News list for "下文"

DeepSeek推出NSA,用于超快速的长上下文训练和推理

2月18日讯,DeepSeek推出NSA。DeepSeek称,NSA是一种与硬件一致且本机可训练的稀疏注意力机制,用于超快速的长上下文训练和推理。通过针对现代硬件的优化设计,NSA加快了推理速度,同时降低了预训练成本,而不会影响性能。在一般基准测试、长上下文任务和基于指令的推理上,它的表现与完全注意力模型相当甚至更好。

clock
2025-02-18 08:34:43
Disclaimer:
1. The information provided does not constitute investment advice. Investors should make independent decisions and bear all risks themselves.
2. The copyright of this content belongs to the original author. The views expressed herein are solely those of the author and do not represent the stance or position of this website.