How AMD Quark trains, quantizes, and serves EAGLE3 speculative-decoding drafts with vLLM on AMD Instinct GPUs, delivering up to 2.00x throughput gains for Kimi-K2.5 and 1.79x for MiniMax-M2.5.
EAGLE3 Speculative Decoding on AMD Instinct GPUs: Training and Serving with vLLM and AMD Quark
calendar_today
July 13, 2026
person
Larry Li, Chao Li, Haichen Zhang, Chun Fang, Andy Luo, Spandan Tiwari, and Ashish Sirasao
domain
vllm