From ZDNet: 2024-04-04 17:45:00

OpenAI announces an expansion to its Custom Models Program and introduces six new features for its fine-tuning API to better cater to developers and enterprises. The program helps in customizing AI models for specific applications, reducing costs, and improving performance. Thousands of organizations have already trained hundreds of thousands of models using the API.

To address the need for extreme customization beyond fine-tuning, OpenAI introduced the Custom Model program in November 2023 for “select” organizations with large proprietary datasets. New features include assisted fine-tuning and custom-trained models. Assisted fine-tuning uses techniques like PEFT methods, while custom-trained models allow organizations to create purpose-built AI models from scratch.

Developers can now utilize features like epoch-based checkpoint creation, comparative UI in the playground, support for third-party platform integrations, validation metrics, hyperparameter configuration, and fine-tuning dashboard improvements with OpenAI’s self-serve fine-tuning API. OpenAI aims to help organizations maximize model performance and create models with a deeper understanding of their business or domain.



Read more at ZDNet: OpenAI makes it easier for developers to fine-tune and build AI models