4 results for tag "speculative-decoding"
A large collection of Claude Code skill templates sponsored by Z.AI, providing ready-to-use development skill configurations across various domains.
A speculative decoding skill from the AI Research Engineering Skills Library, providing techniques for accelerating large language model inference using speculative decoding methods.
Accelerates LLM inference by 1.5-3.6Γ using speculative decoding, draft models, and parallel token generation techniques.