News for hackers. Stuff that matters.

A comment on LISA: Layerwise Importance Sampling for Memory-Efficient LLM Fine-Tuning.

Re:LISA: Layerwise Importance Sampling for Memory-Efficient ... by convexstrictly on Wednesday March 27, 2024 @01:19PM (#39841943)
An author claims better performance than LoRA in 50% of the time.

https://twitter.com/Rui45898440/status/1772996453557997924

[ Parent | on HN ]

There are no comments on this comment yet.