Instructions to use SceneWorks/flux1-dev-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use SceneWorks/flux1-dev-mlx with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("SceneWorks/flux1-dev-mlx", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - MLX
How to use SceneWorks/flux1-dev-mlx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir flux1-dev-mlx SceneWorks/flux1-dev-mlx
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Draw Things
- DiffusionBee
- Atomic Chat
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: flux-1-dev-non-commercial-license
|
| 4 |
+
license_link: https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
library_name: diffusers
|
| 8 |
+
base_model: black-forest-labs/FLUX.1-dev
|
| 9 |
+
tags:
|
| 10 |
+
- mlx
|
| 11 |
+
- apple-silicon
|
| 12 |
+
- flux
|
| 13 |
+
- quantized
|
| 14 |
+
- text-to-image
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
# FLUX.1-dev — MLX quant matrix (bf16 / Q8 / Q4)
|
| 18 |
+
|
| 19 |
+
Pre-quantized, **MLX-ready** repackagings of [`black-forest-labs/FLUX.1-dev`](https://huggingface.co/black-forest-labs/FLUX.1-dev)
|
| 20 |
+
for the [SceneWorks](https://github.com/SceneWorks/SceneWorks) native Apple-Silicon worker (`mlx-gen`).
|
| 21 |
+
Each tier is a complete, self-contained turnkey snapshot that loads directly with no in-app
|
| 22 |
+
conversion peak (epic 8506).
|
| 23 |
+
|
| 24 |
+
| Tier | Subdir | Approx. size |
|
| 25 |
+
|--------|---------|--------------|
|
| 26 |
+
| Q4 | `q4/` | ~8.7 GB |
|
| 27 |
+
| Q8 | `q8/` | ~17 GB |
|
| 28 |
+
| bf16 | `bf16/` | ~31 GB |
|
| 29 |
+
|
| 30 |
+
All four components are packed in Q4/Q8 — the DiT transformer, the CLIP + T5 text encoders, and the
|
| 31 |
+
VAE's mid-block attention — using plain asymmetric group-affine quantization (group size 64),
|
| 32 |
+
byte-identical to the worker's load-time quantization. The bf16 tier is the dense source, mirrored.
|
| 33 |
+
|
| 34 |
+
## License & attribution — NON-COMMERCIAL
|
| 35 |
+
|
| 36 |
+
**These weights are governed by the FLUX.1 [dev] Non-Commercial License v1.1.1** (see `LICENSE.md`),
|
| 37 |
+
inherited unchanged from the upstream Black Forest Labs release. This repository only re-packages the
|
| 38 |
+
weights (quantization + MLX layout) and adds no new training. Use is permitted for non-commercial
|
| 39 |
+
purposes only, per that license. © Black Forest Labs. Original model card:
|
| 40 |
+
https://huggingface.co/black-forest-labs/FLUX.1-dev
|