ignaciolopezluna020/gguf repository overview

⁠SmolLM2

SmolLM2-360M is a compact language model with 360 million parameters, designed to run efficiently on-device while performing a wide range of language tasks. Trained on 4 trillion tokens from a diverse mix of datasets—including FineWeb-Edu, DCLM, The Stack, and newly curated filtered sources—it delivers strong performance in instruction following, knowledge, and reasoning. The instruct version was developed through supervised fine-tuning (SFT) on a blend of public and proprietary datasets, followed by Direct Preference Optimization (DPO) using UltraFeedback.

⁠Intended uses

SmolLM2 is designed for:

• Chat assistants
• Text-extraction
• Rewriting and summarization

⁠Characteristics

⁠Available model variants

¹: VRAM estimated based on model characteristics.

⁠Use this AI model with Docker Model Runner

First, pull the model:

docker model pull ai/smollm2

Then run the model:

docker model run ai/smollm2

For more information on Docker Model Runner, explore the documentation⁠.

⁠Benchmark performance

⁠Links

• SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model⁠