--- title: README emoji: 🐠 colorFrom: yellow colorTo: yellow sdk: static pinned: false --- # Hugging Face Research The science team at Hugging Face is dedicated to advancing machine learning research in ways that maximize value for the whole community. Our work focuses on three core areas of tooling, datasets and open models. This is the release timeline so far (you can click on each element!):
🐳Open-R1
πŸ› οΈSmolAgents
Jan 2025
Dec 2024
🀏SmolLM2
πŸ€“SmolVLM
Nov
🐢Idefics 3
Aug
πŸ₯‡Win AIMO
🐢Docmatix
🀏SmolLM
Jul
🍷FineWeb-Edu
πŸ‘©β€πŸ«Stanford CS25
Jun
πŸ“ˆWSD Analysis
May
🍷FineWeb
πŸ•΅οΈJAT Agent
🐢Idefics 2
Apr
πŸͺZephyr Gemma
πŸͺCosmopedia
Mar
βš™οΈNanotron Release
βš™οΈDatatrove Release
⭐️The Stack v2
⭐️StarCoder2
Feb
πŸ”₯ Warming up
Jan 2024
### πŸ› οΈ Tooling & Infrastructure The foundation of ML research is tooling and infrastructure and we are working on a range of tools such as [datatrove](www.github.com/huggingface/datatrove), [nanotron](www.github.com/huggingface/nanotron), [TRL](www.github.com/huggingface/trl), [LeRobot](www.github.com/huggingface/lerobot), and [lighteval](www.github.com/huggingface/lighteval). ### πŸ“‘ Datasets High quality datasets are the powerhouse of LLMs and require special care and skills to build. We focus on building high-quality datasets such as [no-robots](https://huggingface.co/datasets/HuggingFaceH4/no_robots), [FineWeb](https://huggingface.co/datasets/HuggingFaceFW/fineweb), [The Stack](https://huggingface.co/datasets/bigcode/the-stack-v2), and [FineVideo](https://huggingface.co/datasets/HuggingFaceFV/finevideo). ### πŸ€– Open Models The datatsets and training recipes of most state-of-the-art models are not released. We build cutting-edge models and release the full training pipeline as well fostering more innovation and reproducibility, such as [Zephyr](https://huggingface.co/HuggingFaceH4/zephyr-7b-beta), [StarCoder2](https://huggingface.co/bigcode/starcoder2-15b), or [SmolLM2](https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct). ### 🌸 Collaborations Research and collaboration go hand in hand. That's why we like to organize and participate in large open collaborations such as [BigScience](https://bigscience.huggingface.co) and [BigCode](https://www.bigcode-project.org), as well as lots of smaller partnerships such as [Leaderboards on the Hub](https://huggingface.co/blog?tag=leaderboard). ### βš™οΈ Infrastructre The research team is organized in small teams with typically <4 people and the science cluster consists of 96 x 8xH100 nodes as well as an auto-scalable CPU cluster for dataset processing. In this setup, even a small research team can build and push out impactful artifacts. ### πŸ“– Educational material Besides writing tech reports of research projects we also like to write more educational content to help newcomers get started to the field or practitioners. We built for example the [alignment handbook](https://github.com/huggingface/alignment-handbook), the [evaluation guidebook](https://github.com/huggingface/evaluation-guidebook), the [pretraining tutorial](https://www.youtube.com/watch?v=2-SPH9hIKT8), or the [FineWeb blog](https://huggingface.co/spaces/HuggingFaceFW/blogpost-fineweb-v1). ### πŸ€— Join us! We are actively hiring for both full-time and internships. Check out [hf.co/jobs](https://hf.co/jobs)