Home > AI Solutions > Artificial Intelligence > White Papers > Model Customization for Code Creation with Red Hat OpenShift AI on Dell AI Optimized Infrastructure > Overview
AI is becoming integrated into all aspects of our lives and having a real impact on almost all the ways we conduct business and provide services ranging from chat to code development. As businesses continue to define what their AI journey will look like, most are aware that they must start such initiatives to stay relevant.
One of the limitations of the base large language models is their generic behavior which needs customization and adaptation to a specific business use case using customization techniques. Also, LLMs, once trained and generated, do not have access to information beyond the date that they were trained.
There are numerous techniques that can be used to overcome these limitations, one of them is retrieval augmented generation (RAG). RAG extends the functionality of the LLMs by retrieving facts from an external knowledge base hosted using a vector database such as Redis.
Another technique that can be used to fine-tune the model is using a specific dataset. This approach modifies the parameters of the base model making them customized to a business use case such as code development, chatbot, digital assistant, or language translator and transcriber. Fine-tuning aims to maintain the original capabilities of a pretrained model while adapting it to suit more specialized use cases.
Before deciding which approach to use, one should consider the pros and cons of training your base model, fine-tuning an existing model and RAG:
This technical white paper provides an example of fine-tuning an LLM running on Dell Technologies AI optimized infrastructure demonstrating a robust developer-friendly ecosystem provided by Red Hat OpenShift AI. The use case consists of a scenario where administrators and operators can manage the life cycle of a large language model from the model customization using distributed fine-tuning to model serving and inferencing to add business value. An ecosystem solution is defined which is developer-friendly, providing a centralized hub where supported software components are available greatly increasing the downstream productivity of large language models.