HuggingFaceTB

This is the home for smol models (SmolLM) and high quality pre-training datasets. We released:

• FineWeb-Edu: a filtered version of FineWeb dataset for educational content, paper available here.
• Cosmopedia: the largest open synthetic dataset, with 25B tokens and 30M samples. It contains synthetic textbooks, blog posts, and stories, posts generated by Mixtral. Blog post available here.
• Smollm-Corpus: the pre-training corpus of SmolLM: Cosmopedia v0.2, FineWeb-Edu dedup and Python-Edu. Blog post available here.
• SmolLM models and SmolLM2 models: a series of strong small models in three sizes: 135M, 360M and 1.7B
• SmolVLM: a 2 billion Vision Language Model (VLM) built for on-device inference. It uses SmolLM2-1.7B as a language backbone. Blog post available here.

News 🗞️

• SmolLM2: you can find our most capable model SmolLM2-1.7B here: https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct and our training and evaluation toolkit at: https://github.com/huggingface/smollm
• We released our SFT mix SmolTalk, a 1M samples synthetic dataset to improve instruction following, chat and reasoning: https://hf.co/datasets/HuggingFaceTB/smoltalk
• SmolVLM: a lightweight 2B Vision Language Model available here https://huggingface.co/HuggingFaceTB/SmolVLM-Instruct