3 ms·

The [schnell] model variant is Apache-licensed and is open sourced on Hugging Face: https://huggingface.co/black-forest-labs/FLUX.1-schnell https://huggingface.

The [schnell] model variant is Apache-licensed and is open sourced on Hugging Face: https://huggingface.co/black-forest-labs/FLUX.1-schnell https://huggingface.co/black-forest-labs/FLUX.1-schnell It is very fast and very good at rendering text, and appears to have a text encoder such that the model can handle both text and positioning much better: https://x.com/minimaxir/status/1819041076872908894 https://x.com/minimaxir/status/1819041076872908894 A fun consequence of better text rendering is that it means text watermarks from its training data appear more clearly: https://x.com/minimaxir/status/1819045012166127921 https://x.com/minimaxir/status/1819045012166127921

dheera 2y ago
Thank you. Their website is super hard to navigate and I can't find a "DOWNLOAD" button.
- minimaxir 2y ago
  Note that actually running the model without a A100 GPU or better will be tricker than usual given its size (12B parameters, 24GB on disk). There is a PR to that repo for a diffusers implementation, which may run on a cheap L4 GPU w/ enable_model_cpu_offload(): https://huggingface.co/black-forest-labs/FLUX.1-schnell/commit/20cfea0b729f7f0ca79fab63201b16ef0b6e8df1 https://huggingface.co/black-forest-labs/FLUX.1-schnell/comm...
  - s-macke 2y ago
    Got it running. But it is a special setup. * NVIDIA Jetson AGX Orin Dev. Kit with 64 GB shared RAM. * Default configuration for flux-dev. (FP16, 50 steps) * 33GB GPU RAM usage. * 4 minutes 20 seconds per image at around 50 Watt power usage.
  - CuriouslyC 2y ago
    3090 TIs should be able to handle it without much in the way of tricks for a "reasonable" (for the HN crowd) price.
    - fl0id 2y ago
      higher ram apple silicon should be able to run it too. if they don't use some ancient pytorch version or something.
    - phkahler 2y ago
      Why not on a CPU with 32 or 64 GB of RAM?
      - CuriouslyC 2y ago
        Performance, mostly. It'll work but image generation is shitty to do slowly compared to text inference.
      - holoduke 2y ago
        Much slower memory and limited parallelism. Gpu ÷- 8k pr more cuda cores vs +-16 on regular cpu. Less mem swapping between operations. Gpu much much faster.
  - dheera 2y ago
    You don't need an A100, you can get a used 32GB V100 for $2K-$3K. It's probably the absolute best bang-for-buck inference GPU at the moment. Not for speed but just the fact that there are models you can actually fit on it that you can't fit on a gaming card, and as long as you can fit the model, it is still lightyears better than CPU inference.
    - Morphiak 2y ago
      Why this versus the 2 3090s (with nvlink for marginal gains) and 48GB for 2$K ?
nwoli 2y ago
That’s not really fair to conclude that the training data contains vanity fair images since the prompt includes “by Vanity Fair”. I could write “with text that says Shutterstock” in the prompt but that doesn’t necessairly mean the dataset contains that
- minimaxir 2y ago
  The logo has the same exact copyrighted typography as the real Vanity Fair logo. I've also reproduced the same-copyrighted-typography with other brands with identical composition as copyrighted images. Just asking it "Vanity Fair cover story about Shrek" at a 3:2 ratio gives it a composition identical to a Vanity Fair cover very consistently (subject is in front of logo typography partially obscuring it) The image linked has a traditional www watermark in the lower-left as well. Even something innocous as a "Super Mario 64" prompt shows a copyright watermark: https://x.com/minimaxir/status/1819093418246631855 https://x.com/minimaxir/status/1819093418246631855
  - fennecbutt 2y ago
    If the training data includes a public blog post which has a screenshot of a vanity fair piece? It's like GRRM complaining that LLMs can reproduce chunks of text from his books "they fed my novels into it" Oh yeah? It's definitely not all the parts of your book quoted in millions of places online, including several dedicated wiki style sites? That wouldn't be it, right?
  - Carrok 2y ago
    On my list of AI concerns, whether or not Vanity Fair has it’s copyright infringed does not appear.
    - minitoar 2y ago
      First they came for fashion magazines, and I said nothing.
      - jMyles 2y ago
RobotToaster 2y ago
How does the licence work when there's a bunch of restrictions at the bottom of that page that seem to contradict the licence?
- minimaxir 2y ago
  IANAL but I suspect that "Out-of-Scope Use" has no legal authority.

