I know some of you were waiting fo it, no wait no more ! SVD Keyframe Interpolation @gradio demo is available Hugging Face ! Space link : huggingface.co/spaces/fffilon…

New model alert! 🔥LLaVa-OneVision is now in the Transformers library A powerful series (0.5B/7B/72/B) for single, multi-image, and video scenarios. Successor of LLaVa-NeXT. SOTA open model on Video-MME: video-mme.github.io/home_page.html… Definitely worth a look besides Qwen2-VL 1/2

Yes. So many people seem to think writing a pinokio launcher is some super technical thing only I can do, but the whole point of pinokio is that it makes writing launchers super easy. If you can follow a project readme to install and run some project, you can write a launcher

The weights of our 70B model are available today on Hugging Face here: huggingface.co/mattshumer/Ref… Hyperbolic API available later today. Next week, we will release the weights of Reflection-405B, along with a short report going into more detail on our process and findings.

Gemini Flash is now tied with gpt-4o for #2 on the lmsys *vision* leaderboard! combine that with a 1M context length and you can do some seriously cool multimodal work for super cheap ⚡️⚡️⚡️

🔥🔥 R users can now easily use Gradio 🔥🔥

It is so much fun to play with Gradio. If you know JS/CSS, there is no limit of what you can do with it. Here, I am trying to create an application for writing tool with LLM-powered analysis. At then end of dev, it will be shared on Hugging Face Space.

It’s only 9 AM and the community has already submitted 8 new papers on hf.co/papers Hugging Face 🚀🔥 Time to dive into some cool AI research 💡

if the main demo is busy looks like you can try the new 70B model here on Hugging Face : huggingface.co/spaces/feather…

🙋🏻‍♂️hey there folks, just released a coding model under 10B parameters with 125K context window , achieving (very high!) SOTA scores on evals. you can try it out for yourself on Hugging Face 👇🏻📷 models : huggingface.co/collections/01… Gradio demo : huggingface.co/spaces/Tonic/Y…

Reflection-Llama-3.1-70B by Matt Shumer Sahil Chaudhary Glaive AI is #1 trending on HF. I've said it and will say it again: you don't need to be a big tech to fine-tune, optimize and run your own models for your specific constraints and you will benefit massively from it.

Fluxgym: Train FLUX LoRA LOCALLY with LOW VRAM FLUX LoRAs are super exciting, but there has not been an easy way to train them LOCALLY using LOW VRAM machines. Fluxgym is a Gradio app for training your own Flux LoRAs on your own 12G, 16G, 20G+ VRAM computer for free.

Here's the Fluxgym GitHub repository if you want to fork or install manually: github.com/cocktailpeanut…

Video-to-video is now available in the official CogVideoX-5B Space 🔥 Try it out 🎥 ➡️🎥 huggingface.co/spaces/THUDM/C…

Great news--looks like I was wasting 4GB VRAM with the previous code, it wasn't clearing up memory after running Florence. Just pushed an update that saves 4GB. This means, theoretically you can run Fluxgym on: - 16G VRAM - 12G VRAM - 8GB VRAM (!!!) But will need to verify.

I've been working behind the scenes to make this runnable on TPU with Hugging Face, but I'm sorry, I was overly ambitious. We'll have to wait for the optimization steps to make it work with reasonable resources. 🤗

At the moment, Fluxgym provides Flux-dev based LoRA training only. Because you can actually take the Flux-dev LoRA and use it with Flux-schnell for 4 step inference. But maybe I'm missing something. Anyone know if there are any benefits to training on top of Flux-schnell?

updates: 1. completely forgot about GitHub pages, so now there's a static site 2. upgraded from Google AI Studio gemini flash to gemini pro 3. cleaned up the repo #buildinpublic