import torch
from diffusers import FluxFillPipeline
from diffusers.utils import load_image
image = load_image("https://huggingface.co/datasets/diffusers/diffusers-images-docs/resolve/main/cup.png")
mask = load_image("https://huggingface.co/datasets/diffusers/diffusers-images-docs/resolve/main/cup_mask.png")
# switch to "mps" for apple devices
pipe = FluxFillPipeline.from_pretrained("SamuelTallet/FLUX.1-Fill-dev-SDNQ-4bit-dynamic-hadamard256", dtype=torch.bfloat16, device_map="cuda")
image = pipe(
prompt="a white paper cup",
image=image,
mask_image=mask,
height=1632,
width=1232,
guidance_scale=30,
num_inference_steps=50,
max_sequence_length=512,
generator=torch.Generator("cpu").manual_seed(0)
).images[0]
image.save(f"flux-fill-dev.png")This is FLUX.1 Fill [dev] optimized using SDNQ with UINT4 dynamic quantization and Hadamard Rotation (Group size: 256).
Sample
| Input | Mask | Output |
|---|---|---|
 |
 |
 |
Prompt:
a white paper cup
Seed: 0
Usage
Install Torch, Diffusers (Git), SDNQ 0.2.0+ and Triton.
Use CFG 30 and 50 steps.
- Downloads last month
- 35
Model tree for SamuelTallet/FLUX.1-Fill-dev-SDNQ-4bit-dynamic-hadamard256
Base model
black-forest-labs/FLUX.1-Fill-dev