nunchaku

nunchaku copied to clipboard

Any possibility to get int4 for Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro

2

I see int4 is available for Fill, Canny, Depth, etc.. https://huggingface.co/collections/mit-han-lab/svdquant-67493c2c2e62a1fc6e93f45c Please also provide for ControlNet-Union-Pro

nitinmukesh

[Bug] turbo lora only support dev, not support fill

6

lora link: https://huggingface.co/alimama-creative/FLUX.1-Turbo-Alpha/tree/main dev work with FLUX.1-Turbo-Alpha.lora  fill no work  Loading configuration from F:\ComfyUI\ComfyUI\models\diffusion_models\svdq-int4-flux.1-fill-dev\comfy_config.json model_type FLUX !!! Exception during processing !!! Traceback (most recent call last): File "F:\ComfyUI\ComfyUI\execution.py",...

xueqing0622

Make layers (specifically the first normalization layer) public

4

Some acceleration techniques, such as [TeaCache](https://github.com/ali-vilab/TeaCache/tree/main/TeaCache4FLUX) rely on having access for inference to individual layers in the FLUX transformer. To make them interoperable with Nunchaku, the Nunchaku library should export...

andreaferretti

PuLID with Nunckaku

1

Hi, can PuLID be used with svdquant (Nunckaku) in ComfyUI please? Thanks to let us know.

OCOMATA

fp4用controlnet union 确认是官方的fp16，但是一到采样器就会自动退出，感谢大佬

1

 fp4用controlnet union 确认是官方的fp16，但是一到采样器就会自动退出，感谢大佬解答支持

qiujingjing

linux python=3.10.15, torch2.6, uses Nunchaku-0.1.4 %2Btorch2.6-cp310-cp310-linux_x86_64.whl installed nunchaku, but the plugin import still failed

2

The comfyui execution background information is as follows： [START] Security scan [DONE] Security scan ## ComfyUI-Manager: installing dependencies done. ** ComfyUI startup time: 2025-04-08 22:21:53.837 ** Platform: Linux ** Python...

nongyu-gpu

Injecting quantized module !!! Exception during processing !!! CUDA error: out of memory Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

2

please help whith this, i tried reinstall nunchaku and deepcompressor, just the same error Using xformers attention in VAE VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16 2025-04-07 01:58:30.960384...

usupehmat

Error during inference: diffusers.configuration_utils.ConfigMixin.load_config() got multiple values for keyword argument 'cache_dir'

2

I want to store the models in a specific folder instead of default huggingface cache. getting error with this code Windows 11 ```python transformer_repo = "mit-han-lab/svdq-int4-flux.1-dev" base_model_path = "models/Flux.1-dev" print("Here0")...

nitinmukesh

Implement batch

1

I implemented the Batch function. Could you please review the implementation? @ita9naiwa

Bluear7878

cannot import name 'to_diffusers' from 'nunchaku.lora.flux'

4


LoserCoderLi