ComfyUI Node

GGUF_LLM

A ComfyUI node in ListHelper/LLM with 15 inputs and 3 outputs.

By dseditor·Created about a year ago·Updated 6 months ago· 71
GGUF_LLM
  • image_1
  • image_2
  • image_3
  • text
  • used_seed
  • prompts
modelDownload: QwenVL
prompt
prompt_templateimage_to_prompt.md
system_prompt
max_tokens3072
temperature0.7
top_p0.90
top_k40
seed0
keep_model_loadedfalse
mmproj_fileDownload: QwenVL mmproj
auto_install_llama_cpptrue
CategoryListHelper/LLM

Inputs (15)

NameTypeDefaultDescription
modelCOMBODownload: QwenVL6 options: Download: Z-Image, Download: Z-Image (Abliterated), Download: Qwen, Download: Qwen (Abliterated), Download: QwenVL, Download: QwenVL(2B)
promptSTRING
prompt_templateCOMBOimage_to_prompt.md8 options: Custom, extract_person_features.md, image_to_prompt.md, photography_en.md, photography_zh.md, qwen2512_en.md, +2
system_promptSTRING
max_tokensINT30721–8192
temperatureFLOAT0.70–2
top_pFLOAT0.900–1
top_kINT400–100
seedoptINT00–18446744073709550000
keep_model_loadedoptBOOLEANfalseKeep model in memory after inference
mmproj_fileoptCOMBODownload: QwenVL mmprojVision model mmproj file (auto-enabled when any image is provided and model is VL type)
image_1optIMAGEFirst input image for vision model (auto-enables vision mode for VL models)
image_2optIMAGESecond input image for vision model (optional, for multi-image analysis)
image_3optIMAGEThird input image for vision model (optional, for multi-image analysis)
auto_install_llama_cppoptBOOLEANtrueAuto-install/repair llama-cpp-python (supports all platforms/CUDA versions, prioritizes Basic version for compatibility)

Outputs (3)

NameTypeDescription
textSTRING
used_seedINT
promptsSTRING