ComfyUI Extension: LoRA Range Loader

Authored by esp-dev

Created

Updated

0 stars

Run ComfyUI workflows without the setup

No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.

ComfyUI custom node: load a LoRA from current selection and automatically advance current within a first/last filename range.

Looking for a different extension?

Custom Nodes (1)

README

comfyui-lora-range-loader (ComfyUI)

LoRA Range Loader icon

Custom node for ComfyUI that loads a LoRA like the core LoraLoaderModelOnly, but adds a filename range sweep workflow for testing many LoRA checkpoints from the same training run.

What this node does

  • Adds the node Load LoRA Range (Model Only).
  • Loads the LoRA selected in current exactly like the core model-only LoRA loader.
  • Adds first and last fields to define an allowed filename range.
  • Adds mode to control how current changes after each queued run.
  • Supports these modes:
    • fixed: keep current unchanged
    • increment: move to the next filename in the sorted first..last range and wrap to first
    • decrement: move backward in the sorted first..last range and wrap to last
    • random: pick a random filename from the sorted first..last range

How it works

The node is split into two parts:

  • Backend Python loads only the LoRA selected in current.
  • Frontend JavaScript updates current after the workflow is queued.

This means the LoRA used for the current execution is always the value visible in current at queue time, and the next value is prepared for the following run.

Range rules

  • The list of LoRAs is sorted by filename in ascending order.
  • first must come before or equal to last in that sorted list.
  • If first comes after last, automatic switching is disabled.
  • If current is outside the first..last range, the node still loads that LoRA, but increment, decrement, and random do not change it.

LoRA loading behavior

  • The node reuses the core LoraLoader implementation.
  • It does not load first or last into memory.
  • It only loads current.
  • It inherits the same single-entry cache as the core loader, so repeating the same current value does not reload the LoRA file unnecessarily.

Compatibility

Backend compatibility:

  • The Python node is lightweight and relies only on ComfyUI core APIs.
  • No extra Python packages are required.

Frontend compatibility:

  • Full automatic switching of current requires a modern ComfyUI frontend that supports widget queue lifecycle hooks.
  • If the frontend extension does not run, the backend node still works as a manual current LoRA loader.

How to use

  1. Add Load LoRA Range (Model Only) to the graph.
  2. Connect your MODEL input.
  3. Choose first, last, and current from the LoRA dropdowns.
  4. Choose mode.
  5. Run the workflow repeatedly.
  6. Observe current advancing for the next run according to mode.

Install via ComfyUI-Manager

  1. Open Manager in ComfyUI.
  2. Install this node from the list or by repository URL.
  3. Restart ComfyUI or refresh node definitions.

Manual install

  1. Clone or copy this repository into ComfyUI/custom_nodes/.
  2. Restart ComfyUI.

No additional Python dependencies are needed.

Included files

Notes for publishing

  • The Registry package name in pyproject.toml is intended to be stable and should be treated as immutable after publishing.
  • If you want a different final GitHub repository URL or package name, change it before the first Registry publish.

Run ComfyUI workflows without the setup

No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.

Learn more