Skip to content
Better HN
Top
Best
Ask
Show
New
Jobs
Search
⌘K
LISA: Layerwise Importance Sampling for Memory-Efficient LLM Fine-Tuning
(opens in new tab)
(arxiv.org)
3 points
convexstrictly
2y ago
1 comments
Save
Share
1 comments
1 comments · 1 top-level
top
newest
oldest
convexstrictly
OP
2y ago
An author claims better performance than LoRA in 50% of the time.
https://twitter.com/Rui45898440/status/1772996453557997924
j
/
k
navigate · click thread line to collapse