dheera 2y ago

Thank you. Their website is super hard to navigate and I can't find a "DOWNLOAD" button.

minimaxir 2y ago
Note that actually running the model without a A100 GPU or better will be tricker than usual given its size (12B parameters, 24GB on disk). There is a PR to that repo for a diffusers implementation, which may run on a cheap L4 GPU w/ enable_model_cpu_offload(): https://huggingface.co/black-forest-labs/FLUX.1-schnell/commit/20cfea0b729f7f0ca79fab63201b16ef0b6e8df1 https://huggingface.co/black-forest-labs/FLUX.1-schnell/comm...
- s-macke 2y ago
  Got it running. But it is a special setup. * NVIDIA Jetson AGX Orin Dev. Kit with 64 GB shared RAM. * Default configuration for flux-dev. (FP16, 50 steps) * 33GB GPU RAM usage. * 4 minutes 20 seconds per image at around 50 Watt power usage.
- CuriouslyC 2y ago
  3090 TIs should be able to handle it without much in the way of tricks for a "reasonable" (for the HN crowd) price.
  - fl0id 2y ago
    higher ram apple silicon should be able to run it too. if they don't use some ancient pytorch version or something.
  - phkahler 2y ago
    Why not on a CPU with 32 or 64 GB of RAM?
    - CuriouslyC 2y ago
      Performance, mostly. It'll work but image generation is shitty to do slowly compared to text inference.
    - holoduke 2y ago
      Much slower memory and limited parallelism. Gpu ÷- 8k pr more cuda cores vs +-16 on regular cpu. Less mem swapping between operations. Gpu much much faster.
- dheera 2y ago
  You don't need an A100, you can get a used 32GB V100 for $2K-$3K. It's probably the absolute best bang-for-buck inference GPU at the moment. Not for speed but just the fact that there are models you can actually fit on it that you can't fit on a gaming card, and as long as you can fit the model, it is still lightyears better than CPU inference.
  - Morphiak 2y ago
    Why this versus the 2 3090s (with nvlink for marginal gains) and 48GB for 2$K ?

nwoli 2y ago

That’s not really fair to conclude that the training data contains vanity fair images since the prompt includes “by Vanity Fair”. I could write “with text that says Shutterstock” in the prompt but that doesn’t necessairly mean the dataset contains that

minimaxir 2y ago
The logo has the same exact copyrighted typography as the real Vanity Fair logo. I've also reproduced the same-copyrighted-typography with other brands with identical composition as copyrighted images. Just asking it "Vanity Fair cover story about Shrek" at a 3:2 ratio gives it a composition identical to a Vanity Fair cover very consistently (subject is in front of logo typography partially obscuring it) The image linked has a traditional www watermark in the lower-left as well. Even something innocous as a "Super Mario 64" prompt shows a copyright watermark: https://x.com/minimaxir/status/1819093418246631855 https://x.com/minimaxir/status/1819093418246631855
- fennecbutt 2y ago
  If the training data includes a public blog post which has a screenshot of a vanity fair piece? It's like GRRM complaining that LLMs can reproduce chunks of text from his books "they fed my novels into it" Oh yeah? It's definitely not all the parts of your book quoted in millions of places online, including several dedicated wiki style sites? That wouldn't be it, right?
- Carrok 2y ago
  On my list of AI concerns, whether or not Vanity Fair has it’s copyright infringed does not appear.
  - minitoar 2y ago
    First they came for fashion magazines, and I said nothing.
    - jMyles 2y ago

RobotToaster 2y ago

How does the licence work when there's a bunch of restrictions at the bottom of that page that seem to contradict the licence?

minimaxir 2y ago
IANAL but I suspect that "Out-of-Scope Use" has no legal authority.