Accept the FLUX.1 [dev] Non-Commercial License to access these files

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

These files are a Derivative of FLUX.1 Fill [dev] and are licensed by Black Forest Labs Inc. under the FLUX.1 [dev] Non-Commercial License (LICENSE.md in this repository): non-commercial use only, subject to its usage restrictions and content-filtering requirements. Your rights to use them are granted directly by Black Forest Labs under that license. For commercial use, get a license from Black Forest Labs (bfl.ai).

FLUX.1 Fill Dev GGUF

GGUF builds of FLUX.1 Fill Dev, made by Moonbox for on-device use. Each precision is one file at the root of this repository; pick one by the memory you have. Only the model's own weights are converted: the other parts of the pipeline (text encoder, VAE, tokenizer) come from the original repository unless the section below says otherwise.

Files

The last column is how far each file is from the original BF16 weights: the relative RMS error of its dequantized weights, averaged over all parameters. Image-quality scores against BF16 outputs are added as each precision passes the image eval.

Precision	File	Size	sha256	Checked against BF16
F16	`flux-1-fill-dev-F16.gguf`	23.81 GB	`9e3477a55e42`	0.0000007 % (6.8e-9)
BF16	`flux-1-fill-dev-BF16.gguf`	23.81 GB	`cb2cd4851402`	exact
Q8_0	`flux-1-fill-dev-Q8_0.gguf`	12.72 GB	`bf51a9ae3c7a`	0.55 %
Q6_K	`flux-1-fill-dev-Q6_K.gguf`	9.85 GB	`977f6ae24a71`	1.8 %
Q5_K_M	`flux-1-fill-dev-Q5_K_M.gguf`	8.42 GB	`abf0fb9d3b05`	3.4 %
Q4_K_M	`flux-1-fill-dev-Q4_K_M.gguf`	6.93 GB	`bba7d531b663`	6.8 %

How to run

These files are the FLUX.1 Fill [dev] transformer only (12 B parameters, flux1-fill-dev.safetensors converted as is). FLUX.1 Fill fills a masked region of a picture from a text prompt (inpainting and outpainting). Text encoders and VAE are separate, the same as FLUX.1 [dev]'s:

VAE: ae.safetensors from black-forest-labs/FLUX.1-Fill-dev.
Text encoders: clip_l.safetensors and t5xxl_fp16.safetensors (or t5xxl_fp8_e4m3fn.safetensors) from comfyanonymous/flux_text_encoders.

stable-diffusion.cpp (we ran master-1091, commit b9d7176). The mask is white where the picture is repainted:

sd-cli --diffusion-model flux-1-fill-dev-Q4_K_M.gguf --vae ae.safetensors --clip_l clip_l.safetensors \
  --t5xxl t5xxl_fp16.safetensors -i picture.png --mask mask.png -p "a glowing golden full moon with craters" \
  --cfg-scale 1.0 --guidance 30 --sampling-method euler --steps 28 -W 1024 -H 1024 --diffusion-fa

Tested here: Q4_K_M inpainted a 1024 × 1024 picture (20 steps) in about a minute end to end on an RTX 3080 Laptop GPU (16 GB) with 32 GB of RAM, adding --mmap --max-vram 14 to the command above: 36 s of sampling, 14 s of text encoding. Add --strength 1.0 to repaint the masked region from scratch (sd-cli's default, 0.75, keeps some of the original's colours).

FLUX.1 Fill is guidance-distilled: CFG 1 with a high distilled guidance (30, as in Black Forest Labs' example).

ComfyUI with ComfyUI-GGUF: load it with Unet Loader (GGUF) (tagged flux) and use the stock FLUX.1 Fill inpainting workflow (InpaintModelConditioning, FluxGuidance 30).

Which file: Q8_0 is the closest to the original at about half the size; Q6_K and Q5_K_M trade a little fidelity for memory; Q4_K_M is the smallest. BF16 / F16 are the unquantized weights.

License. The FLUX.1 [dev] Non-Commercial License (LICENSE.md and NOTICE.md in this repository): non-commercial use only. These files are not an official product of Black Forest Labs. For commercial use, get a license from Black Forest Labs.

How these were made

From FLUX.1 Fill Dev at revision 358293da03: the original BF16 weights converted to GGUF F16 / BF16, then quantized with llama.cpp's llama-quantize. Every file was checked against the original before upload; the sha256 column above is what the files must hash to.

License

FLUX.1-dev Non-Commercial (gated). These files are a derivative of the original model and carry its license: it is copied in this repository, and it governs your use of these files. The original model card follows unchanged.

Original model card

FLUX.1 Fill [dev] is a 12 billion parameter rectified flow transformer capable of filling areas in existing images based on a text description. For more information, please read our blog post.

Key Features

Cutting-edge output quality, second only to our state-of-the-art model FLUX.1 Fill [pro].
Blends impressive prompt following with completing the structure of your source image.
Trained using guidance distillation, making FLUX.1 Fill [dev] more efficient.
Open weights to drive new scientific research, and empower artists to develop innovative workflows.
Generated outputs can be used for personal, scientific, and commercial purposes as described in the FLUX.1 [dev] Non-Commercial License.

Usage

We provide a reference implementation of FLUX.1 Fill [dev], as well as sampling code, in a dedicated github repository. Developers and creatives looking to build on top of FLUX.1 Fill [dev] are encouraged to use this as a starting point.

API Endpoints

The FLUX.1 models are also available in our API bfl.ml

Diffusers

To use FLUX.1 Fill [dev] with the 🧨 diffusers python library, first install or upgrade diffusers

pip install -U diffusers

Then you can use FluxFillPipeline to run the model

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")

pipe = FluxFillPipeline.from_pretrained("black-forest-labs/FLUX.1-Fill-dev", torch_dtype=torch.bfloat16).to("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")

To learn more check out the diffusers documentation

Limitations

This model is not intended or able to provide factual information.
As a statistical model this checkpoint might amplify existing societal biases.
The model may fail to generate output that matches the prompts.
Prompt following is heavily influenced by the prompting-style.
There may be slight-color shifts in areas that are not filled in
Filling in complex textures may produce lines at the edges of the filled-area.

Out-of-Scope Use

The model and its derivatives may not be used

In any way that violates any applicable national, federal, state, local or international law or regulation.
For the purpose of exploiting, harming or attempting to exploit or harm minors in any way; including but not limited to the solicitation, creation, acquisition, or dissemination of child exploitative content.
To generate or disseminate verifiably false information and/or content with the purpose of harming others.
To generate or disseminate personal identifiable information that can be used to harm an individual.
To harass, abuse, threat

Downloads last month: -

GGUF

Model size

12B params

Architecture

flux

Hardware compatibility

4-bit

5-bit

6-bit

8-bit

16-bit

Model tree for moonbox-dev/FLUX.1-Fill-dev-GGUF

Base model

black-forest-labs/FLUX.1-Fill-dev

Quantized

(25)

this model