Randall DeFauwThe Data “Center of Excellence”In my previous posts I’ve talked a lot about technical details of Generative AI. But I’ve been thinking recently about how generative AI…Oct 1, 2023Oct 1, 2023
Randall DeFauwFine-tuning Llama 2 on Trainium instancesLlama 2 is a powerful and popular large-language model (LLM) published by Meta. There’s a lot of interest in fine-tuning Llama 2 with…Sep 19, 2023Sep 19, 2023
Randall DeFauwUsing Python imports to inject more relevant context into code generation promptsA recent research paper proposed a method for injecting more relevant context into code generation prompts for LLMs. In a nutshell, rather…Sep 13, 2023Sep 13, 2023
Randall DeFauwRunning the HELM benchmarks on Amazon EC2The HELM testing suite is probably the most well-known framework for comparing LLM performance. Several folks have asked about how to use…Jul 25, 2023Jul 25, 2023
Randall DeFauwEvaluating Large Language ModelsOne of the interesting things about building generative AI solutions is the impressive number of models you can pick from. If you want to…Jul 18, 2023Jul 18, 2023
Randall DeFauwA shared responsibility model for generative AIOne of the more useful things that AWS published is the shared responsibility model for cloud computing. This model succinctly describes…Jul 2, 2023Jul 2, 2023