Get models like Phi-2, Mistral, and LLaVA running locally on a Raspberry Pi with Ollama
Inside Phi 2: Microsoft's Small Language Model explores the development of tailored AI models that minimize resource usage, with a focus on small language models (SLMs). The article discusses Microsoft Research's approach to building generative AI models, highlighting their "textbooks are all you need" strategy used in training their Phi series SLMs. Key aspects include curating high-quality training data, generating synthetic content, and fine-tuning the model with domain-specific information. This innovative approach has resulted in surprisingly good outcomes, with some benchmarked SLMs performing similarly or even better than larger LLMs like GPT.
Bookmarks on this page are managed by an admin user. 