vLLM投机解码参数调优:Kimi-K2.5-Eagle3-FP8的num_speculative_tokens配置详解
vLLM投机解码参数调优:Kimi-K2.5-Eagle3-FP8的num_speculative_tokens配置详解 【免费下载链接】Kimi-K2.5-Eagle3-FP8 项目地址: https://ai.gitcode.com/hf_mirrors/amd/Kimi-K2.5-Eagle3-FP8
vLLM投机解码(Speculative Decoding)是…
2026/8/17 22:52:36