Releasing weights for FLUX.1 Krea • Hacker News Board

Re: Releasing weights for FLUX.1 Krea

#31

Post by qlm » Thu, Jul 31, 2025, 9:23 PM UTC

Images still look AI generated

Re: Releasing weights for FLUX.1 Krea

#32

Post by sangwulee » Thu, Jul 31, 2025, 9:26 PM UTC

swyx wrote: post #25

sangwulee wrote: post #4

Hi! I'm lead researcher on Krea-1. FLUX.1 Krea is a 12B rectified flow model distilled from Krea-1, designed to be compatible with FLUX architecture. Happy to answer any technical questions :)

thanks for doing this! what does " designed to be compatible with FLUX architecture" mean and why is that important?

FLUX.1 is one of the most popular open weights text-to-image models. We distilled Krea-1 to FLUX.1 [dev] model so that the community can adopt it seamlessly into existing ecosystem. Any finetuning code, workflows, etc that was built on top of FLUX.1 [dev] can be reused with our model :)

Re: Releasing weights for FLUX.1 Krea

#33

Post by artninja1988 » Thu, Jul 31, 2025, 9:30 PM UTC

Thanks for the release to the team! Any plans on doing the same for FLUX.1 Kontext?

Re: Releasing weights for FLUX.1 Krea

#34

Post by leftstrokeviral » Thu, Jul 31, 2025, 9:31 PM UTC

How much data is the model trained on?

Re: Releasing weights for FLUX.1 Krea

#35

Post by paparicio » Thu, Jul 31, 2025, 9:32 PM UTC

Great, these people are crazy. CONGRATS!!!

Re: Releasing weights for FLUX.1 Krea

#36

Post by sangwulee » Thu, Jul 31, 2025, 9:33 PM UTC

bluehark wrote: post #29

Do you have an NVIDIA optimized version? Similar to how RTX accelerated FLUX.1 Kontext: https://blogs.nvidia.com/blog/rtx-ai-garage-flux-kontext-nim...

We have not added a separate RTX accelerated version for FLUX.1 Krea, but the model is fully compatible with existing FLUX.1 dev codebase. I don't think we made a separate onnx export for it though. Doing 4~8 bit quantized version with SVDQuant would be a nice follow up so that the checkpoint is more friendly for consumer grade hardware.

Re: Releasing weights for FLUX.1 Krea

#37

Post by dvrp » Thu, Jul 31, 2025, 9:38 PM UTC

paparicio wrote: post #35

Great, these people are crazy. CONGRATS!!!

Haha thanks! Do you happen to have a use-case for it?

Re: Releasing weights for FLUX.1 Krea

#38

Post by sangwulee » Thu, Jul 31, 2025, 9:39 PM UTC

bluehark wrote: post #30

How large was the dataset used for post-training?

We used two types of datasets for post-training. Supervised finetuning data and preference data used for RLHF stage. You can actually use less than < 1M samples to significantly boost the aesthetics. Quality matters A LOT. Quantity helps with generalisation and stability of the checkpoints though.

Re: Releasing weights for FLUX.1 Krea

#39

Post by dvrp » Thu, Jul 31, 2025, 9:39 PM UTC

qlm wrote: post #31

Images still look AI generated

Yeah, there are still imperfections. But, it’s surprising to us how much the quality can be improved without the need of a whole pre-training (re-) run.

Re: Releasing weights for FLUX.1 Krea

#40

Post by dvrp » Thu, Jul 31, 2025, 9:41 PM UTC

leftstrokeviral wrote: post #34

How much data is the model trained on?

Copying and pasting Sangwu’s answer:

We used two types of datasets for post-training. Supervised finetuning data and preference data used for RLHF stage. You can actually use less than < 1M samples to significantly boost the aesthetics. Quality matters A LOT. Quantity helps with generalisation and stability of the checkpoints though.