Nitpick: this is not open weights, this is weights available. The license restricts many things like commercial, NSFW, etc.
Releasing weights for FLUX.1 Krea
71–80 of 112 posts
Re: Releasing weights for FLUX.1 Krea
#72I've never gotten one to make what I am thinking of:
A Galton board. At the top, several inches apart are two holes from which balls drop. One drops blue balls, the other red balls. They form a merged distribution below in columns, demonstrating dual overlapping normal distributions
Imagine one of these: https://imgur.com/a/DiAOTzJ but with two spouts at the top dropping different colored balls
Its attempts: https://imgur.com/undefined https://imgur.com/a/uecXDzI
Re: Releasing weights for FLUX.1 Krea
#73I've never gotten one to make what I am thinking of: A Galton board. At the top, several inches apart are two holes from which balls drop. One drops blue balls, the other red balls. They form a merged distribution below in columns, demonstrating dual overlapping normal distributions Imagine one of these: https://imgur.com/a/DiAOTzJ but with two spouts at the top dropping different colored balls Its attempts: https://…
Have you tried building one irl? I can't find a video of a double one
Re: Releasing weights for FLUX.1 Krea
#74Nitpick: this is not open weights, this is weights available. The license restricts many things like commercial, NSFW, etc.
I mean this started with Stable Diffusion 1.x->XL which were only loosely open, and has just gotten worse with progressively farther from open licensed image gen models being described as “open weights”, but, yes, Flux.1 Krea (like the weights-available versions of Flux.1 from BFL itself) is not open even to the degree of the older versions of Stable Diffusion; weights available and “free-as-in-beer licensed for certain uses”, sure, but not open.
Re: Releasing weights for FLUX.1 Krea
#75Can someone ELI5 why the safetensor file is 23.8 GB, given the 12B parameter model? Does the model use closer to 24 GB of VRAM or 12 GB of VRAM. I've always associated a 1 billion parameter = 1 GB of VRAM. Is this estimate inaccurate?
A parameter can be any size float. Lots of downloadable models are FP8 (8 bits per parameter), but it appears this model is FP16 (16 bits per parameter) Often, the training is done in FP16 then quantized down to FP8 or FP4 for distribution.
I think they are bfloat16, not FP16, but they are both 16bpw formats, so it doesn't make a size difference.
Re: Releasing weights for FLUX.1 Krea
#76Earlier quoted context omitted.
Is it possible (or do people already do this), to train a classifier to identify the AI look and use it as an adversary to try and maximise both 'quality' and 'not that sort of quality'?
I actually tried a few experiments in early exploration stages! I trained a small classifier to judge AI vs non-AI images. Use it as a reward model to do small RL / post training experiments. Sadly, was not too successful. We found that directly finetuning the model on high quality photorealistic image was most reliable. Another note about preference optimisation and RL is that it has really high quality ceiling but…
I could see how you might need a multi channel classifier so that one exists on a range (A) of -1 = "This looks like AI" to 1="This does not look like AI" and another(R) where 1="The above factor is relevant to this image" to 0="The AI-ness of this image is not a meaningful concept
Then optimise for max (Quality + A*R)
Arguably amplitude of A should do R but I think the AI-ness and the AI-ness-relevance are distinct concepts (It could be highly relevant but it can't tell what it should be).
Re: Releasing weights for FLUX.1 Krea
#77Earlier quoted context omitted.
How is the data collected?
The highest quality finetuning data was hand curated internally. I would say our post training pipeline is quite similar to SeedDream 2.0 ~ 3.0 series from ByteDance. Similar to them, we use extensive quality filters and internal models to get the highest quality possible. Even from there, we still hand curate a hand-picked subset.
[dead]
Re: Releasing weights for FLUX.1 Krea
#78hey hn! I'm one of the founders at Krea. we prepared a blogpost about how we trained FLUX Krea if you're interested in learning more: https://www.krea.ai/blog/flux-krea-open-source-release
Off topic but did you really hide scroll bars on the website? Why...? .scrollbar-hide { -ms-overflow-style: none; scrollbar-width: none; }
UI brought to you by vibe code
Re: Releasing weights for FLUX.1 Krea
#79Earlier quoted context omitted.
Off topic but did you really hide scroll bars on the website? Why...? .scrollbar-hide { -ms-overflow-style: none; scrollbar-width: none; }
UI brought to you by vibe code
nah, Krea is just from that side of design twitter where you don't uppercase letters and you can break the rules sometimes. very atypography-coded.
Re: Releasing weights for FLUX.1 Krea
#80Nice release. Ran some preliminary tests using the 12b Txt2Img Krea model. Its biggest wins seems to be raw speed (and possibly realism) but perhaps unsurprisingly did not score any higher on the leaderboard for prompt adherence than the normal Flux.1D model.
https://genai-showdown.specr.net
On another note, there seem to be some indication that Wan 2.2+ future models might end up becoming significant players in the T2I space though you'll probably need a metric ton of LoRAs to cover some of the lack of image diversity.