Overview

phamhai/Llama-3.2-1B-Instruct-Frog is a 1 billion parameter instruction-tuned model, built upon Meta's Llama-3.2-1B-Instruct. It features an extended context length of 131K tokens. The primary focus of this model is to provide enhanced support for Vietnamese language tasks, especially within Retrieval-Augmented Generation (RAG) frameworks, where the base Llama-3.2 models showed limitations.

Key Capabilities

• Vietnamese RAG Optimization: Specifically trained to improve performance in Vietnamese RAG scenarios.
• Efficient Deployment: Optimized for fast inference, making it suitable for deployment on on-premise and edge devices such as laptops, smartphones, NVIDIA Jetson Xavier, and Raspberry Pi.
• Instruction Following: Capable of handling various instruction-based tasks, including Question Answering, Summarization, Function Calling, and Question Rewriting, as demonstrated by provided examples.

Limitations

• The 1B parameter version may exhibit poorer prompt understanding and lower accuracy in certain tasks like summarization and entity extraction in Function Calling compared to larger models.

Good For

• Developers building Vietnamese-centric RAG applications.
• Projects requiring a lightweight LLM for edge device deployment.
• Use cases where fast inference and Vietnamese language proficiency are critical.