Speculative Decoding via Early-exiting for Faster LLM Inference with Thompson Sampling Control Mechanism

Re-assign community