Load Tuner Data

LoRA AutoTuner's whole value proposition is that it does an expensive sweep once and hands you the winning config. But "once" only stays true if you can get that result back later without re-running the sweep - that's what this node is for. It loads a previously saved .tuner/.json file from disk and hands it back as TUNER_DATA, ready to feed into a Merge Selector.

What it is and why you'd use it

Pair it mentally with Save Tuner Data: you run the AutoTuner once (which can be genuinely slow on a big stack), save the result, and from then on Load Tuner Data + Merge Selector reproduces that exact ranked outcome in seconds, with no sweep at all. It's also how you share or archive a tuning result - a saved tuner file is portable in a way a live AutoTuner run in your graph isn't.

How it works

It's a simple file loader with a dropdown. Point it at a saved tuner data file (from the folder LoRA Optimizer's tuner-data nodes use), and it deserializes the ranked configs back into a TUNER_DATA object. Per the README, it auto-reloads if the underlying file changes on disk, so if you overwrite a saved tuner file with a fresh sweep, you don't need to restart ComfyUI to pick up the change.

The inputs and outputs that matter

• tuner_data_file - a dropdown of saved tuner data files. That's the only input; there's no model or LoRA stack connection here, because this node isn't running anything, just reading a file.

Outputs: tuner_data - feed straight into a Merge Selector's tuner_data input to apply a specific ranked config. prompt, description, and metadata_info - the same trio LoRA Metadata Reader exposes, here reading whatever context was saved alongside the tuner data rather than off a LoRA file. Useful for reminding yourself what a saved config was actually for six weeks later, without having to open the raw file.

How to install it

ComfyUI Manager: search "LoRA Optimizer", install, restart. Manually:

cd ComfyUI/custom_nodes/
git clone https://github.com/ethanfel/ComfyUI-LoRA-Optimizer.git

Restart - nodes land under loaders. No downloads; this node reads files you already saved from a previous AutoTuner run.

Common issues & troubleshooting

The dropdown is empty. You haven't saved any tuner data yet. Run LoRA AutoTuner once, use Save Tuner Data to write the result to disk, and this dropdown will populate on the next graph refresh.

Loaded data doesn't match your current stack. This node doesn't validate that a saved tuner result still applies to whatever lora_stack you're currently feeding into Merge Selector - that's on you to keep straight. A tuner file saved for one combination of LoRAs won't necessarily produce a sensible merge if you've since swapped one of the files out; re-run the AutoTuner if your stack has changed.

You want the latest result, not an old save. If you're actively iterating and re-running the AutoTuner each time, you likely don't need this node at all - connect the AutoTuner's own tuner_data output straight to Merge Selector. Load Tuner Data earns its place once you're done iterating and want to reuse a settled result across sessions without paying for the sweep again.