MobileLLM is a lightweight large language model (LLM) framework developed by Facebook Research, optimized for on-device deployment where computational and memory efficiency are critical. Introduced in the ICML 2024 paper “MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases”, it focuses on delivering strong reasoning and generalization capabilities in models under one billion parameters. The framework integrates several architectural innovations—SwiGLU activation, deep and thin network design, embedding sharing, and grouped-query attention (GQA)—to achieve a superior trade-off between model size, inference speed, and accuracy. MobileLLM demonstrates remarkable performance, with the 125M and 350M variants outperforming previous state-of-the-art models of the same scale by up to 4.3% on zero-shot commonsense reasoning tasks.
Features
- Optimized transformer architecture for sub-billion parameter LLMs
- Combines SwiGLU activation, embedding sharing, and grouped-query attention
- Supports distributed multi-node pretraining with PyTorch ≥ 2.0
- Delivers state-of-the-art zero-shot reasoning results across multiple tasks
- Includes reproducible training and evaluation pipelines for multiple model sizes
- Scalable design philosophy extending from 125M to 1.5B parameters
Project Samples
