GIFT-SW: Gaussian noise Injected Fine-Tuning of Salient Weights for LLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhelnin, Maxim, Moskvoretskii, Viktor, Shvetsov, Egor, Venediktov, Egor, Krylova, Mariya, Zuev, Aleksandr, Burnaev, Evgeny
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912004408606720
author Zhelnin, Maxim
Moskvoretskii, Viktor
Shvetsov, Egor
Venediktov, Egor
Krylova, Mariya
Zuev, Aleksandr
Burnaev, Evgeny
author_facet Zhelnin, Maxim
Moskvoretskii, Viktor
Shvetsov, Egor
Venediktov, Egor
Krylova, Mariya
Zuev, Aleksandr
Burnaev, Evgeny
contents Parameter Efficient Fine-Tuning (PEFT) methods have gained popularity and democratized the usage of Large Language Models (LLMs). Recent studies have shown that a small subset of weights significantly impacts performance. Based on this observation, we introduce a novel PEFT method, called Gaussian noise Injected Fine Tuning of Salient Weights (GIFT-SW). Our method updates only salient columns, while injecting Gaussian noise into non-salient ones. To identify these columns, we developeda generalized sensitivity metric that extends and unifies metrics from previous studies. Experiments with LLaMA models demonstrate that GIFT-SW outperforms full fine-tuning and modern PEFT methods under the same computational budget. Moreover, GIFT-SW offers practical advantages to recover performance of models subjected to mixed-precision quantization with keeping salient weights in full precision.
format Preprint
id arxiv_https___arxiv_org_abs_2408_15300
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle GIFT-SW: Gaussian noise Injected Fine-Tuning of Salient Weights for LLMs
Zhelnin, Maxim
Moskvoretskii, Viktor
Shvetsov, Egor
Venediktov, Egor
Krylova, Mariya
Zuev, Aleksandr
Burnaev, Evgeny
Machine Learning
Artificial Intelligence
Parameter Efficient Fine-Tuning (PEFT) methods have gained popularity and democratized the usage of Large Language Models (LLMs). Recent studies have shown that a small subset of weights significantly impacts performance. Based on this observation, we introduce a novel PEFT method, called Gaussian noise Injected Fine Tuning of Salient Weights (GIFT-SW). Our method updates only salient columns, while injecting Gaussian noise into non-salient ones. To identify these columns, we developeda generalized sensitivity metric that extends and unifies metrics from previous studies. Experiments with LLaMA models demonstrate that GIFT-SW outperforms full fine-tuning and modern PEFT methods under the same computational budget. Moreover, GIFT-SW offers practical advantages to recover performance of models subjected to mixed-precision quantization with keeping salient weights in full precision.
title GIFT-SW: Gaussian noise Injected Fine-Tuning of Salient Weights for LLMs
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2408.15300