CodeLlama-13b-Instruct-hf Overview

This model is the 13 billion parameter instruction-tuned variant within the Code Llama family, developed by Meta. It is an auto-regressive language model utilizing an optimized transformer architecture, trained between January and July 2023. The Code Llama series includes base models for general code tasks, Python-specific models, and instruct-tuned versions like this one, which are designed for instruction following and safer deployment in applications.

Key Capabilities

• Code Completion: Generates missing parts of code.
• Infilling: Fills in code within existing structures.
• Instructions/Chat: Responds to natural language instructions for code-related tasks.

Intended Use Cases

CodeLlama-13b-Instruct-hf is suitable for commercial and research applications requiring code synthesis and understanding in English and programming languages. It is particularly intended for use in code assistant and generation tools where instruction following and safety are priorities. Meta developed this model, and its use is governed by a custom commercial license. More details are available in the research paper and its arXiv page.