CFE Lora Params

CFE Lora Params is a collector node: you feed it a list of LoRAs with strengths, and it bundles them into a single LORA_PARAMS object. That's the honest description, and here's the honest catch - nothing else in this pack can consume that object. There's no "CFE Lora Applier" waiting for it, no node downstream that reads LORA_PARAMS. Right now it's a partial utility that only makes sense inside the author's private workflow, which the README flat-out admits exists ("just for my own workflow").

The source code tells the story. A commented-out import at the top of the pack (from .mine_nodes import *) strongly suggests the author keeps a private node set that this thing was built to feed. It also tells you where the code came from: the file header says "Most of this was stolen from rgthree Power Lora Loader, because it was awesome" - and it's true, the fuzzy LoRA-filename matching is lifted straight from there.

How it works

The node uses a flexible input type, which means the UI shows you a growing list of LoRA entries you can add to. Each entry carries four fields the code looks for:

• on - a toggle; the LoRA only gets collected when it's on.
• lora - the filename, matched fuzzy against your models/loras folder (exact path, basename, no-extension - it'll find it).
• strength - the model strength.
• strengthTwo - optional clip strength; if you leave it out, the code reuses strength for both, exactly like rgthree's loader.

It filters the inputs, skips anything with zero strength, and outputs a dict of {"loras": [...], "strengths": [[model, clip], ...]}.

Install

Unremarkable for this pack - no dependencies at all:

cd ComfyUI/custom_nodes
git clone https://github.com/CpreForEver/CFE_comfyui

Restart, and it's under CFE/loras. Or ComfyUI Manager → CFE_comfyui.

Should you use it?

For a beginner: probably not, and I'd rather tell you that than let you burn an afternoon. Because the output type has no consumer in this pack, the node is dead weight unless you're building your own custom node that understands LORA_PARAMS, or the author wires up an applier in a future commit. If your goal is "load a LoRA on Flux," use the stock LoraLoader, or install rgthree-comfy and use the actual Power Lora Loader this one was copied from - that one does the whole job (stacking, toggling, trigger words, and applying to your model) instead of just preparing the shopping list. This node is one half of a feature the pack hasn't shipped yet. File it under "interesting to watch, useless today," and check back after the author's next commit.