target_ instantiation paths.

How do I set up an Optuna hyperparameter sweep with Hydra?▼

Setting up an Optuna hyperparameter sweep with Hydra involves creating sweep configuration templates using Bayesian or TPE samplers, enabling systematic multirun execution across model and data parameters for your training script.

Why does my Hydra experiment composition fail with inconsistent YAML configs?▼

Hydra experiment composition fails when YAML configs are hand-written inconsistently, but generating modular config hierarchies with DRY defaults and resolution-friendly _target_ patterns ensures valid composition.

Can I use Hydra defaults overrides to manage model, data, and logger configs separately?▼

Yes, you can use Hydra defaults overrides to manage configs separately by generating structured YAML files that align with Hydra's composition system for model, data, trainer, and logger components.

What is the best way to structure ML configs for repeatable experiments?▼

The best way to structure ML configs for repeatable experiments is applying modular best-practice naming, DRY defaults composition, and sweep-ready multirun configuration suitable for src/train.py execution.

Do I need PyTorch Lightning to use Hydra config generation?▼

Hydra config generation is designed for PyTorch Lightning and Hydra-based training projects, producing valid YAML templates aligned with correct _target_ instantiation paths for these specific frameworks.