ML-SpecQD: Multi-Level Speculative Decoding with Quantized Drafts

MQ
Re-assign community