huggingface的hf-mem工具更新,新增将混合专家模型(MoE)内存估计分解为基础权重、路由专家和KV缓存三个部分的功能。
RT @alvarobartt: Latest hf-mem now breaks down Mixture-of-Experts (MoE) memory estimations into base weights, routed experts, and KV cach…
likes: 33 | retweets: 6 | replies: 1 | views: 6074