Text-to-Image
Transformers
English
SupraDiT
feature-extraction
small
supra
image
flux
img
t2i
from scratch
custom_code
Instructions to use SupraLabs/Supra2-IMG with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use SupraLabs/Supra2-IMG with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("SupraLabs/Supra2-IMG", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Question regarding DiT usage
#3
by Incorporo-user - opened
Hi,
Is there any specific reason why you decided to build this project with DiT instead of U-Nets? (eg: FCDM)
I am currently working on a flow matching model for glyph creation and I am looking for small architectures since fonts should be an easier task so having a smaller model <1B would make it better in inference stage.
Hey there! We just chose DiT because it's more modern than U-Nets actually.
But I think U-Nets are good too.
Please try both architectures and tell us which was better - we'd actually like to know that π€
Best,
LH