Combining Large and Small LLMs to Boost Inference Time and Quality | Towards Data Science

Implementing Speculative and Contrastive Decoding