We are excited to see vLLM as an option for local apps in the Hugging Face hub! It comes with easy snippets to quickly test out the model.

It's now so easy add images to the gallery of your LoRA on Hugging Face 🤯 🪄 ① Generate an image with the Widget 🖼️ ② Press "Add to model card gallery" 🔥

We are announcing Llama-3.1-SuperNova, a Llama-3.1-70B-Instruct model offline distilled from Llama-3.1-405B-Instruct. It's ridiculously strong, particularly in instruction following and math. It's available to play with at supernova.arcee.ai. Read more about the model and

Here we go again 🤗 Here is a step-by-step guide on how to upload 12B+ models to Hugging Face Step 1. pip install huggingface_hub Step 2. huggingface-cli upload-folder <repo-id> <local-path> --repo-type=model That's it. Enjoy fast download speeds!

Mistral released Pixtral 12B Vision Language Model 🔥 Some notes on the release: 1. Text backbone: Mistral Nemo 12B 2. Vision Adapter: 400M 3. Uses GeLU (for vision adapter) & 2D RoPE (for vision encoder) 4. Larger vocabulary - 131,072 5. Three new special tokens - `img`,

LLaMA-Omni, a new model for speech interaction 🦙Based on Llama 3.1 8B Instruct ⚡️Low-latency speech 🚀Simultaneous text and speech generation 🤏Trained with 4 GPUs in less than 3 days Model: hf.co/ICTNLP/Llama-3… Paper: hf.co/papers/2409.06…

Are you tired of having to download datasets to explore them? You can now run SQL directly in the Hugging Face Hub dataset viewer 🤗As an example, check out positive samples in the IMDB dataset huggingface.co/datasets/stanf…

Google releases DataGemma 👀Challenge hallucinations by using real data from Data Commons 🧠Gemma 2 fine-tuned for RAG and Retrieval Interleaved Generation (RIG) Models: huggingface.co/collections/go… Blog: blog.google/technology/ai/…

Can LLMs Generate Novel Research Ideas? 🤔 🏫Recruit 100 NLP researchers to write ideas 🫣Blind reviews of both LLMs and human ideas Result: LLM ideas are rated more novel but less feasible Nice in-depth write-up in their paper! huggingface.co/papers/2409.04…

This is how the Hugging Face team is preparing for the PyTorch Conference next week🤗 See you soon and come to our party for some nice swag!

Announcing SAM 2 Studio and Core ML Segment Anything 2! I'm super excited about on-device ML, and firmly believe that it will be a big part of the future of AI. We converted Segment Anything 2 to Core ML and wrote a demo app that you can use on your Mac 🔥

Introducing SAM 2 Studio ⚡️Fully local image segmentation app 👀Privacy-first: nothing sent to a server ⭐️Using Segment Anything 2 + CoreML Models: huggingface.co/collections/ap… Download app: huggingface.co/coreml-project… GH repo: github.com/huggingface/sa…

A quick update that we have release the inference code and model for DepthCrafter. Code: github.com/Tencent/DepthC… Model: huggingface.co/tencent/DepthC… Project page: depthcrafter.github.io

Personal update: I'h joining Hugging Face today!

🧶 Introducing DataCraft: build synthetic datasets using natural language! Creating good quality synthetic data is difficult. It’s a trial and error process and requires a lot of tricks. DataCraft puts Argilla's dataset generation best practices in your hands within a no

I'm thrilled to unveil our revamped Inference API docs! We've tackled your feedback head-on: clearer rate limits, dedicated PRO section, better code examples, and detailed parameter lists for each task. Deploying AI made simple. Dive in: huggingface.co/docs/api-infer…

Segment Anything 2 (SAM 2) by AI at Meta running 100% on-device powered by Apple CoreML! ⚡ Takes fraction of a second to run inference on Mac or iPhone! > Apache licensed optimised model checkpoints - tiny, small, base ad large! > Open source application to annote any image in

How much memory do you need to use a model? 🤗 Powered by EleutherAI Cookbook, you can now try this demo to - Estimate memory to infer or train a LLM - Compute the parameters count - Theoretical FLOPs for training huggingface.co/spaces/derek-t…