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

LRM
Re-assign community