Fetching the paper…

SWIFT: On-the-Fly Self-Speculative Decoding for LLM Inference Acceleration · Around