llm-adapter
PublicA plug-and-play adapter architecture that efficiently adapts large language models to downstream tasks. Essential for client-server architecture language model services. A practical approach to adapting LLMs' hidden activations without modifying the base model.