mergekit — Skill

What problem does it solve?

It solves the problem of combining multiple pretrained LLM checkpoints, fine-tunes, or LoRA adapters into a single merged .safetensors artifact without doing full retraining.

Core Features & Use Cases

• Out-of-core tensor streaming merges: run merges without loading entire models into RAM by streaming tensors layer-by-layer.
• Task-vector and delta-aware merge methods: apply methods such as linear blending, slerp/nuslerp interpolation, task arithmetic, TIES/DARE/DELLA, and breadcrumbs-based pruning for interference reduction.
• Frankenmerging via layer slices: assemble a model by selecting and optionally blending contiguous layer ranges from different sources.
• Tokenizer and LoRA handling: configure tokenizer sources and extract LoRA adapters approximating finetune deltas.

Quick Start

Use mergekit to merge two LoRA- or fine-tuned checkpoints into a single .safetensors model by telling the AI to run the mergekit-yaml CLI on a provided YAML config that specifies the merge method, base model (when required), input model paths, and output directory.