New: Cookbooks and AI ExplanationsStep-by-Step recipes to solve problems connected to Roadmaps and Cheat Sheets. Need more details? Use AI buttons for structured and simple explanations with concrete examples throughout the whole platform.Take a look
I fine-tuned one adapter for tone and another for a skill, and now I need both without training a third model from scratch.
What you'll have at the end
One merged model that keeps both adapters' behaviors, checked by testing it against each adapter's own held-out examples separately
You need
Two LoRA adapters already trained on the exact same base model checkpoint, one for tone and one for a skill, each with its own held-out examples you can check the merged result against.
Not covered
Merging adapters that were trained on different base model checkpoints, since the trim-elect-merge math has no way to reconcile two different starting points.
Leans on
Pick starting values for LoRA rank, alpha, and learning rate
if the two adapters you want to merge don't share a rank yet and you need to pick one before retraining either
Most serving stacks refuse a bare LoRA adapter
for what to do if your serving stack still won't accept the merged adapter as a single file after folding it into the base weights
Checked 18 Aug 2026
Part of the LLM Fine-tuning cookbook