Fine-tuned models are now in limited public beta for GitHub Copilot Enterprise

GitHub Copilot is being enhanced with fine-tuned models designed to offer more personalized and contextually accurate coding assistance tailored to individual organizational needs. Available in limited public beta for Copilot Enterprise users, these models allow customization based on proprietary codebases and coding practices. By training on private repositories and incorporating user interaction telemetry, Copilot can adapt to unique coding environments, improving code quality and productivity. The fine-tuning process uses the LoRA method for efficient training, powered by Azure OpenAI Service for scalability and security, ensuring that data remains private and is not shared across models. Research indicates that Copilot users experience increased productivity and job satisfaction, with improvements in pull request metrics and build success rates. Fine-tuned models enable organizations in various sectors, such as finance and healthcare, to address specific coding requirements and enhance developer efficiency.