Login
From:
科学空间|Scientific Spaces
(Uncensored)
subscribe
低精度Attention可能存在有偏的舍入误差
https://kexue.fm/archives/11371
links
backlinks
前段时间笔者在arXiv上刷到了论文《Why Low-Precision Transformer Training Fails: An Analysis on Flash Attention》，...
Roast topics
Find topics
Roast it!
Roast topics
Find topics
Find it!
Roast topics
Find topics
Find it!
