PhoneLM:an Efficient and Capable Small Language Model Family through Principled Pre-training
This work addresses the need for efficient SLMs tailored to device hardware, which is crucial for on-device AI applications, representing an incremental improvement in SLM design.
The paper tackles the problem of designing small language models (SLMs) for on-device deployment by proposing a principle of architecture searching for optimal runtime efficiency before pre-training, resulting in the PhoneLM family (0.5B and 1.5B versions) that achieves state-of-the-art capability-efficiency tradeoff for similar parameter sizes.
The interest in developing small language models (SLM) for on-device deployment is fast growing. However, the existing SLM design hardly considers the device hardware characteristics. Instead, this work presents a simple yet effective principle for SLM design: architecture searching for (near-)optimal runtime efficiency before pre-training. Guided by this principle, we develop PhoneLM SLM family (currently with 0.5B and 1.5B versions), that acheive the state-of-the-art capability-efficiency tradeoff among those with similar parameter size. We fully open-source the code, weights, and training datasets of PhoneLM for reproducibility and transparency, including both base and instructed versions. We also release a finetuned version of PhoneLM capable of accurate Android Intent invocation, and an end-to-end Android demo. All materials are available at https://github.com/UbiquitousLearning/PhoneLM.