LISA: Layerwise Importance Sampling for Memory-Efficient LLM Fine-Tuning (arxiv.org) 3 points by convexstrictly 2y ago ↗ HN
[–] convexstrictly 2y ago ↗ An author claims better performance than LoRA in 50% of the time.https://twitter.com/Rui45898440/status/1772996453557997924
1 comment
[ 0.28 ms ] story [ 11.0 ms ] threadhttps://twitter.com/Rui45898440/status/1772996453557997924