Welcome to the sd-optim Wiki!

This wiki provides detailed documentation, guides, and explanations for using sd-optim. The goal is to centralize information that was previously scattered or included in configuration file comments.

Overview

sd-optim is an opinionated tool for optimizing model merges using various optimization methods/libraries like Optuna or BayesianOptimization with the help of the powerful sd-mecha merging library.

Navigation

• Getting Started: Basic setup and installation. (TODO)
• Configuration: Detailed explanation of config.yaml and optimization_guide.yaml.
• Optimization Strategies: Guidance on choosing optimizer settings, samplers, and acquisition functions.
• Scoring: Information about available image scorers and their configuration.
• Merge Methods: Documentation for supported sd-mecha merge methods and custom methods. (TODO: Link or add content from mm_docs.md)
• Advanced Usage: Tips for advanced scenarios like recipe optimization. (TODO)