←新闻详情

面壁智能 OpenBMB更新于 10月21日 22:46

MiniCPM5-2B FIT-GGUF 按体积预算分四档

@TeksEdge Yes — MiniCPM5-2B FIT-GGUF is meant to hit a size budget, not a fixed Q-level. Those four tiers are fidelity-to-BF16 tradeoffs and drop into normal ll

AI 导读

@TeksEdge 是的——MiniCPM5-2B FIT-GGUF 旨在满足体积预算,而非固定的 Q-level。这四个档位是相对 BF16 的保真度权衡,可直接放入常规的 llama.cpp / LM Studio。

正文

当前显示原文,中文全文译文尚未就绪。

@TeksEdge Yes — MiniCPM5-2B FIT-GGUF is meant to hit a size budget, not a fixed Q-level. Those four tiers are fidelity-to-BF16 tradeoffs and drop into normal llama.cpp / LM Studio.

来源

面壁智能 OpenBMB