Nodes/VRGameDevGirl Video Enhancement Nodes/🧠 VRGDG SuperGemma GGUF Chat 🧠
ComfyUI Node

🧠 VRGDG SuperGemma GGUF Chat 🧠

A ComfyUI node in VRGDG/LLM with 40 inputs and 3 outputs.

By vrgamegirl19·Created about a year ago·Updated 22 days ago· 533
🧠 VRGDG SuperGemma GGUF Chat 🧠
  • image1
  • image2
  • image3
  • image4
  • image5
  • image6
  • image7
  • image8
  • image9
  • image10
  • image11
  • image12
  • image13
  • image14
  • image15
  • image16
  • image17
  • image18
  • image19
  • image20
  • image21
  • image22
  • image23
  • image24
  • text
  • used_model
  • status
◄model_file[No Gemma GGUF found in models/LLM]►
◄mmproj_file[No mmproj GGUF found in models/LLM]►
◄task_presettext_to_image►
◄custom_instructions►
◄user_input►
◄trigger_word►
◄image_count0►
◄advancedfalse►
◄unload_after_runfalse►
◄n_ctx262144►
◄n_gpu_layers99►
◄n_threads8►
◄chat_format►
◄temperature0.60►
◄top_p0.95►
◄max_new_tokens800►
CategoryVRGDG/LLM

Inputs (40)

NameTypeDefaultDescription
model_fileCOMBO[No Gemma GGUF found in models/LLM]Gemma GGUF models found under ComfyUI/models/LLM. Only .gguf files with 'gemma' in the name are shown.
mmproj_fileCOMBO[No mmproj GGUF found in models/LLM]mmproj GGUF files found under ComfyUI/models/LLM. Required only when using image inputs.
task_presetCOMBOtext_to_imageSelect a task preset with built-in instructions.
custom_instructionsSTRINGUsed only when task_preset is custom. Enter your own full instruction block.
user_inputSTRINGYour task details and creative direction for the selected preset.
trigger_wordSTRINGOptional LoRA/training trigger token. Used only by Captioner preset.
image_countINT00–24How many optional image inputs to show on the node.
advancedBOOLEANfalseShow advanced GGUF runtime controls such as context, GPU layers, threads, sampler, and token limits.
unload_after_runBOOLEANfalseIf enabled, unload the GGUF model from cache after this run to free RAM/VRAM.
n_ctxINT262144512–262144GGUF context window. SuperGemma defaults to the full 262144-token training context; lower this if RAM/VRAM is too high.
n_gpu_layersINT99-1–200How many layers to offload to GPU. Use -1 to offload all supported layers.
n_threadsINT81–128CPU threads used by llama.cpp.
chat_formatSTRINGOptional llama.cpp chat format override. Leave blank to use the model default.
temperatureFLOAT0.600–2Higher = more creative variation, lower = more deterministic output.
top_pFLOAT0.950–1Nucleus sampling cutoff.
max_new_tokensINT80032–32000Maximum number of output tokens.
image1optIMAGEOptional reference image input.
image2optIMAGEOptional reference image input.
image3optIMAGEOptional reference image input.
image4optIMAGEOptional reference image input.
image5optIMAGEOptional reference image input.
image6optIMAGEOptional reference image input.
image7optIMAGEOptional reference image input.
image8optIMAGEOptional reference image input.
image9optIMAGEOptional reference image input.
image10optIMAGEOptional reference image input.
image11optIMAGEOptional reference image input.
image12optIMAGEOptional reference image input.
image13optIMAGEOptional reference image input.
image14optIMAGEOptional reference image input.
image15optIMAGEOptional reference image input.
image16optIMAGEOptional reference image input.
image17optIMAGEOptional reference image input.
image18optIMAGEOptional reference image input.
image19optIMAGEOptional reference image input.
image20optIMAGEOptional reference image input.
image21optIMAGEOptional reference image input.
image22optIMAGEOptional reference image input.
image23optIMAGEOptional reference image input.
image24optIMAGEOptional reference image input.

Outputs (3)

NameTypeDescription
textSTRING—
used_modelSTRING—
statusSTRING—