LISA: Layerwise Importance Sampling for Memory-Efficient LLM Fine-Tuning arxiv.org 3 points by convexstrictly 2 years ago · 1 comment Reader PiP Save Collapse all Expand all 1 thread convexstrictlyOP 2 years ago An author claims better performance than LoRA in 50% of the time.https://twitter.com/Rui45898440/status/1772996453557997924