Nodes/Ray's Orenchi/🧰 Ray's Model Tools: LoRA QKV Loader
ComfyUI Node

🧰 Ray's Model Tools: LoRA QKV Loader

Load and stack LoRAs with +, trash and a strength per row. Automatically checks the connected model's native parameter map, repairs unambiguous trainer naming and supported fused/separate QKV mismatches, and otherwise uses native adapters. Applies to the MODEL only. Zero strength skips a row; an empty list passes the model through. Incompatible or unsafe repairs are reported.

By MutantSparrowΒ·Created 4 months agoΒ·Updated 3 days agoΒ· 1
🧰 Ray's Model Tools: LoRA QKV Loader
  • model
  • MODEL
β—„lora_nameβ–Ί
β—„strength1.00β–Ί
β—„additional_loras[]β–Ί
CategoryπŸ‘‘ Ray/🧰 Model Tools

Inputs (4)

NameTypeDefaultDescription
modelMODELDiffusion model to patch. The selected LoRAs are applied in row order to a clone; text encoders are not modified.
lora_nameSTRINGSelect a registered LoRA from the searchable folder dropdown. Search includes every subfolder.
strengthFLOAT1.00-100–100First LoRA's model-strength multiplier. Zero skips it; negative values reverse its direction. Each additional row has its own strength.
additional_lorasSTRING[]Additional LoRA rows with independent strengths, applied in order.

Outputs (1)

NameTypeDescription
MODELMODELCloned model with compatible LoRAs applied in row order at their individual strengths. The original model and text encoders are unchanged.