SDSAT: Accelerating LLM Inference through Speculative Decoding with Semantic Adaptive Tokens

Re-assign community