Smol L M2 1.7 B Instruct q4f16_1 M L C W E B G P U

Introduction

SmolLM2 is a family of compact language models available in three parameter sizes: 135M, 360M, and 1.7B. This repository hosts the WASM-compiled 1.7B model version, optimized for WebLLM.

Architecture

The SmolLM2-1.7B model demonstrates substantial improvements over its predecessor, SmolLM1-1.7B, in areas such as instruction following, knowledge acquisition, reasoning, and mathematics.

Training

The model was trained on 11 trillion tokens using a diverse set of datasets, including FineWeb-Edu, DCLM, The Stack, and new mathematics and coding datasets. It underwent supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) with UltraFeedback.

Guide: Running Locally

1. Clone Repository: Clone the SmolLM2-1.7B repository to your local machine.
2. Install Dependencies: Ensure required dependencies for WASM and WebLLM are installed.
3. Run Model: Execute the model using an appropriate runtime environment for WebLLM.
4. Cloud GPUs: Consider using cloud GPU services like AWS, GCP, or Azure for more efficient performance.

License

The SmolLM2-1.7B model is licensed under the Apache 2.0 License.