Scholar Feed

LLM quantization

QLoRA

QLoRA: Efficient Finetuning of Quantized LLMs

Superseded baseline#20 of 80 most-superseded · first seen May 23, 2023

Superseded — cited as a baseline and beaten by newer methods

3 papers critique it · 1 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites QLoRA as a baseline.

Limited to fine-tuning (not training from scratch) and requires GPU hardware.
— True 4-Bit Quantized Convolutional Neural Network Training on CPU: Achieving Full-Precision Parity
“However, these methods only apply quantization to the weight during fine-tuning to reduce memory consumption.”
— RoLoRA: Fine-tuning Rotated Outlier-free LLMs for Effective Weight-Activation Quantization
“Although QLoRA substantially reduces GPU memory and fine-tuning time while maintaining strong performance at the 4-bit level, it suffers from considerable performance degradation below 4 bits.”
— Enhancing Ultra-Low-Bit Quantization of Large Language Models Through Saliency-Aware Partial Retraining

Beaten on benchmarks

Head-to-head results where a newer method reports beating QLoRA. Values are copied from the source paper's tables — verify against the cited paper.

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.