Nodes/ComfyUI-FatMex-Nodes/Fat Mex Model Loader
ComfyUI Node

Fat Mex Model Loader

All-in-one model loader. Pick a preset to load the correct UNET, CLIP, and VAE. Optionally stack up to 2 LoRAs. Supports all Fat Mex model families: Klein 9B, Qwen Image Edit, Z-Image Turbo, Chroma HD, etc.

By FatMex·Created 6 months ago·Updated 6 months ago· 3
Fat Mex Model Loader
    • model
    • clip
    • vae
    presetKlein 9B
    sage_attentiontrue
    lora_1none
    lora_1_strength1.00
    lora_2none
    lora_2_strength1.00
    weight_dtype_overridedefault
    CategoryFat Mex

    Inputs (7)

    NameTypeDefaultDescription
    presetCOMBOKlein 9BModel preset to load. Each preset defines the UNET, CLIP, VAE, and default sampling settings.
    sage_attentionoptBOOLEANtrueEnable sage attention for faster inference (requires sageattention package).
    lora_1optCOMBOnoneFirst LoRA to apply to the model.
    lora_1_strengthoptFLOAT1.00-10–10Strength of the first LoRA.
    lora_2optCOMBOnoneSecond LoRA to apply to the model.
    lora_2_strengthoptFLOAT1.00-10–10Strength of the second LoRA.
    weight_dtype_overrideoptCOMBOdefaultOverride weight data type. 'default' uses the preset's setting.

    Outputs (3)

    NameTypeDescription
    modelMODELThe loaded diffusion model.
    clipCLIPThe CLIP text encoder.
    vaeVAEThe VAE for encoding/decoding images.