Hacker News
Flattening Every Memory Peak in Long-Context Mixture-of-Experts Training
13 points by gmays
ago
|
1 comments
ricardobeat
[-]
The abstract reads like slop. Authors working on LLMs should know better than that.