when will "unquantized_layers.safetensors" and "config.json" for kontext-dev be available? Thank you!!!!!
as the title
No need for these files starting from nunchaku v0.3.0.
Just download the model and load its path, and then you can use it.
Appreciate it!
I am using "NunchakuFluxTransformer2dModel" to load it and it requires a "config.json", "unquantized_layers.safetensors", and "transformer_blocks.safetensors". Are there any new pipeline which can load it like how it does for Flux.1-dev or is this just Comfyui:
"
from diffusers import FluxPipeline
from nunchaku.models.transformer_flux import NunchakuFluxTransformer2dModel
transformer = NunchakuFluxTransformer2dModel.from_pretrained("mit-han-lab/svdq-int4-flux.1-dev")
pipeline = FluxPipeline.from_pretrained(
"black-forest-labs/FLUX.1-dev", transformer=transformer, torch_dtype=torch.bfloat16
).to("cuda")"
When will the SVDQuant version come out? Thank you!
You can use this branch: https://github.com/mit-han-lab/nunchaku/tree/feat/kontext
I will merge it later today.
The usage: https://github.com/mit-han-lab/nunchaku/blob/feat/kontext/examples/flux.1-kontext-dev.py
Thank you very much!
0.3.3 installed + nunchaku-0.3.2.dev20250630+torch2.7-cp312-cp312-win_amd64 and everytime black screen