Nodes/ComfyUI_FluxMod/FluxMod Mini Unified Model Loader
ComfyUI Node

FluxMod Mini Unified Model Loader

The FluxMod loader with an extra patch slot

By lodestone-rock·Created 2 years ago·Updated about a year ago· 123
FluxMod Mini Unified Model Loader
    • model
    unet_name
    guidance_name
    quant_mode
    lite_patch_unet_name

    Upfront honesty: this one's obscure enough that it doesn't show up in the README's own node table at all - that table lists FluxModCheckpointLoader, KSamplerMod, FluxModSamplerWrapper, and SkipLayerForward, and this "Mini" loader isn't among them. Zero search impressions so far bears that out; almost nobody's looking for it by name yet. What follows is grounded in what the node's schema actually exposes, not in documentation that doesn't exist for it.

    Structurally it's the same shape as the regular FluxModDiffusionLoader, plus one more field. You get unet_name (your base Flux Dev or Schnell checkpoint, .safetensors or .GGUF if ComfyUI-GGUF is installed), guidance_name (the modulation addon - universal_modulator.safetensors from lodestone-horizon/flux-essence), and quant_mode (bf16, float8_e4m3fn, or float8_e5m2), exactly as on the full loader. The addition is lite_patch_unet_name, an enum slot for a second, separately-selectable model file layered on top. Going by the name alone - and this is inference, not documented behavior - it reads as a slot for a smaller supplementary patch on top of the base checkpoint, in the same spirit as the modulation addon itself, rather than a second full model. Treat that as an educated guess rather than a fact until the README catches up.

    Inputs and outputs

    unet_name, guidance_name, quant_mode behave identically to the main FluxModDiffusionLoader - see that node's page for the fuller rundown of each. lite_patch_unet_name is the one new field; like guidance_name, it's populated from whatever files you drop in the relevant models folder, and defaults to "None" if nothing's selected. Output is a single model, same as every other loader in this pack - wire it into KSamplerMod if you picked a float8 quant_mode, or a regular KSampler if you're in bf16.

    How to install it

    Via ComfyUI Manager: search "ComfyUI_FluxMod" in Custom Nodes Manager, install, restart. Manually:

    cd ComfyUI/custom_nodes
    git clone https://github.com/lodestone-rock/ComfyUI_FluxMod.git
    

    Restart ComfyUI and refresh your browser. Requirements match the main loader: an official Flux Dev or Schnell checkpoint, the universal_modulator.safetensors addon, and ComfyUI-GGUF installed separately if you want GGUF files to appear in unet_name.

    Common issues & troubleshooting

    lite_patch_unet_name only shows "None" as an option. That field pulls from a models folder the same way unet_name and guidance_name do - if nothing else is populated, you likely haven't placed a compatible file where this node expects to find it, or no such file exists for your setup yet given how sparsely documented this node is.

    Same float8-quant sampler requirement as the main loader. If quant_mode is float8_e4m3fn or float8_e5m2, you still need KSamplerMod or FluxModSamplerWrapper downstream - this node doesn't change that rule.

    Not sure whether to use this or the regular FluxModDiffusionLoader. If you don't have a specific file to put in lite_patch_unet_name, there's no reason to reach for this variant - use the main loader and leave this one for whenever the extra patch slot is actually documented or you find a workflow that specifically calls for it.

    Categoryadvanced/FluxMod

    Inputs (4)

    NameTypeDefaultDescription
    unet_nameCOMBOThis loads both .safetensors and .GGUF models if ComfyUI-GGUF is installed.
    guidance_nameCOMBO1 options: None
    quant_modeCOMBO3 options: bf16, float8_e4m3fn (8 bit), float8_e5m2 (also 8 bit)
    lite_patch_unet_nameCOMBO1 options: None

    Outputs (1)

    NameTypeDescription
    modelMODEL