Self-calibration for Language Model Quantization and Pruning

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Williams, Miles, Chrysostomou, George, Aletras, Nikolaos
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908446919491584
author Williams, Miles
Chrysostomou, George
Aletras, Nikolaos
author_facet Williams, Miles
Chrysostomou, George
Aletras, Nikolaos
contents Quantization and pruning are fundamental approaches for model compression, enabling efficient inference for language models. In a post-training setting, state-of-the-art quantization and pruning methods require calibration data, a small set of unlabeled examples. Conventionally, this is randomly sampled web text, aiming to reflect the model training data. However, this poses two key problems: (1) unrepresentative calibration examples can harm model performance, and (2) organizations increasingly avoid releasing model training data. In this paper, we propose self-calibration as a solution. Our approach requires no external data, instead leveraging the model itself to generate synthetic calibration data, with a view to better approximating the pre-training data distribution. We extensively compare the performance of self-calibration with several baselines, across a variety of models, compression methods, and tasks. Our approach proves consistently competitive in maximizing downstream task performance, frequently outperforming even using real data.
format Preprint
id arxiv_https___arxiv_org_abs_2410_17170
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Self-calibration for Language Model Quantization and Pruning
Williams, Miles
Chrysostomou, George
Aletras, Nikolaos
Computation and Language
Quantization and pruning are fundamental approaches for model compression, enabling efficient inference for language models. In a post-training setting, state-of-the-art quantization and pruning methods require calibration data, a small set of unlabeled examples. Conventionally, this is randomly sampled web text, aiming to reflect the model training data. However, this poses two key problems: (1) unrepresentative calibration examples can harm model performance, and (2) organizations increasingly avoid releasing model training data. In this paper, we propose self-calibration as a solution. Our approach requires no external data, instead leveraging the model itself to generate synthetic calibration data, with a view to better approximating the pre-training data distribution. We extensively compare the performance of self-calibration with several baselines, across a variety of models, compression methods, and tasks. Our approach proves consistently competitive in maximizing downstream task performance, frequently outperforming even using real data.
title Self-calibration for Language Model Quantization and Pruning
topic Computation and Language
url https://arxiv.org/abs/2410.17170