GitHub - huggingface/smollm: Everything about the SmolLM2 and SmolVLM family of models (original) (raw)

Smol Models 🤏

Welcome to Smol Models, a family of efficient and lightweight AI models from Hugging Face. Our mission is to create powerful yet compact models, for text and vision, that can run effectively on-device while maintaining strong performance.

News 📰

• Introducing FineMath, the best public math pretraining dataset 🚀
• Added continual pretraining code for Llama 3.2 3B on FineMath & FineWeb-Edu with nanotron

💬 SmolLM2 (Language Model)

SmolLM2 is our family of compact language models available in three sizes:

• SmolLM2-135M: Ultra-lightweight model for basic text tasks
• SmolLM2-360M: Balanced model for general use
• SmolLM2-1.7B: Our most capable language model, available at 🤏 SmolLM2-1.7B-Instruct here.

All models have instruction-tuned versions optimized for assistant-like interactions. Find them in our SmolLM2 collection.

👁️ SmolVLM (Vision Language Model)

SmolVLM is our compact multimodal model that can:

• Process both images and text and perform tasks like visual QA, image description, and visual storytelling
• Handle multiple images in a single conversation
• Run efficiently on-device

Repository Structure

smollm/
├── text/               # SmolLM2 related code and resources
├── vision/            # SmolVLM related code and resources
└── tools/             # Shared utilities and inference tools
    ├── smol_tools/    # Lightweight AI-powered tools
    ├── smollm_local_inference/
    └── smolvlm_local_inference/

Getting Started

SmolLM2

from transformers import AutoModelForCausalLM, AutoTokenizer

checkpoint = "HuggingFaceTB/SmolLM2-1.7B-Instruct" tokenizer = AutoTokenizer.from_pretrained(checkpoint) model = AutoModelForCausalLM.from_pretrained(checkpoint)

messages = [{"role": "user", "content": "Write a 100-word article on 'Benefits of Open-Source in AI research"}] input_text = tokenizer.apply_chat_template(messages, tokenize=False)

SmolVLM

from transformers import AutoProcessor, AutoModelForVision2Seq

processor = AutoProcessor.from_pretrained("HuggingFaceTB/SmolVLM-Instruct") model = AutoModelForVision2Seq.from_pretrained("HuggingFaceTB/SmolVLM-Instruct")

messages = [ { "role": "user", "content": [ {"type": "image"}, {"type": "text", "text": "What's in this image?"} ] } ]

Ecosystem

Resources

Documentation

• SmolLM2 Documentation
• SmolVLM Documentation
• Local Inference Guide

Pretrained Models

• SmolLM2 Models Collection
• SmolVLM Model

Datasets

• SmolTalk - Our instruction-tuning dataset
• FineMath - Mathematics pretraining dataset
• FineWeb-Edu - Educational content pretraining dataset