Instructions to use fal/FLUX.2-Tiny-AutoEncoder with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use fal/FLUX.2-Tiny-AutoEncoder with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("fal/FLUX.2-Tiny-AutoEncoder", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Synthetic Data?
#2
by yukiarimo - opened
Was FLUX 2 trained on synthetic data?
Hello!
This repository is not for FLUX.2, but instead for an autoencoder that is compatible with FLUX.2. The training details of the main model would be available through BFL, though I'm not sure what they've published - check out their blog post announcing FLUX.2 for more details.
This model was trained only on real-world data from the linked LAION dataset.
benjamin-paine changed discussion status to closed
I know what Auto Encoder is. I just wondered, cause I’ve tried to encode-decode one of my photos from a camera and pass it through AI detectors.
Real -> obviously, gets 1% AI on all
Reconstructed -> gets 80%+ AI on all
So, I wonder why