Hugging Face

🌍 Explore how geoengineering could cool our planet! 🚀 With a Gradio Dashboard that subs as a website😍! 👉 This latest Gradio demo dives into the effects of stratospheric aerosol injection (SAI), a cutting-edge proposal to fight global warming. 🌡️ 🤗 SAI involves dispersing reflective particles high in the atmosphere to reflect sunlight, potentially cooling the Earth's surface. While this method could mitigate some immediate climate change impacts, it's not a replacement for essential reductions in greenhouse gas emissions. 💪 👩🔬 Powered by the FaIR climate emulator from the University of Leeds and Oxford University, this demo allows you to simulate Earth’s temperature response to various CO2 and SAI scenarios. Choose from standard emissions scenarios, called Shared Socioeconomic Pathways (SSPs), and adjust SAI levels based on historical volcanic data like the 1991 Mount Pinatubo eruption. 📊 The Gradio demo provides interactive plots to visualize global and regional temperature projections up to the end of the century, with and without SAI, including uncertainty estimates. 💯 Discover the potential impacts of different geoengineering scenarios on our climate. Check out this impressive dashboard built with Gradio, and see firsthand how our tools can help visualize complex climate solutions. Dive into the demo here: [Link to demo] #ClimateChange #Geoengineering #AI #DataVisualization #ClimateAction

🚀 Stability AI's Stable Diffusion 3 (SD3) - ControlNet (Canny) is already here! 🤩 Explore for free on Hugging Face Spaces and create your next masterpiece✨ Demo from Sylvain F.: https://lnkd.in/gmzN9nHw SD3 ControlNet weights: https://lnkd.in/gJ3cpQzv 🚀 Launch your next AI App with Gradio's Image Component! Visualize results within seconds. Dive into image-based GenAI with Gradio and build the UI of your dream!📸✨ Visit: Gradio.dev

Slides from my talk at the Center of AI, UCL last week: "Controlling Diffusion Models: Assorted Approaches" https://lnkd.in/g4vq7eQC

A new batch of Hugging Face dog bandanas is available 🔥 Connect with / send a DM to Adina Yakefu if you want one and she will send one. The only condition is to post a picture of your puppies wearing it 🐶

everything you need to know about multimodal AI, most outstanding models, vision language models, tools and more in 12 mins!

releasing: smol vision Hugging Face 🌼 A repository with notebooks on shrinking, optimizing, speeding-up, customizing large vision models!

Fuck yeah! Nemotron 4 340B is out! 🔥 > Chonky beast beats Mixtral 8x22B, Claude sonnet, Llama3 70B, Qwen 2 and competes with GPT 4 > Release Base, Instruct and Reward model > Trained on 9T tokens > 8T pre-training + 1T for continual training for increased quality > Instruct model trained on 98% Synthetic data! LFG! > Trained on English, 50+ natural languages and 40+ coding languages > June 2023 training cut-off > Apache 2.0 like license - explicitly allows synthetic data generation and commercial use ⚡ > GPT styled - used Group Query Attention (GQA) & RoPE embeddings > Uses 10K HelpSteer2 dataset (human annotated) to train the Reward model - also released! > Run inference on 8x H100 Bonus: They release a 340B reward model, too - it ranks #1 on the RewardBench 🫡 Congratulations to Nvidia for open sourcing-; it is such a brilliant model!

Meta announces An Image is Worth More Than 16x16 Patches Exploring Transformers on Individual Pixels paper page: https://buff.ly/3VHOUTB This work does not introduce a new method. Instead, we present an interesting finding that questions the necessity of the inductive bias -- locality in modern computer vision architectures. Concretely, we find that vanilla Transformers can operate by directly treating each individual pixel as a token and achieve highly performant results. This is substantially different from the popular design in Vision Transformer, which maintains the inductive bias from ConvNets towards local neighborhoods (e.g. by treating each 16x16 patch as a token). We mainly showcase the effectiveness of pixels-as-tokens across three well-studied tasks in computer vision: supervised learning for object classification, self-supervised learning via masked autoencoding, and image generation with diffusion models. Although directly operating on individual pixels is less computationally practical, we believe the community must be aware of this surprising piece of knowledge when devising the next generation of neural architectures for computer vision.

🔥 Depth Anything is back with version 2 - and is 10x faster than other current methods! 💪 Models of various sizes (from 25 million to 1.3 billion parameters) are available on Hugging Face Hub✨ Demo link & more😉👇 🚀 Depth Anything V2: Unveiling the secrets of powerful depth estimation!📏 - Trains only on Synthetic images🎨 - 10x faster & more accurate than other SD-based models⚡ - Introduces new benchmark with precise annotations & diverse scenes 📊 Models on 🤗: https://lnkd.in/gT5Khy9G 🚀 Experience the magic of Depth Anything V2 with Gradio demo: https://lnkd.in/gzh2VeAi 👉Upload your own images and explore the predicted depth maps 👉 Test the models on diverse scenes and share your findings

It's been extremely rewarding to support Navigate360 in their mission to keep school communities safe online through our Expert Support Program at Hugging Face 🤗 From careful data annotation to fine-tuning open models that detect harmful content in text and images to prompt engineering LLMs for threat understanding, this team is innovating with open-source and open-science to prevent self-harm, stop bullying, and detect emerging threats and violence against others. I look forward to continuing our work together to keep schools and their communities safe. 👏 Kudos to this awesome team Jeff Jarrard John DeJesus Dustin Eagar Vivienne May Uy