Extensions/ComfyUI-QuantOps
ComfyUI Extension

ComfyUI-QuantOps

[DEPRECATED]

By silveroxides·Created 8 months ago·Updated a day ago· 55
silveroxides/ComfyUI-QuantOps
Nodes10
On cloudLocal install
Categoryloaders/quantized
Stars55
Updateda day ago

Nodes (10)

Load Diffusion Model (BNB 4-bit)

BNB NF4/FP4 UNETs without the bitsandbytes pain

loaders/quantized
Load VAE (No Dynamic VRAM)

The VAE loader that skips ComfyUI's dynamic VRAM paging — and why you might want that

loaders/quantized
Load CLIP (Quantized)

Your text encoder is now the VRAM bottleneck — load a quantized one

loaders/quantized
Load CLIP (Quantized, Simple)

The no-thinking version of Load CLIP (Quantized)

loaders/quantized
Load DualCLIP (Quantized)

Two quantized text encoders at once, from CLIP-L + T5 to Qwen + Gemma

loaders/quantized
Load DualCLIP (Quantized, Simple)

The set-and-forget dual text encoder loader

loaders/quantized
Load Checkpoint (Quantized)

Load a quantized checkpoint without knowing which layer is which format

loaders/quantized
Load Checkpoint (Quantized, Simple)

The 'just load my quantized checkpoint' loader

loaders/quantized
Load Diffusion Model (Quantized)

Load a quantized diffusion model in the format you actually want

loaders/quantized
Load Diffusion Model (Quantized, Simple)

Auto-detect your quantized UNET and get on with it

loaders/quantized
Readme

ComfyUI-QuantOps IS NO LONGER NEEDED SO IT HAS BEEN DEPRECATED

Support for int8 ConvRot quants is built into ComfyUI now

Extended quantization layouts for ComfyUI, enabling loading and inference with models quantized by convert_to_quant.

This is experimental and due to lack of proper support and merging of PR in ComfyUI, do not expect this to work without putting in the effort. I don't have the time or the energy to keep this up and will close entire project if i keep getting bunch of low effort issues posted expecting me go serve a fix up on a silver platter.

tl;dr Go complain at ComfyOrg. Not here.

The following is the last update I make regarding this.

In order to use int8_tensorwise(RTX 30xx-series or newer GPU) you will need the following:

  • torch 2.10+cu130 or higher
  • installed the latest of my custom comfy-kitchen fork wheels with the int8-tensorwise support
  • enable the use of triton backend by using --enable-triton-backend launch argument in ComfyUI

Step 1: Install Triton Activate your virtual environment used by ComfyUI and install triton. For Windows you need to use this but linux can install latest triton as usual.

# for torch 2.10 and 2.11
pip install -U "triton-windows<3.7"
# for torch 2.12
pip install -U "triton-windows<3.8"

Step 3: Install my comfy-kitchen Download the latest uploaded version matching you python of my pre-compiled .whl file from my HuggingFace repository (Latest as of 13 June 2026)

Install it directly pointing to the file path:

pip install --no-deps --force-reinstall --no-cache-dir "path/to/comfy-kitchen.whl"

Step 4: Install/Update ComfyUI-QuantOps You just need to ensure it's fully up to date to read the new model formats. Run these commands:

cd custom_nodes/ComfyUI-QuantOps
git pull

When launching Comfyui add launch argument:

--enable-triton-backend

You can get most of the models here: https://huggingface.co/silveroxides

License

MIT License

Acknowledgements